2023-04-27 12:19:53,962 INFO [train.py:976] (3/8) Training started 2023-04-27 12:19:53,962 INFO [train.py:986] (3/8) Device: cuda:3 2023-04-27 12:19:53,964 INFO [train.py:995] (3/8) {'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 12:19:53,964 INFO [train.py:997] (3/8) About to create model 2023-04-27 12:19:54,654 INFO [zipformer.py:178] (3/8) 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 12:19:54,669 INFO [train.py:1001] (3/8) Number of model parameters: 70369391 2023-04-27 12:19:57,221 INFO [train.py:1016] (3/8) Using DDP 2023-04-27 12:19:58,267 INFO [multidataset.py:46] (3/8) About to get multidataset train cuts 2023-04-27 12:19:58,268 INFO [multidataset.py:49] (3/8) Loading LibriSpeech in lazy mode 2023-04-27 12:19:58,288 INFO [multidataset.py:65] (3/8) Loading GigaSpeech 1998 splits in lazy mode 2023-04-27 12:20:00,752 INFO [multidataset.py:72] (3/8) Loading CommonVoice in lazy mode 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:230] (3/8) Enable MUSAN 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:231] (3/8) About to get Musan cuts 2023-04-27 12:20:03,028 INFO [asr_datamodule.py:255] (3/8) Enable SpecAugment 2023-04-27 12:20:03,028 INFO [asr_datamodule.py:256] (3/8) Time warp factor: 80 2023-04-27 12:20:03,028 INFO [asr_datamodule.py:266] (3/8) Num frame mask: 10 2023-04-27 12:20:03,029 INFO [asr_datamodule.py:279] (3/8) About to create train dataset 2023-04-27 12:20:03,029 INFO [asr_datamodule.py:306] (3/8) Using DynamicBucketingSampler. 2023-04-27 12:20:07,523 INFO [asr_datamodule.py:321] (3/8) About to create train dataloader 2023-04-27 12:20:07,524 INFO [asr_datamodule.py:435] (3/8) About to get dev-clean cuts 2023-04-27 12:20:07,525 INFO [asr_datamodule.py:442] (3/8) About to get dev-other cuts 2023-04-27 12:20:07,526 INFO [asr_datamodule.py:352] (3/8) About to create dev dataset 2023-04-27 12:20:07,764 INFO [asr_datamodule.py:369] (3/8) About to create dev dataloader 2023-04-27 12:20:25,619 INFO [train.py:904] (3/8) Epoch 1, batch 0, loss[loss=7.376, simple_loss=6.677, pruned_loss=6.976, over 16811.00 frames. ], tot_loss[loss=7.376, simple_loss=6.677, pruned_loss=6.976, over 16811.00 frames. ], batch size: 39, lr: 2.50e-02, grad_scale: 2.0 2023-04-27 12:20:25,620 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 12:20:32,881 INFO [train.py:938] (3/8) Epoch 1, validation: loss=6.911, simple_loss=6.238, pruned_loss=6.721, over 944034.00 frames. 2023-04-27 12:20:32,881 INFO [train.py:939] (3/8) Maximum memory allocated so far is 12902MB 2023-04-27 12:20:36,293 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:20:52,226 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:20:55,573 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=86.28 vs. limit=5.0 2023-04-27 12:21:03,471 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=7.17 vs. limit=2.0 2023-04-27 12:21:17,133 INFO [train.py:904] (3/8) Epoch 1, batch 50, loss[loss=1.213, simple_loss=1.075, pruned_loss=1.237, over 16893.00 frames. ], tot_loss[loss=2.182, simple_loss=1.977, pruned_loss=1.967, over 749851.51 frames. ], batch size: 116, lr: 2.75e-02, grad_scale: 2.0 2023-04-27 12:21:21,665 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=7.90 vs. limit=2.0 2023-04-27 12:21:23,264 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=109.68 vs. limit=5.0 2023-04-27 12:21:27,141 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=16.07 vs. limit=2.0 2023-04-27 12:21:46,499 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:22:02,559 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=9.29 vs. limit=2.0 2023-04-27 12:22:02,791 WARNING [train.py:894] (3/8) Grad scale is small: 0.001953125 2023-04-27 12:22:02,791 INFO [train.py:904] (3/8) Epoch 1, batch 100, loss[loss=1.215, simple_loss=1.033, pruned_loss=1.433, over 16671.00 frames. ], tot_loss[loss=1.638, simple_loss=1.458, pruned_loss=1.615, over 1326119.16 frames. ], batch size: 57, lr: 3.00e-02, grad_scale: 0.00390625 2023-04-27 12:22:05,007 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=89.20 vs. limit=5.0 2023-04-27 12:22:05,137 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=50.73 vs. limit=2.0 2023-04-27 12:22:13,669 INFO [optim.py:368] (3/8) 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 12:22:20,963 WARNING [optim.py:388] (3/8) Scaling gradients by 0.0112030990421772, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:21,110 INFO [optim.py:450] (3/8) 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 12:22:37,260 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=2.44 vs. limit=2.0 2023-04-27 12:22:43,720 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:22:44,872 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=20.20 vs. limit=2.0 2023-04-27 12:22:46,127 WARNING [optim.py:388] (3/8) Scaling gradients by 0.0022801109589636326, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:46,235 INFO [optim.py:450] (3/8) 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] (3/8) Scaling gradients by 0.04246773198246956, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:49,754 INFO [optim.py:450] (3/8) 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] (3/8) Scaling gradients by 0.000716241542249918, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:51,454 INFO [optim.py:450] (3/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.93, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.876e+12, grad_sumsq = 4.542e+13, orig_rms_sq=4.131e-02 2023-04-27 12:22:52,994 INFO [train.py:904] (3/8) Epoch 1, batch 150, loss[loss=0.9798, simple_loss=0.8325, pruned_loss=1.064, over 16702.00 frames. ], tot_loss[loss=1.401, simple_loss=1.229, pruned_loss=1.448, over 1764639.43 frames. ], batch size: 134, lr: 3.25e-02, grad_scale: 0.00390625 2023-04-27 12:22:53,729 WARNING [optim.py:388] (3/8) Scaling gradients by 0.049951765686273575, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:53,836 INFO [optim.py:450] (3/8) 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] (3/8) Scaling gradients by 0.00609818659722805, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:58,647 INFO [optim.py:450] (3/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.49, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.379e+10, grad_sumsq = 3.140e+11, orig_rms_sq=4.392e-02 2023-04-27 12:23:16,873 WARNING [optim.py:388] (3/8) Scaling gradients by 0.059935860335826874, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:16,977 INFO [optim.py:450] (3/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.63, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=1.808e+08, grad_sumsq = 3.916e+09, orig_rms_sq=4.617e-02 2023-04-27 12:23:28,342 WARNING [optim.py:388] (3/8) Scaling gradients by 0.060559310019016266, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:28,481 INFO [optim.py:450] (3/8) 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 12:23:32,130 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=9.68 vs. limit=2.0 2023-04-27 12:23:40,695 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=8.54 vs. limit=2.0 2023-04-27 12:23:42,816 WARNING [train.py:894] (3/8) Grad scale is small: 0.00390625 2023-04-27 12:23:42,816 INFO [train.py:904] (3/8) Epoch 1, batch 200, loss[loss=1.045, simple_loss=0.8763, pruned_loss=1.118, over 17039.00 frames. ], tot_loss[loss=1.264, simple_loss=1.097, pruned_loss=1.321, over 2101213.19 frames. ], batch size: 55, lr: 3.50e-02, grad_scale: 0.0078125 2023-04-27 12:23:51,003 INFO [optim.py:368] (3/8) 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] (3/8) Scaling gradients by 0.002041660714894533, model_norm_threshold=541.4743041992188 2023-04-27 12:23:51,112 INFO [optim.py:450] (3/8) 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] (3/8) Scaling gradients by 0.02974529005587101, model_norm_threshold=541.4743041992188 2023-04-27 12:24:00,688 INFO [optim.py:450] (3/8) 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] (3/8) Scaling gradients by 0.01955481991171837, model_norm_threshold=541.4743041992188 2023-04-27 12:24:01,588 INFO [optim.py:450] (3/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.84, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=6.426e+08, grad_sumsq = 1.649e+10, orig_rms_sq=3.897e-02 2023-04-27 12:24:30,805 INFO [train.py:904] (3/8) Epoch 1, batch 250, loss[loss=0.8888, simple_loss=0.7414, pruned_loss=0.9106, over 15998.00 frames. ], tot_loss[loss=1.159, simple_loss=0.9987, pruned_loss=1.203, over 2370398.41 frames. ], batch size: 35, lr: 3.75e-02, grad_scale: 0.0078125 2023-04-27 12:24:33,594 WARNING [optim.py:388] (3/8) Scaling gradients by 0.057925041764974594, model_norm_threshold=541.4743041992188 2023-04-27 12:24:33,736 INFO [optim.py:450] (3/8) 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 12:25:16,077 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:25:20,573 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:25:21,111 WARNING [train.py:894] (3/8) Grad scale is small: 0.0078125 2023-04-27 12:25:21,112 INFO [train.py:904] (3/8) Epoch 1, batch 300, loss[loss=0.876, simple_loss=0.7373, pruned_loss=0.8249, over 16434.00 frames. ], tot_loss[loss=1.091, simple_loss=0.9335, pruned_loss=1.118, over 2592594.94 frames. ], batch size: 146, lr: 4.00e-02, grad_scale: 0.015625 2023-04-27 12:25:30,077 INFO [optim.py:368] (3/8) 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 12:25:52,563 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=4.00 vs. limit=2.0 2023-04-27 12:26:12,599 INFO [train.py:904] (3/8) Epoch 1, batch 350, loss[loss=0.9417, simple_loss=0.7768, pruned_loss=0.9035, over 17191.00 frames. ], tot_loss[loss=1.047, simple_loss=0.8896, pruned_loss=1.057, over 2758613.64 frames. ], batch size: 46, lr: 4.25e-02, grad_scale: 0.015625 2023-04-27 12:26:18,648 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:26:25,717 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=3.96 vs. limit=2.0 2023-04-27 12:26:52,672 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:27:06,004 INFO [train.py:904] (3/8) Epoch 1, batch 400, loss[loss=0.8898, simple_loss=0.7249, pruned_loss=0.8473, over 16876.00 frames. ], tot_loss[loss=1.008, simple_loss=0.8507, pruned_loss=1.003, over 2876241.70 frames. ], batch size: 42, lr: 4.50e-02, grad_scale: 0.03125 2023-04-27 12:27:16,010 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=13.73 vs. limit=2.0 2023-04-27 12:27:17,878 INFO [optim.py:368] (3/8) 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 12:27:18,423 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=11.53 vs. limit=2.0 2023-04-27 12:27:46,059 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:27:56,255 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:27:59,545 INFO [train.py:904] (3/8) Epoch 1, batch 450, loss[loss=0.9108, simple_loss=0.7447, pruned_loss=0.8254, over 16888.00 frames. ], tot_loss[loss=0.9811, simple_loss=0.8219, pruned_loss=0.9601, over 2971936.78 frames. ], batch size: 116, lr: 4.75e-02, grad_scale: 0.03125 2023-04-27 12:28:18,101 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8307, 3.8610, 3.8423, 3.8563, 3.8053, 3.8527, 3.8409, 3.8227], device='cuda:3'), covar=tensor([0.0092, 0.0086, 0.0070, 0.0091, 0.0088, 0.0100, 0.0080, 0.0135], device='cuda:3'), in_proj_covar=tensor([0.0009, 0.0008, 0.0009, 0.0009, 0.0009, 0.0009, 0.0008, 0.0009], device='cuda:3'), out_proj_covar=tensor([8.4822e-06, 8.7141e-06, 8.5400e-06, 8.4388e-06, 8.6427e-06, 8.5042e-06, 8.6032e-06, 8.7001e-06], device='cuda:3') 2023-04-27 12:28:51,183 INFO [train.py:904] (3/8) Epoch 1, batch 500, loss[loss=0.8872, simple_loss=0.7138, pruned_loss=0.8074, over 17182.00 frames. ], tot_loss[loss=0.9663, simple_loss=0.8036, pruned_loss=0.9298, over 3043202.85 frames. ], batch size: 46, lr: 4.99e-02, grad_scale: 0.0625 2023-04-27 12:28:56,856 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.8549, 5.8408, 5.8687, 5.8651, 5.7752, 5.8747, 5.6449, 5.8621], device='cuda:3'), covar=tensor([0.0120, 0.0079, 0.0121, 0.0129, 0.0147, 0.0115, 0.0145, 0.0107], device='cuda:3'), in_proj_covar=tensor([0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008], device='cuda:3'), out_proj_covar=tensor([8.0127e-06, 8.1371e-06, 7.8675e-06, 8.0769e-06, 7.9874e-06, 8.1565e-06, 8.1890e-06, 8.1044e-06], device='cuda:3') 2023-04-27 12:29:01,368 INFO [optim.py:368] (3/8) 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 12:29:04,662 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=3.00 vs. limit=2.0 2023-04-27 12:29:44,933 INFO [train.py:904] (3/8) Epoch 1, batch 550, loss[loss=0.9028, simple_loss=0.729, pruned_loss=0.7881, over 16416.00 frames. ], tot_loss[loss=0.954, simple_loss=0.7881, pruned_loss=0.9018, over 3104036.03 frames. ], batch size: 165, lr: 4.98e-02, grad_scale: 0.0625 2023-04-27 12:29:58,166 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:03,891 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:08,905 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=2.26 vs. limit=2.0 2023-04-27 12:30:12,925 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=3.47 vs. limit=2.0 2023-04-27 12:30:26,898 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:38,300 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:30:38,745 INFO [train.py:904] (3/8) Epoch 1, batch 600, loss[loss=0.9657, simple_loss=0.7776, pruned_loss=0.823, over 17257.00 frames. ], tot_loss[loss=0.9417, simple_loss=0.7733, pruned_loss=0.8734, over 3159637.62 frames. ], batch size: 52, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:30:43,426 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=4.64 vs. limit=2.0 2023-04-27 12:30:48,386 INFO [optim.py:368] (3/8) 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 12:31:02,541 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:31:07,557 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:31:28,002 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:31:30,996 INFO [train.py:904] (3/8) Epoch 1, batch 650, loss[loss=0.8518, simple_loss=0.6872, pruned_loss=0.7033, over 16494.00 frames. ], tot_loss[loss=0.9267, simple_loss=0.7578, pruned_loss=0.8418, over 3193970.51 frames. ], batch size: 68, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:31:31,343 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:31:32,143 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:32:22,394 INFO [train.py:904] (3/8) Epoch 1, batch 700, loss[loss=0.9504, simple_loss=0.7697, pruned_loss=0.7595, over 16837.00 frames. ], tot_loss[loss=0.9164, simple_loss=0.7482, pruned_loss=0.8112, over 3231878.09 frames. ], batch size: 57, lr: 4.98e-02, grad_scale: 0.25 2023-04-27 12:32:31,774 INFO [optim.py:368] (3/8) 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 12:33:00,829 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:33:05,904 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:33:14,006 INFO [train.py:904] (3/8) Epoch 1, batch 750, loss[loss=0.8539, simple_loss=0.7056, pruned_loss=0.6412, over 16678.00 frames. ], tot_loss[loss=0.8995, simple_loss=0.7354, pruned_loss=0.7739, over 3243516.64 frames. ], batch size: 57, lr: 4.97e-02, grad_scale: 0.25 2023-04-27 12:33:51,297 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:34:05,995 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 12:34:06,256 INFO [train.py:904] (3/8) Epoch 1, batch 800, loss[loss=0.8201, simple_loss=0.677, pruned_loss=0.6052, over 12268.00 frames. ], tot_loss[loss=0.8732, simple_loss=0.7173, pruned_loss=0.7271, over 3263833.07 frames. ], batch size: 247, lr: 4.97e-02, grad_scale: 0.5 2023-04-27 12:34:07,458 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([6.1805, 5.8651, 6.1702, 6.1789, 6.1785, 6.1783, 6.1799, 6.1789], device='cuda:3'), covar=tensor([0.0146, 0.1715, 0.0278, 0.0150, 0.0117, 0.0154, 0.0223, 0.0138], device='cuda:3'), in_proj_covar=tensor([0.0016, 0.0018, 0.0018, 0.0016, 0.0015, 0.0016, 0.0018, 0.0016], device='cuda:3'), out_proj_covar=tensor([1.5442e-05, 1.7554e-05, 1.6711e-05, 1.4865e-05, 1.5181e-05, 1.4975e-05, 1.6767e-05, 1.5107e-05], device='cuda:3') 2023-04-27 12:34:17,530 INFO [optim.py:368] (3/8) 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 12:34:54,099 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:34:58,586 INFO [train.py:904] (3/8) Epoch 1, batch 850, loss[loss=0.7077, simple_loss=0.6026, pruned_loss=0.483, over 16291.00 frames. ], tot_loss[loss=0.8406, simple_loss=0.6952, pruned_loss=0.6768, over 3270728.30 frames. ], batch size: 164, lr: 4.96e-02, grad_scale: 0.5 2023-04-27 12:35:50,488 INFO [train.py:904] (3/8) Epoch 1, batch 900, loss[loss=0.6908, simple_loss=0.5874, pruned_loss=0.4663, over 16849.00 frames. ], tot_loss[loss=0.8052, simple_loss=0.6715, pruned_loss=0.6258, over 3286127.98 frames. ], batch size: 42, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:35:57,618 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:00,422 INFO [optim.py:368] (3/8) 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 12:36:08,843 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:36:14,460 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:36:37,791 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 950, loss[loss=0.6498, simple_loss=0.5779, pruned_loss=0.3965, over 17169.00 frames. ], tot_loss[loss=0.7735, simple_loss=0.6507, pruned_loss=0.5806, over 3290889.09 frames. ], batch size: 46, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:36:44,110 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:37:17,404 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1229, 4.1947, 4.0309, 3.8631, 5.1024, 4.7724, 3.7364, 3.8962], device='cuda:3'), covar=tensor([0.4956, 0.5362, 0.5559, 0.4870, 0.1383, 0.2969, 0.5026, 1.3293], device='cuda:3'), in_proj_covar=tensor([0.0033, 0.0037, 0.0038, 0.0026, 0.0031, 0.0035, 0.0024, 0.0030], device='cuda:3'), out_proj_covar=tensor([2.4909e-05, 2.6430e-05, 2.8312e-05, 2.0642e-05, 2.3958e-05, 2.6969e-05, 1.8612e-05, 2.4052e-05], device='cuda:3') 2023-04-27 12:37:20,301 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=4.05 vs. limit=2.0 2023-04-27 12:37:35,472 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:37:35,994 INFO [train.py:904] (3/8) Epoch 1, batch 1000, loss[loss=0.6114, simple_loss=0.5361, pruned_loss=0.3806, over 16931.00 frames. ], tot_loss[loss=0.7405, simple_loss=0.6287, pruned_loss=0.5372, over 3303941.02 frames. ], batch size: 109, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:37:46,287 INFO [optim.py:368] (3/8) 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 12:38:21,603 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 1050, loss[loss=0.5656, simple_loss=0.5065, pruned_loss=0.3354, over 16850.00 frames. ], tot_loss[loss=0.7096, simple_loss=0.6083, pruned_loss=0.4977, over 3310331.09 frames. ], batch size: 90, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:38:45,002 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-04-27 12:39:12,046 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:39:13,693 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 12:39:21,796 INFO [train.py:904] (3/8) Epoch 1, batch 1100, loss[loss=0.5678, simple_loss=0.5097, pruned_loss=0.3333, over 16502.00 frames. ], tot_loss[loss=0.6802, simple_loss=0.5891, pruned_loss=0.4615, over 3316694.08 frames. ], batch size: 75, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:39:33,202 INFO [optim.py:368] (3/8) 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 12:39:54,365 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 2023-04-27 12:39:57,600 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1005, 3.7623, 3.0107, 2.9956, 2.8162, 3.1650, 2.9292, 3.3906], device='cuda:3'), covar=tensor([0.6520, 0.3033, 0.3273, 0.4027, 0.5105, 0.4247, 0.5180, 0.2967], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0081, 0.0087, 0.0091, 0.0088, 0.0089, 0.0090, 0.0075], device='cuda:3'), out_proj_covar=tensor([7.3248e-05, 7.7114e-05, 7.6255e-05, 7.9818e-05, 7.6116e-05, 7.8970e-05, 7.4697e-05, 6.5096e-05], device='cuda:3') 2023-04-27 12:40:17,543 INFO [train.py:904] (3/8) Epoch 1, batch 1150, loss[loss=0.5742, simple_loss=0.5272, pruned_loss=0.3214, over 17273.00 frames. ], tot_loss[loss=0.6526, simple_loss=0.5712, pruned_loss=0.4288, over 3321478.58 frames. ], batch size: 52, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:40:19,514 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:41:03,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([6.0455, 5.7106, 5.8853, 6.0699, 6.0647, 6.0469, 6.0690, 6.0383], device='cuda:3'), covar=tensor([0.0183, 0.1067, 0.0373, 0.0153, 0.0177, 0.0167, 0.0169, 0.0166], device='cuda:3'), in_proj_covar=tensor([0.0049, 0.0074, 0.0064, 0.0050, 0.0048, 0.0050, 0.0053, 0.0049], device='cuda:3'), out_proj_covar=tensor([4.0181e-05, 6.1752e-05, 5.2138e-05, 3.7509e-05, 3.9750e-05, 3.8124e-05, 4.4503e-05, 3.8620e-05], device='cuda:3') 2023-04-27 12:41:10,398 INFO [train.py:904] (3/8) Epoch 1, batch 1200, loss[loss=0.5632, simple_loss=0.4875, pruned_loss=0.3466, over 16290.00 frames. ], tot_loss[loss=0.6301, simple_loss=0.5572, pruned_loss=0.4013, over 3329575.81 frames. ], batch size: 165, lr: 4.93e-02, grad_scale: 2.0 2023-04-27 12:41:11,983 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-27 12:41:12,659 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:41:14,331 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=6.14 vs. limit=5.0 2023-04-27 12:41:21,557 INFO [optim.py:368] (3/8) 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 12:41:25,744 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:41:29,944 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:41:33,285 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-27 12:41:34,762 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:41:57,148 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:03,008 INFO [train.py:904] (3/8) Epoch 1, batch 1250, loss[loss=0.5173, simple_loss=0.4728, pruned_loss=0.2901, over 16817.00 frames. ], tot_loss[loss=0.6112, simple_loss=0.5448, pruned_loss=0.3795, over 3331403.17 frames. ], batch size: 83, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:42:14,149 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([6.6760, 6.3781, 6.4287, 6.7099, 6.6952, 6.6833, 6.7037, 6.6725], device='cuda:3'), covar=tensor([0.0214, 0.1024, 0.0643, 0.0181, 0.0207, 0.0215, 0.0235, 0.0193], device='cuda:3'), in_proj_covar=tensor([0.0053, 0.0079, 0.0069, 0.0052, 0.0052, 0.0055, 0.0057, 0.0052], device='cuda:3'), out_proj_covar=tensor([4.2953e-05, 6.6407e-05, 5.6490e-05, 3.8627e-05, 4.1874e-05, 4.1024e-05, 4.6785e-05, 4.0401e-05], device='cuda:3') 2023-04-27 12:42:17,895 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:24,719 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:49,321 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:56,587 INFO [train.py:904] (3/8) Epoch 1, batch 1300, loss[loss=0.5609, simple_loss=0.5042, pruned_loss=0.322, over 16900.00 frames. ], tot_loss[loss=0.5957, simple_loss=0.5347, pruned_loss=0.3616, over 3332877.17 frames. ], batch size: 109, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:43:07,776 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 1, batch 1350, loss[loss=0.5347, simple_loss=0.4889, pruned_loss=0.298, over 16830.00 frames. ], tot_loss[loss=0.5777, simple_loss=0.5237, pruned_loss=0.3419, over 3326327.83 frames. ], batch size: 96, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:44:49,662 INFO [train.py:904] (3/8) Epoch 1, batch 1400, loss[loss=0.4775, simple_loss=0.4631, pruned_loss=0.2417, over 16844.00 frames. ], tot_loss[loss=0.5623, simple_loss=0.5144, pruned_loss=0.3255, over 3326258.64 frames. ], batch size: 42, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:45:00,101 INFO [optim.py:368] (3/8) 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 12:45:27,766 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:45:40,955 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-27 12:45:46,438 INFO [train.py:904] (3/8) Epoch 1, batch 1450, loss[loss=0.4937, simple_loss=0.4653, pruned_loss=0.2619, over 16882.00 frames. ], tot_loss[loss=0.5469, simple_loss=0.5045, pruned_loss=0.3102, over 3325326.16 frames. ], batch size: 116, lr: 4.90e-02, grad_scale: 2.0 2023-04-27 12:46:13,256 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6274, 4.5009, 4.5683, 4.9348, 4.6751, 4.7101, 4.8772, 4.7364], device='cuda:3'), covar=tensor([0.0688, 0.1585, 0.1215, 0.0378, 0.0728, 0.0579, 0.0438, 0.0479], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0092, 0.0090, 0.0069, 0.0077, 0.0077, 0.0075, 0.0072], device='cuda:3'), out_proj_covar=tensor([5.5927e-05, 8.0856e-05, 7.3518e-05, 4.6762e-05, 5.6320e-05, 5.4499e-05, 5.7509e-05, 5.2598e-05], device='cuda:3') 2023-04-27 12:46:24,574 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 12:46:25,983 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8240, 3.5820, 3.5628, 3.7219, 3.3409, 3.7388, 3.6407, 3.3581], device='cuda:3'), covar=tensor([0.0935, 0.0767, 0.1189, 0.0638, 0.1128, 0.0925, 0.0896, 0.0912], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0060, 0.0075, 0.0061, 0.0067, 0.0067, 0.0059, 0.0063], device='cuda:3'), out_proj_covar=tensor([6.1588e-05, 4.9835e-05, 7.1898e-05, 5.7463e-05, 6.2538e-05, 5.9665e-05, 5.4619e-05, 5.7177e-05], device='cuda:3') 2023-04-27 12:46:34,552 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:46:41,351 INFO [train.py:904] (3/8) Epoch 1, batch 1500, loss[loss=0.5328, simple_loss=0.4951, pruned_loss=0.2882, over 16252.00 frames. ], tot_loss[loss=0.5345, simple_loss=0.4967, pruned_loss=0.298, over 3310810.10 frames. ], batch size: 165, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:46:44,282 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:46:51,035 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:46:52,704 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.422e+02 5.181e+02 6.478e+02 8.944e+02 1.260e+03, threshold=1.296e+03, percent-clipped=1.0 2023-04-27 12:46:57,640 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1900, 4.1202, 3.9909, 3.8771, 3.9326, 3.2951, 3.9074, 2.5289], device='cuda:3'), covar=tensor([0.0817, 0.0969, 0.1453, 0.0630, 0.1420, 0.2192, 0.1440, 0.2335], device='cuda:3'), in_proj_covar=tensor([0.0031, 0.0024, 0.0028, 0.0028, 0.0021, 0.0027, 0.0027, 0.0023], device='cuda:3'), out_proj_covar=tensor([2.5963e-05, 1.9222e-05, 2.3585e-05, 2.1771e-05, 1.7403e-05, 2.2657e-05, 2.1145e-05, 2.0746e-05], device='cuda:3') 2023-04-27 12:47:38,771 INFO [train.py:904] (3/8) Epoch 1, batch 1550, loss[loss=0.4652, simple_loss=0.4687, pruned_loss=0.2233, over 17133.00 frames. ], tot_loss[loss=0.524, simple_loss=0.4904, pruned_loss=0.2876, over 3319671.12 frames. ], batch size: 47, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:47:39,034 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:48:35,336 INFO [train.py:904] (3/8) Epoch 1, batch 1600, loss[loss=0.4784, simple_loss=0.4686, pruned_loss=0.2405, over 17189.00 frames. ], tot_loss[loss=0.5182, simple_loss=0.4878, pruned_loss=0.2808, over 3321629.96 frames. ], batch size: 46, lr: 4.88e-02, grad_scale: 4.0 2023-04-27 12:48:47,164 INFO [optim.py:368] (3/8) 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 12:48:47,448 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:49:05,618 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:49:32,287 INFO [train.py:904] (3/8) Epoch 1, batch 1650, loss[loss=0.4928, simple_loss=0.4708, pruned_loss=0.2567, over 16721.00 frames. ], tot_loss[loss=0.5109, simple_loss=0.4847, pruned_loss=0.2729, over 3329779.36 frames. ], batch size: 83, lr: 4.87e-02, grad_scale: 4.0 2023-04-27 12:49:57,883 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:50:29,818 INFO [train.py:904] (3/8) Epoch 1, batch 1700, loss[loss=0.4447, simple_loss=0.4607, pruned_loss=0.2079, over 17119.00 frames. ], tot_loss[loss=0.5048, simple_loss=0.4823, pruned_loss=0.2664, over 3331987.46 frames. ], batch size: 48, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:50:41,826 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.189e+02 5.218e+02 6.714e+02 8.049e+02 1.427e+03, threshold=1.343e+03, percent-clipped=1.0 2023-04-27 12:51:28,762 INFO [train.py:904] (3/8) Epoch 1, batch 1750, loss[loss=0.505, simple_loss=0.4802, pruned_loss=0.2647, over 16864.00 frames. ], tot_loss[loss=0.4964, simple_loss=0.4779, pruned_loss=0.2589, over 3329685.08 frames. ], batch size: 116, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:52:15,499 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:52:27,927 INFO [train.py:904] (3/8) Epoch 1, batch 1800, loss[loss=0.4339, simple_loss=0.4482, pruned_loss=0.206, over 17111.00 frames. ], tot_loss[loss=0.493, simple_loss=0.4773, pruned_loss=0.255, over 3308155.86 frames. ], batch size: 47, lr: 4.85e-02, grad_scale: 4.0 2023-04-27 12:52:37,345 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:53:10,956 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 1850, loss[loss=0.4601, simple_loss=0.4583, pruned_loss=0.2292, over 16851.00 frames. ], tot_loss[loss=0.4858, simple_loss=0.4737, pruned_loss=0.2489, over 3302431.62 frames. ], batch size: 96, lr: 4.84e-02, grad_scale: 4.0 2023-04-27 12:53:32,837 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:32,925 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:48,604 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:51,779 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8169, 4.7342, 4.9956, 4.8705, 4.4328, 4.8290, 4.4059, 4.8220], device='cuda:3'), covar=tensor([0.0498, 0.0592, 0.0308, 0.0471, 0.1492, 0.0521, 0.0915, 0.1810], device='cuda:3'), in_proj_covar=tensor([0.0034, 0.0035, 0.0035, 0.0027, 0.0035, 0.0034, 0.0031, 0.0030], device='cuda:3'), out_proj_covar=tensor([2.5465e-05, 2.6523e-05, 2.5930e-05, 2.2666e-05, 2.9478e-05, 2.5249e-05, 2.6253e-05, 2.6042e-05], device='cuda:3') 2023-04-27 12:54:02,341 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5657, 4.5413, 4.3645, 4.8091, 4.6780, 4.8347, 4.6772, 4.6478], device='cuda:3'), covar=tensor([0.0295, 0.0315, 0.0518, 0.0368, 0.0462, 0.0260, 0.0301, 0.0398], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0090, 0.0116, 0.0093, 0.0090, 0.0091, 0.0084, 0.0093], device='cuda:3'), out_proj_covar=tensor([8.4424e-05, 8.4284e-05, 1.2021e-04, 9.6133e-05, 8.7405e-05, 8.9669e-05, 8.1359e-05, 9.2313e-05], device='cuda:3') 2023-04-27 12:54:22,008 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:54:23,727 INFO [train.py:904] (3/8) Epoch 1, batch 1900, loss[loss=0.3715, simple_loss=0.3992, pruned_loss=0.1696, over 15932.00 frames. ], tot_loss[loss=0.4745, simple_loss=0.4677, pruned_loss=0.2402, over 3306793.11 frames. ], batch size: 35, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:54:36,241 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.249e+02 5.423e+02 6.685e+02 8.506e+02 1.861e+03, threshold=1.337e+03, percent-clipped=8.0 2023-04-27 12:54:44,599 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:54:49,106 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 12:55:00,098 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:55:22,815 INFO [train.py:904] (3/8) Epoch 1, batch 1950, loss[loss=0.3764, simple_loss=0.4117, pruned_loss=0.1693, over 17254.00 frames. ], tot_loss[loss=0.465, simple_loss=0.4631, pruned_loss=0.2328, over 3304031.41 frames. ], batch size: 45, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:55:42,108 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 2000, loss[loss=0.4013, simple_loss=0.4099, pruned_loss=0.1964, over 16802.00 frames. ], tot_loss[loss=0.4592, simple_loss=0.4601, pruned_loss=0.2285, over 3301693.01 frames. ], batch size: 102, lr: 4.82e-02, grad_scale: 8.0 2023-04-27 12:56:36,802 INFO [optim.py:368] (3/8) 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 12:56:58,912 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 12:57:27,604 INFO [train.py:904] (3/8) Epoch 1, batch 2050, loss[loss=0.416, simple_loss=0.4356, pruned_loss=0.1982, over 16811.00 frames. ], tot_loss[loss=0.447, simple_loss=0.4533, pruned_loss=0.2199, over 3313564.00 frames. ], batch size: 96, lr: 4.81e-02, grad_scale: 8.0 2023-04-27 12:58:05,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2096, 4.0928, 3.8743, 4.3169, 4.2237, 4.3721, 4.3182, 4.2729], device='cuda:3'), covar=tensor([0.0249, 0.0315, 0.0630, 0.0379, 0.0387, 0.0268, 0.0306, 0.0310], device='cuda:3'), in_proj_covar=tensor([0.0097, 0.0096, 0.0129, 0.0105, 0.0096, 0.0101, 0.0089, 0.0101], device='cuda:3'), out_proj_covar=tensor([9.2624e-05, 9.5196e-05, 1.3799e-04, 1.1253e-04, 1.0011e-04, 1.0136e-04, 8.8285e-05, 1.0585e-04], device='cuda:3') 2023-04-27 12:58:19,066 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:58:32,223 INFO [train.py:904] (3/8) Epoch 1, batch 2100, loss[loss=0.4282, simple_loss=0.4542, pruned_loss=0.2011, over 16477.00 frames. ], tot_loss[loss=0.4381, simple_loss=0.4486, pruned_loss=0.2134, over 3319942.04 frames. ], batch size: 68, lr: 4.80e-02, grad_scale: 16.0 2023-04-27 12:58:41,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9003, 4.9403, 4.5167, 5.0967, 4.9967, 5.1238, 5.1610, 4.8915], device='cuda:3'), covar=tensor([0.0265, 0.0302, 0.0633, 0.0417, 0.0371, 0.0258, 0.0231, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0097, 0.0132, 0.0105, 0.0097, 0.0102, 0.0089, 0.0102], device='cuda:3'), out_proj_covar=tensor([9.5006e-05, 9.7334e-05, 1.4073e-04, 1.1388e-04, 1.0177e-04, 1.0191e-04, 8.8630e-05, 1.0630e-04], device='cuda:3') 2023-04-27 12:58:45,932 INFO [optim.py:368] (3/8) 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 12:59:19,748 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 2150, loss[loss=0.4631, simple_loss=0.4652, pruned_loss=0.2304, over 15468.00 frames. ], tot_loss[loss=0.4323, simple_loss=0.4453, pruned_loss=0.2094, over 3323939.79 frames. ], batch size: 190, lr: 4.79e-02, grad_scale: 16.0 2023-04-27 12:59:39,943 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6979, 4.8854, 4.6839, 4.9658, 4.6297, 4.9739, 4.7051, 4.8490], device='cuda:3'), covar=tensor([0.0354, 0.0529, 0.0399, 0.0231, 0.0416, 0.0266, 0.0304, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0113, 0.0102, 0.0081, 0.0102, 0.0091, 0.0095, 0.0083], device='cuda:3'), out_proj_covar=tensor([7.8752e-05, 9.9651e-05, 8.5501e-05, 5.8242e-05, 7.9344e-05, 7.1070e-05, 7.5069e-05, 6.8001e-05], device='cuda:3') 2023-04-27 13:00:32,066 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:40,252 INFO [train.py:904] (3/8) Epoch 1, batch 2200, loss[loss=0.4227, simple_loss=0.4319, pruned_loss=0.2067, over 16682.00 frames. ], tot_loss[loss=0.426, simple_loss=0.4416, pruned_loss=0.205, over 3317802.36 frames. ], batch size: 89, lr: 4.78e-02, grad_scale: 16.0 2023-04-27 13:00:53,194 INFO [optim.py:368] (3/8) 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 13:00:56,683 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:07,600 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:01:11,873 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:35,450 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9912, 5.1197, 4.8157, 4.8821, 5.0237, 5.2013, 5.2847, 4.8268], device='cuda:3'), covar=tensor([0.0454, 0.0770, 0.0903, 0.1247, 0.1177, 0.0608, 0.0645, 0.1320], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0149, 0.0131, 0.0145, 0.0159, 0.0116, 0.0108, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:3') 2023-04-27 13:01:44,909 INFO [train.py:904] (3/8) Epoch 1, batch 2250, loss[loss=0.3297, simple_loss=0.3752, pruned_loss=0.1421, over 16863.00 frames. ], tot_loss[loss=0.4226, simple_loss=0.4401, pruned_loss=0.2024, over 3308906.18 frames. ], batch size: 42, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:01:56,177 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 13:02:05,718 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:02:08,817 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:02:19,048 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9690, 4.4990, 3.9359, 4.9484, 5.1195, 2.1526, 4.9135, 4.2696], device='cuda:3'), covar=tensor([0.0939, 0.0189, 0.0609, 0.0060, 0.0160, 0.2264, 0.0092, 0.0111], device='cuda:3'), in_proj_covar=tensor([0.0064, 0.0036, 0.0060, 0.0039, 0.0035, 0.0075, 0.0036, 0.0023], device='cuda:3'), out_proj_covar=tensor([6.2538e-05, 3.7536e-05, 5.5725e-05, 3.0210e-05, 3.6464e-05, 6.6080e-05, 3.0887e-05, 2.2550e-05], device='cuda:3') 2023-04-27 13:02:49,154 INFO [train.py:904] (3/8) Epoch 1, batch 2300, loss[loss=0.3894, simple_loss=0.4351, pruned_loss=0.1719, over 16609.00 frames. ], tot_loss[loss=0.4167, simple_loss=0.4365, pruned_loss=0.1983, over 3303055.81 frames. ], batch size: 57, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:03:01,865 INFO [optim.py:368] (3/8) 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 13:03:07,888 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:08,207 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.76 vs. limit=5.0 2023-04-27 13:03:39,615 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:03:53,139 INFO [train.py:904] (3/8) Epoch 1, batch 2350, loss[loss=0.3967, simple_loss=0.441, pruned_loss=0.1762, over 16744.00 frames. ], tot_loss[loss=0.4103, simple_loss=0.4326, pruned_loss=0.1939, over 3303444.05 frames. ], batch size: 57, lr: 4.76e-02, grad_scale: 16.0 2023-04-27 13:03:53,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8634, 2.9902, 2.8270, 2.6835, 2.8919, 2.7805, 2.8981, 2.7385], device='cuda:3'), covar=tensor([0.0309, 0.0287, 0.0335, 0.0326, 0.0351, 0.0412, 0.0367, 0.0346], device='cuda:3'), in_proj_covar=tensor([0.0051, 0.0049, 0.0045, 0.0048, 0.0049, 0.0053, 0.0056, 0.0050], device='cuda:3'), out_proj_covar=tensor([4.8612e-05, 4.7942e-05, 4.3723e-05, 4.5821e-05, 4.3802e-05, 5.4793e-05, 5.6873e-05, 4.8296e-05], device='cuda:3') 2023-04-27 13:04:54,412 INFO [train.py:904] (3/8) Epoch 1, batch 2400, loss[loss=0.3912, simple_loss=0.4293, pruned_loss=0.1766, over 16572.00 frames. ], tot_loss[loss=0.4087, simple_loss=0.4324, pruned_loss=0.1924, over 3303577.72 frames. ], batch size: 75, lr: 4.75e-02, grad_scale: 16.0 2023-04-27 13:04:55,632 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 13:05:07,296 INFO [optim.py:368] (3/8) 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 13:05:18,919 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4011, 5.6050, 5.3242, 5.7721, 5.1905, 5.3967, 5.4647, 5.6940], device='cuda:3'), covar=tensor([0.0306, 0.0490, 0.0325, 0.0165, 0.0434, 0.0246, 0.0248, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0130, 0.0113, 0.0091, 0.0115, 0.0102, 0.0107, 0.0092], device='cuda:3'), out_proj_covar=tensor([9.5869e-05, 1.1519e-04, 9.4755e-05, 6.7556e-05, 9.3424e-05, 8.1910e-05, 8.8648e-05, 7.9250e-05], device='cuda:3') 2023-04-27 13:05:40,893 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 13:05:55,891 INFO [train.py:904] (3/8) Epoch 1, batch 2450, loss[loss=0.4339, simple_loss=0.4446, pruned_loss=0.2116, over 16775.00 frames. ], tot_loss[loss=0.4045, simple_loss=0.431, pruned_loss=0.1889, over 3309198.04 frames. ], batch size: 124, lr: 4.74e-02, grad_scale: 16.0 2023-04-27 13:06:51,021 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:06:58,990 INFO [train.py:904] (3/8) Epoch 1, batch 2500, loss[loss=0.3813, simple_loss=0.4178, pruned_loss=0.1724, over 16836.00 frames. ], tot_loss[loss=0.4007, simple_loss=0.4291, pruned_loss=0.1862, over 3313791.17 frames. ], batch size: 42, lr: 4.73e-02, grad_scale: 16.0 2023-04-27 13:07:11,499 INFO [optim.py:368] (3/8) 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 13:07:14,752 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:29,231 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:50,933 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:01,505 INFO [train.py:904] (3/8) Epoch 1, batch 2550, loss[loss=0.374, simple_loss=0.4207, pruned_loss=0.1636, over 16143.00 frames. ], tot_loss[loss=0.3963, simple_loss=0.4254, pruned_loss=0.1835, over 3311937.38 frames. ], batch size: 35, lr: 4.72e-02, grad_scale: 16.0 2023-04-27 13:08:15,979 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:27,819 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5484, 3.9567, 3.6537, 3.0549, 3.1430, 3.6808, 3.8383, 3.4901], device='cuda:3'), covar=tensor([0.0131, 0.0138, 0.0188, 0.0271, 0.0252, 0.0183, 0.0180, 0.0255], device='cuda:3'), in_proj_covar=tensor([0.0033, 0.0035, 0.0039, 0.0040, 0.0036, 0.0039, 0.0039, 0.0036], device='cuda:3'), out_proj_covar=tensor([3.7628e-05, 3.6344e-05, 4.0344e-05, 3.8997e-05, 3.5717e-05, 4.0112e-05, 3.8089e-05, 3.8536e-05], device='cuda:3') 2023-04-27 13:08:32,315 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:50,368 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2341, 4.2714, 4.0533, 4.4592, 4.3849, 4.5316, 4.4366, 4.3233], device='cuda:3'), covar=tensor([0.0301, 0.0267, 0.0736, 0.0358, 0.0359, 0.0213, 0.0271, 0.0320], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0107, 0.0166, 0.0129, 0.0112, 0.0119, 0.0103, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2023-04-27 13:09:07,692 INFO [train.py:904] (3/8) Epoch 1, batch 2600, loss[loss=0.3787, simple_loss=0.4228, pruned_loss=0.1674, over 16759.00 frames. ], tot_loss[loss=0.3926, simple_loss=0.4233, pruned_loss=0.181, over 3318043.88 frames. ], batch size: 57, lr: 4.71e-02, grad_scale: 16.0 2023-04-27 13:09:20,091 INFO [optim.py:368] (3/8) 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 13:09:37,042 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0638, 3.8943, 4.2274, 4.2798, 3.3748, 4.1057, 2.9725, 4.4629], device='cuda:3'), covar=tensor([0.0627, 0.1006, 0.0542, 0.0617, 0.4312, 0.0674, 0.2167, 0.1008], device='cuda:3'), in_proj_covar=tensor([0.0048, 0.0052, 0.0045, 0.0042, 0.0089, 0.0050, 0.0061, 0.0043], device='cuda:3'), out_proj_covar=tensor([4.2030e-05, 4.5761e-05, 3.7428e-05, 4.3410e-05, 8.1242e-05, 4.3376e-05, 5.6328e-05, 4.5868e-05], device='cuda:3') 2023-04-27 13:10:11,887 INFO [train.py:904] (3/8) Epoch 1, batch 2650, loss[loss=0.3612, simple_loss=0.4238, pruned_loss=0.1493, over 16680.00 frames. ], tot_loss[loss=0.3883, simple_loss=0.4217, pruned_loss=0.1774, over 3322570.08 frames. ], batch size: 62, lr: 4.70e-02, grad_scale: 16.0 2023-04-27 13:10:55,313 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:11:09,964 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:11:15,038 INFO [train.py:904] (3/8) Epoch 1, batch 2700, loss[loss=0.4296, simple_loss=0.4538, pruned_loss=0.2027, over 16174.00 frames. ], tot_loss[loss=0.3826, simple_loss=0.4183, pruned_loss=0.1734, over 3326901.67 frames. ], batch size: 35, lr: 4.69e-02, grad_scale: 16.0 2023-04-27 13:11:29,641 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.771e+02 4.333e+02 5.264e+02 6.238e+02 1.242e+03, threshold=1.053e+03, percent-clipped=1.0 2023-04-27 13:12:14,247 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:12:20,231 INFO [train.py:904] (3/8) Epoch 1, batch 2750, loss[loss=0.3249, simple_loss=0.3827, pruned_loss=0.1336, over 17194.00 frames. ], tot_loss[loss=0.3774, simple_loss=0.4156, pruned_loss=0.1696, over 3325671.12 frames. ], batch size: 44, lr: 4.68e-02, grad_scale: 16.0 2023-04-27 13:13:13,430 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9505, 4.0278, 3.8598, 3.5660, 4.0276, 3.8951, 4.1373, 4.1326], device='cuda:3'), covar=tensor([0.0190, 0.0263, 0.0275, 0.0254, 0.0192, 0.0292, 0.0321, 0.0170], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0039, 0.0039, 0.0043, 0.0042, 0.0044, 0.0048, 0.0043], device='cuda:3'), out_proj_covar=tensor([4.9023e-05, 4.3615e-05, 4.2898e-05, 4.5744e-05, 4.4433e-05, 5.4005e-05, 5.2772e-05, 4.5687e-05], device='cuda:3') 2023-04-27 13:13:23,779 INFO [train.py:904] (3/8) Epoch 1, batch 2800, loss[loss=0.3829, simple_loss=0.4128, pruned_loss=0.1765, over 16665.00 frames. ], tot_loss[loss=0.3759, simple_loss=0.4144, pruned_loss=0.1687, over 3335356.09 frames. ], batch size: 134, lr: 4.67e-02, grad_scale: 16.0 2023-04-27 13:13:35,358 INFO [optim.py:368] (3/8) 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 13:13:54,331 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:14:25,999 INFO [train.py:904] (3/8) Epoch 1, batch 2850, loss[loss=0.3352, simple_loss=0.3868, pruned_loss=0.1418, over 17245.00 frames. ], tot_loss[loss=0.3715, simple_loss=0.4111, pruned_loss=0.1659, over 3335285.08 frames. ], batch size: 45, lr: 4.66e-02, grad_scale: 16.0 2023-04-27 13:15:09,609 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:15:27,286 INFO [train.py:904] (3/8) Epoch 1, batch 2900, loss[loss=0.3435, simple_loss=0.3711, pruned_loss=0.158, over 16451.00 frames. ], tot_loss[loss=0.3701, simple_loss=0.4088, pruned_loss=0.1657, over 3338195.58 frames. ], batch size: 75, lr: 4.65e-02, grad_scale: 16.0 2023-04-27 13:15:40,739 INFO [optim.py:368] (3/8) 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 13:16:13,960 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 2950, loss[loss=0.3712, simple_loss=0.4073, pruned_loss=0.1675, over 16475.00 frames. ], tot_loss[loss=0.3707, simple_loss=0.4077, pruned_loss=0.1669, over 3319977.27 frames. ], batch size: 68, lr: 4.64e-02, grad_scale: 16.0 2023-04-27 13:17:29,886 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:17:32,224 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:17:35,316 INFO [train.py:904] (3/8) Epoch 1, batch 3000, loss[loss=0.3658, simple_loss=0.4143, pruned_loss=0.1586, over 17052.00 frames. ], tot_loss[loss=0.3674, simple_loss=0.4056, pruned_loss=0.1646, over 3332346.45 frames. ], batch size: 53, lr: 4.63e-02, grad_scale: 16.0 2023-04-27 13:17:35,317 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 13:17:45,054 INFO [train.py:938] (3/8) Epoch 1, validation: loss=0.2847, simple_loss=0.3895, pruned_loss=0.08992, over 944034.00 frames. 2023-04-27 13:17:45,055 INFO [train.py:939] (3/8) Maximum memory allocated so far is 15576MB 2023-04-27 13:17:59,838 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.008e+02 4.355e+02 5.212e+02 6.501e+02 1.071e+03, threshold=1.042e+03, percent-clipped=0.0 2023-04-27 13:18:09,263 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:36,946 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:41,576 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:18:50,094 INFO [train.py:904] (3/8) Epoch 1, batch 3050, loss[loss=0.3078, simple_loss=0.3695, pruned_loss=0.123, over 17202.00 frames. ], tot_loss[loss=0.3669, simple_loss=0.4053, pruned_loss=0.1642, over 3317508.49 frames. ], batch size: 44, lr: 4.62e-02, grad_scale: 16.0 2023-04-27 13:19:27,656 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:19:53,489 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2300, 3.1374, 2.5778, 2.6429, 2.4773, 1.8463, 3.3113, 3.3366], device='cuda:3'), covar=tensor([0.0185, 0.0201, 0.0273, 0.0667, 0.0598, 0.0896, 0.0102, 0.0132], device='cuda:3'), in_proj_covar=tensor([0.0036, 0.0035, 0.0046, 0.0070, 0.0063, 0.0067, 0.0033, 0.0031], device='cuda:3'), out_proj_covar=tensor([4.2979e-05, 4.0661e-05, 4.6080e-05, 7.0557e-05, 6.0720e-05, 6.4258e-05, 3.4231e-05, 3.5453e-05], device='cuda:3') 2023-04-27 13:19:53,978 INFO [train.py:904] (3/8) Epoch 1, batch 3100, loss[loss=0.5039, simple_loss=0.4877, pruned_loss=0.26, over 12437.00 frames. ], tot_loss[loss=0.3648, simple_loss=0.4037, pruned_loss=0.163, over 3310563.97 frames. ], batch size: 246, lr: 4.61e-02, grad_scale: 16.0 2023-04-27 13:20:07,649 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.721e+02 4.392e+02 5.238e+02 7.570e+02 1.450e+03, threshold=1.048e+03, percent-clipped=8.0 2023-04-27 13:20:10,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0117, 4.9188, 5.1943, 5.2584, 5.5732, 5.1432, 5.0755, 5.2891], device='cuda:3'), covar=tensor([0.0279, 0.0243, 0.0517, 0.0468, 0.0241, 0.0252, 0.0396, 0.0191], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0103, 0.0132, 0.0131, 0.0130, 0.0115, 0.0121, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:3') 2023-04-27 13:20:15,026 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-27 13:20:24,446 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8807, 3.7978, 3.3821, 3.5888, 3.8942, 3.8310, 3.4775, 3.7251], device='cuda:3'), covar=tensor([0.0165, 0.0144, 0.0195, 0.0241, 0.0091, 0.0144, 0.0143, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0053, 0.0048, 0.0068, 0.0067, 0.0048, 0.0055, 0.0063, 0.0060], device='cuda:3'), out_proj_covar=tensor([6.6926e-05, 5.7999e-05, 9.9412e-05, 8.6422e-05, 5.3613e-05, 6.5449e-05, 8.5073e-05, 8.0528e-05], device='cuda:3') 2023-04-27 13:21:00,205 INFO [train.py:904] (3/8) Epoch 1, batch 3150, loss[loss=0.3835, simple_loss=0.41, pruned_loss=0.1785, over 16797.00 frames. ], tot_loss[loss=0.361, simple_loss=0.4004, pruned_loss=0.1608, over 3315457.70 frames. ], batch size: 102, lr: 4.60e-02, grad_scale: 16.0 2023-04-27 13:21:14,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3043, 4.7033, 5.0990, 5.2528, 4.4286, 4.9875, 4.9528, 4.4980], device='cuda:3'), covar=tensor([0.0246, 0.0304, 0.0231, 0.0103, 0.1363, 0.0279, 0.0181, 0.0306], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0075, 0.0120, 0.0086, 0.0145, 0.0098, 0.0088, 0.0094], device='cuda:3'), out_proj_covar=tensor([1.2808e-04, 9.1233e-05, 1.5455e-04, 1.0391e-04, 1.8938e-04, 1.3165e-04, 1.1557e-04, 1.3283e-04], device='cuda:3') 2023-04-27 13:21:38,518 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 3200, loss[loss=0.3637, simple_loss=0.3927, pruned_loss=0.1674, over 15523.00 frames. ], tot_loss[loss=0.3585, simple_loss=0.3988, pruned_loss=0.1591, over 3325210.10 frames. ], batch size: 191, lr: 4.59e-02, grad_scale: 16.0 2023-04-27 13:22:17,347 INFO [optim.py:368] (3/8) 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 13:23:07,318 INFO [train.py:904] (3/8) Epoch 1, batch 3250, loss[loss=0.3741, simple_loss=0.4007, pruned_loss=0.1738, over 16774.00 frames. ], tot_loss[loss=0.357, simple_loss=0.3977, pruned_loss=0.1582, over 3323045.35 frames. ], batch size: 83, lr: 4.58e-02, grad_scale: 16.0 2023-04-27 13:23:42,907 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7244, 4.8336, 4.4155, 5.1171, 4.9922, 5.0783, 5.0716, 4.8882], device='cuda:3'), covar=tensor([0.0359, 0.0282, 0.1337, 0.0493, 0.0503, 0.0231, 0.0300, 0.0342], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0128, 0.0216, 0.0159, 0.0134, 0.0138, 0.0127, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 13:23:56,271 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:24:02,465 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:12,252 INFO [train.py:904] (3/8) Epoch 1, batch 3300, loss[loss=0.3059, simple_loss=0.3768, pruned_loss=0.1176, over 16714.00 frames. ], tot_loss[loss=0.3571, simple_loss=0.3982, pruned_loss=0.158, over 3318060.23 frames. ], batch size: 62, lr: 4.57e-02, grad_scale: 16.0 2023-04-27 13:24:25,122 INFO [optim.py:368] (3/8) 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 13:24:27,821 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0209, 3.1376, 2.6615, 3.1919, 3.4438, 3.2300, 3.3038, 3.1773], device='cuda:3'), covar=tensor([0.0094, 0.0105, 0.0223, 0.0144, 0.0057, 0.0087, 0.0119, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0027, 0.0033, 0.0031, 0.0033, 0.0026, 0.0030, 0.0032, 0.0034], device='cuda:3'), out_proj_covar=tensor([2.5278e-05, 2.8477e-05, 2.8000e-05, 3.0188e-05, 2.2989e-05, 2.3778e-05, 2.7488e-05, 2.8673e-05], device='cuda:3') 2023-04-27 13:24:57,401 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6099, 4.7204, 4.4364, 4.5123, 4.5173, 4.8704, 4.9109, 4.4520], device='cuda:3'), covar=tensor([0.0542, 0.0894, 0.0817, 0.1337, 0.1640, 0.0585, 0.0695, 0.1418], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0186, 0.0152, 0.0162, 0.0198, 0.0138, 0.0132, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2023-04-27 13:25:03,639 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:16,110 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:25:17,767 INFO [train.py:904] (3/8) Epoch 1, batch 3350, loss[loss=0.3746, simple_loss=0.4121, pruned_loss=0.1686, over 16445.00 frames. ], tot_loss[loss=0.3535, simple_loss=0.3962, pruned_loss=0.1554, over 3314193.81 frames. ], batch size: 75, lr: 4.56e-02, grad_scale: 16.0 2023-04-27 13:25:50,435 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 1, batch 3400, loss[loss=0.3626, simple_loss=0.4058, pruned_loss=0.1597, over 16744.00 frames. ], tot_loss[loss=0.3513, simple_loss=0.3948, pruned_loss=0.1539, over 3321666.67 frames. ], batch size: 83, lr: 4.55e-02, grad_scale: 16.0 2023-04-27 13:26:39,219 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.681e+02 4.195e+02 5.225e+02 6.801e+02 1.040e+03, threshold=1.045e+03, percent-clipped=0.0 2023-04-27 13:27:18,546 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 13:27:32,307 INFO [train.py:904] (3/8) Epoch 1, batch 3450, loss[loss=0.3545, simple_loss=0.3835, pruned_loss=0.1628, over 15432.00 frames. ], tot_loss[loss=0.349, simple_loss=0.3928, pruned_loss=0.1526, over 3319893.42 frames. ], batch size: 190, lr: 4.54e-02, grad_scale: 16.0 2023-04-27 13:27:45,287 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 13:28:12,490 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:28:13,674 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3780, 4.1281, 2.8863, 4.2772, 3.7419, 1.7787, 4.3700, 3.8536], device='cuda:3'), covar=tensor([0.0946, 0.0192, 0.0703, 0.0068, 0.0597, 0.1653, 0.0095, 0.0150], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0067, 0.0105, 0.0056, 0.0067, 0.0106, 0.0057, 0.0031], device='cuda:3'), out_proj_covar=tensor([1.2753e-04, 7.5785e-05, 1.0485e-04, 5.3586e-05, 8.2206e-05, 1.0237e-04, 5.8510e-05, 3.5863e-05], device='cuda:3') 2023-04-27 13:28:39,711 INFO [train.py:904] (3/8) Epoch 1, batch 3500, loss[loss=0.351, simple_loss=0.3902, pruned_loss=0.1559, over 16745.00 frames. ], tot_loss[loss=0.3465, simple_loss=0.3911, pruned_loss=0.1509, over 3322316.40 frames. ], batch size: 134, lr: 4.53e-02, grad_scale: 16.0 2023-04-27 13:28:53,812 INFO [optim.py:368] (3/8) 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 13:28:56,158 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3619, 3.6301, 2.6447, 3.5862, 3.4270, 3.5967, 3.0332, 3.6766], device='cuda:3'), covar=tensor([0.0096, 0.0069, 0.0292, 0.0089, 0.0075, 0.0100, 0.0190, 0.0062], device='cuda:3'), in_proj_covar=tensor([0.0023, 0.0020, 0.0027, 0.0022, 0.0025, 0.0019, 0.0031, 0.0020], device='cuda:3'), out_proj_covar=tensor([2.4537e-05, 2.4262e-05, 3.1350e-05, 2.4311e-05, 2.5880e-05, 2.2924e-05, 3.2548e-05, 2.1747e-05], device='cuda:3') 2023-04-27 13:29:16,292 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:34,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9298, 2.6596, 2.7297, 2.6940, 3.2843, 3.3502, 3.4657, 2.7157], device='cuda:3'), covar=tensor([0.0112, 0.0216, 0.0232, 0.0159, 0.0069, 0.0079, 0.0110, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0025, 0.0033, 0.0027, 0.0030, 0.0023, 0.0026, 0.0027, 0.0032], device='cuda:3'), out_proj_covar=tensor([2.5203e-05, 2.9117e-05, 2.5371e-05, 2.7759e-05, 2.0860e-05, 2.1021e-05, 2.3778e-05, 2.7223e-05], device='cuda:3') 2023-04-27 13:29:40,028 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:29:46,546 INFO [train.py:904] (3/8) Epoch 1, batch 3550, loss[loss=0.3819, simple_loss=0.4099, pruned_loss=0.1769, over 12087.00 frames. ], tot_loss[loss=0.3444, simple_loss=0.3889, pruned_loss=0.1499, over 3322962.12 frames. ], batch size: 247, lr: 4.51e-02, grad_scale: 16.0 2023-04-27 13:30:01,642 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2773, 3.4619, 3.3051, 3.1715, 3.3388, 3.4143, 3.5726, 3.4830], device='cuda:3'), covar=tensor([0.0230, 0.0177, 0.0187, 0.0201, 0.0214, 0.0249, 0.0179, 0.0163], device='cuda:3'), in_proj_covar=tensor([0.0045, 0.0033, 0.0032, 0.0042, 0.0034, 0.0038, 0.0043, 0.0038], device='cuda:3'), out_proj_covar=tensor([6.0436e-05, 4.6089e-05, 4.3859e-05, 5.4450e-05, 4.7439e-05, 6.0784e-05, 5.7426e-05, 5.0652e-05], device='cuda:3') 2023-04-27 13:30:44,946 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:30:54,496 INFO [train.py:904] (3/8) Epoch 1, batch 3600, loss[loss=0.2804, simple_loss=0.3441, pruned_loss=0.1083, over 17036.00 frames. ], tot_loss[loss=0.3409, simple_loss=0.3861, pruned_loss=0.1478, over 3314049.11 frames. ], batch size: 41, lr: 4.50e-02, grad_scale: 16.0 2023-04-27 13:31:04,119 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:31:08,701 INFO [optim.py:368] (3/8) 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 13:31:29,226 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 2023-04-27 13:31:43,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9669, 3.9995, 3.8218, 2.2502, 3.6233, 3.8714, 3.6249, 3.8114], device='cuda:3'), covar=tensor([0.0104, 0.0096, 0.0137, 0.1236, 0.0221, 0.0086, 0.0128, 0.0140], device='cuda:3'), in_proj_covar=tensor([0.0042, 0.0044, 0.0050, 0.0086, 0.0043, 0.0044, 0.0042, 0.0050], device='cuda:3'), out_proj_covar=tensor([4.9009e-05, 5.2723e-05, 6.1294e-05, 1.0065e-04, 5.9541e-05, 5.4233e-05, 5.7621e-05, 5.6177e-05], device='cuda:3') 2023-04-27 13:31:51,886 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:31:56,477 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:32:06,055 INFO [train.py:904] (3/8) Epoch 1, batch 3650, loss[loss=0.3662, simple_loss=0.3933, pruned_loss=0.1696, over 11759.00 frames. ], tot_loss[loss=0.3397, simple_loss=0.3836, pruned_loss=0.1479, over 3295977.09 frames. ], batch size: 248, lr: 4.49e-02, grad_scale: 16.0 2023-04-27 13:32:42,334 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:43,568 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:33:20,376 INFO [train.py:904] (3/8) Epoch 1, batch 3700, loss[loss=0.3301, simple_loss=0.3583, pruned_loss=0.151, over 16736.00 frames. ], tot_loss[loss=0.3386, simple_loss=0.3805, pruned_loss=0.1484, over 3275028.32 frames. ], batch size: 83, lr: 4.48e-02, grad_scale: 16.0 2023-04-27 13:33:35,071 INFO [optim.py:368] (3/8) 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 13:33:54,063 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:15,102 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:34:27,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4818, 2.6392, 2.4911, 2.5072, 2.7451, 2.7760, 2.8953, 2.2443], device='cuda:3'), covar=tensor([0.0248, 0.0268, 0.0356, 0.0204, 0.0132, 0.0141, 0.0141, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0027, 0.0039, 0.0037, 0.0035, 0.0030, 0.0033, 0.0034, 0.0032], device='cuda:3'), out_proj_covar=tensor([4.0391e-05, 5.4215e-05, 5.2501e-05, 4.1880e-05, 3.9826e-05, 4.3650e-05, 4.0902e-05, 4.1948e-05], device='cuda:3') 2023-04-27 13:34:33,709 INFO [train.py:904] (3/8) Epoch 1, batch 3750, loss[loss=0.3147, simple_loss=0.3571, pruned_loss=0.1361, over 16741.00 frames. ], tot_loss[loss=0.3384, simple_loss=0.3793, pruned_loss=0.1487, over 3252817.03 frames. ], batch size: 83, lr: 4.47e-02, grad_scale: 16.0 2023-04-27 13:35:05,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5135, 3.3389, 3.1589, 2.2127, 2.4565, 1.8111, 3.4581, 3.8197], device='cuda:3'), covar=tensor([0.0433, 0.0485, 0.0479, 0.2260, 0.1360, 0.1933, 0.0357, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0041, 0.0038, 0.0058, 0.0101, 0.0085, 0.0091, 0.0046, 0.0035], device='cuda:3'), out_proj_covar=tensor([5.6137e-05, 5.0256e-05, 6.2160e-05, 1.0540e-04, 9.0419e-05, 9.2596e-05, 5.3377e-05, 4.6494e-05], device='cuda:3') 2023-04-27 13:35:46,283 INFO [train.py:904] (3/8) Epoch 1, batch 3800, loss[loss=0.3078, simple_loss=0.3687, pruned_loss=0.1235, over 17034.00 frames. ], tot_loss[loss=0.3386, simple_loss=0.379, pruned_loss=0.1491, over 3248389.76 frames. ], batch size: 53, lr: 4.46e-02, grad_scale: 16.0 2023-04-27 13:35:51,754 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1390, 4.0466, 4.2935, 4.4171, 4.5871, 4.1395, 4.1759, 4.4423], device='cuda:3'), covar=tensor([0.0310, 0.0247, 0.0446, 0.0376, 0.0277, 0.0250, 0.0480, 0.0201], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0103, 0.0130, 0.0132, 0.0139, 0.0113, 0.0125, 0.0101], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:3') 2023-04-27 13:36:00,504 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.355e+02 5.070e+02 6.208e+02 8.135e+02 1.675e+03, threshold=1.242e+03, percent-clipped=4.0 2023-04-27 13:36:57,573 INFO [train.py:904] (3/8) Epoch 1, batch 3850, loss[loss=0.3265, simple_loss=0.368, pruned_loss=0.1425, over 16774.00 frames. ], tot_loss[loss=0.3365, simple_loss=0.3772, pruned_loss=0.1478, over 3244083.58 frames. ], batch size: 124, lr: 4.45e-02, grad_scale: 16.0 2023-04-27 13:38:09,799 INFO [train.py:904] (3/8) Epoch 1, batch 3900, loss[loss=0.3228, simple_loss=0.3665, pruned_loss=0.1396, over 16261.00 frames. ], tot_loss[loss=0.3314, simple_loss=0.3728, pruned_loss=0.1449, over 3258530.06 frames. ], batch size: 165, lr: 4.44e-02, grad_scale: 16.0 2023-04-27 13:38:11,829 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:38:24,734 INFO [optim.py:368] (3/8) 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 13:38:46,412 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:38:49,729 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 13:39:11,939 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:39:21,374 INFO [train.py:904] (3/8) Epoch 1, batch 3950, loss[loss=0.3171, simple_loss=0.3542, pruned_loss=0.14, over 16881.00 frames. ], tot_loss[loss=0.3284, simple_loss=0.37, pruned_loss=0.1434, over 3264189.03 frames. ], batch size: 109, lr: 4.43e-02, grad_scale: 16.0 2023-04-27 13:39:26,974 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 13:40:10,820 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 13:40:13,968 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:20,470 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:40:37,508 INFO [train.py:904] (3/8) Epoch 1, batch 4000, loss[loss=0.3358, simple_loss=0.3717, pruned_loss=0.1499, over 16837.00 frames. ], tot_loss[loss=0.326, simple_loss=0.3681, pruned_loss=0.142, over 3264920.97 frames. ], batch size: 102, lr: 4.42e-02, grad_scale: 16.0 2023-04-27 13:40:49,866 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-27 13:40:52,171 INFO [optim.py:368] (3/8) 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 13:40:54,431 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8042, 3.8023, 3.7272, 4.1672, 4.0370, 4.1475, 4.0689, 3.9031], device='cuda:3'), covar=tensor([0.0386, 0.0269, 0.1115, 0.0321, 0.0391, 0.0257, 0.0311, 0.0312], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0127, 0.0217, 0.0157, 0.0133, 0.0137, 0.0126, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 13:41:24,004 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 13:41:50,491 INFO [train.py:904] (3/8) Epoch 1, batch 4050, loss[loss=0.2868, simple_loss=0.355, pruned_loss=0.1093, over 16831.00 frames. ], tot_loss[loss=0.3169, simple_loss=0.3629, pruned_loss=0.1354, over 3267539.16 frames. ], batch size: 83, lr: 4.41e-02, grad_scale: 16.0 2023-04-27 13:42:30,268 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7906, 3.7614, 3.6397, 3.6341, 3.8390, 4.1388, 4.0989, 3.7256], device='cuda:3'), covar=tensor([0.1027, 0.1230, 0.0884, 0.1676, 0.1647, 0.0728, 0.0759, 0.1989], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0182, 0.0148, 0.0158, 0.0188, 0.0139, 0.0142, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2023-04-27 13:42:53,231 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5406, 2.8486, 2.2008, 4.1081, 3.5322, 3.4586, 3.4409, 3.4403], device='cuda:3'), covar=tensor([0.0442, 0.0883, 0.1076, 0.0326, 0.1165, 0.0466, 0.0702, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0048, 0.0053, 0.0045, 0.0043, 0.0089, 0.0050, 0.0067, 0.0052], device='cuda:3'), out_proj_covar=tensor([4.8932e-05, 5.2969e-05, 4.4078e-05, 4.8154e-05, 9.1907e-05, 5.0497e-05, 6.5200e-05, 5.9090e-05], device='cuda:3') 2023-04-27 13:43:04,571 INFO [train.py:904] (3/8) Epoch 1, batch 4100, loss[loss=0.3749, simple_loss=0.41, pruned_loss=0.1699, over 12225.00 frames. ], tot_loss[loss=0.3121, simple_loss=0.361, pruned_loss=0.1316, over 3264417.58 frames. ], batch size: 247, lr: 4.40e-02, grad_scale: 32.0 2023-04-27 13:43:18,968 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 4.086e+02 5.427e+02 7.425e+02 1.337e+03, threshold=1.085e+03, percent-clipped=4.0 2023-04-27 13:43:52,039 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7936, 4.2793, 4.4329, 3.6982, 4.2569, 4.3488, 4.1589, 4.2165], device='cuda:3'), covar=tensor([0.0477, 0.0087, 0.0046, 0.0137, 0.0079, 0.0066, 0.0113, 0.0090], device='cuda:3'), in_proj_covar=tensor([0.0051, 0.0028, 0.0029, 0.0039, 0.0030, 0.0033, 0.0037, 0.0036], device='cuda:3'), out_proj_covar=tensor([7.7720e-05, 4.6188e-05, 4.4263e-05, 5.7435e-05, 4.7402e-05, 5.7040e-05, 5.6056e-05, 5.5152e-05], device='cuda:3') 2023-04-27 13:44:04,968 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 4150, loss[loss=0.3859, simple_loss=0.4335, pruned_loss=0.1691, over 16649.00 frames. ], tot_loss[loss=0.323, simple_loss=0.371, pruned_loss=0.1375, over 3247382.88 frames. ], batch size: 134, lr: 4.39e-02, grad_scale: 32.0 2023-04-27 13:44:39,155 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5005, 1.9241, 1.8269, 2.6236, 3.0541, 2.4940, 2.5062, 2.4282], device='cuda:3'), covar=tensor([0.0076, 0.0360, 0.0219, 0.0335, 0.0057, 0.0097, 0.0237, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0022, 0.0036, 0.0027, 0.0027, 0.0022, 0.0026, 0.0026, 0.0029], device='cuda:3'), out_proj_covar=tensor([2.3642e-05, 3.8648e-05, 2.7776e-05, 2.7404e-05, 1.9081e-05, 2.2503e-05, 2.4217e-05, 2.6626e-05], device='cuda:3') 2023-04-27 13:45:37,099 INFO [train.py:904] (3/8) Epoch 1, batch 4200, loss[loss=0.3526, simple_loss=0.4126, pruned_loss=0.1463, over 16558.00 frames. ], tot_loss[loss=0.3301, simple_loss=0.3798, pruned_loss=0.1402, over 3246699.80 frames. ], batch size: 57, lr: 4.38e-02, grad_scale: 16.0 2023-04-27 13:45:39,595 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:45:39,682 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:45:52,594 INFO [optim.py:368] (3/8) 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 13:46:50,185 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:46:50,856 INFO [train.py:904] (3/8) Epoch 1, batch 4250, loss[loss=0.3775, simple_loss=0.3998, pruned_loss=0.1776, over 11980.00 frames. ], tot_loss[loss=0.3325, simple_loss=0.3828, pruned_loss=0.1411, over 3218832.17 frames. ], batch size: 247, lr: 4.36e-02, grad_scale: 16.0 2023-04-27 13:47:36,657 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:45,314 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:48:04,724 INFO [train.py:904] (3/8) Epoch 1, batch 4300, loss[loss=0.3194, simple_loss=0.3837, pruned_loss=0.1275, over 16615.00 frames. ], tot_loss[loss=0.3305, simple_loss=0.383, pruned_loss=0.139, over 3204817.28 frames. ], batch size: 62, lr: 4.35e-02, grad_scale: 16.0 2023-04-27 13:48:21,408 INFO [optim.py:368] (3/8) 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 13:48:52,579 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:49:17,716 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:20,276 INFO [train.py:904] (3/8) Epoch 1, batch 4350, loss[loss=0.3143, simple_loss=0.3861, pruned_loss=0.1213, over 16859.00 frames. ], tot_loss[loss=0.3337, simple_loss=0.3869, pruned_loss=0.1402, over 3192756.40 frames. ], batch size: 96, lr: 4.34e-02, grad_scale: 16.0 2023-04-27 13:50:04,955 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:50:36,420 INFO [train.py:904] (3/8) Epoch 1, batch 4400, loss[loss=0.2946, simple_loss=0.3657, pruned_loss=0.1117, over 17123.00 frames. ], tot_loss[loss=0.3349, simple_loss=0.3885, pruned_loss=0.1407, over 3185927.96 frames. ], batch size: 47, lr: 4.33e-02, grad_scale: 16.0 2023-04-27 13:50:48,695 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0416, 2.6111, 2.7116, 3.4980, 3.3064, 2.9074, 3.1171, 3.2011], device='cuda:3'), covar=tensor([0.0077, 0.0473, 0.0233, 0.0079, 0.0084, 0.0223, 0.0104, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0029, 0.0055, 0.0038, 0.0035, 0.0031, 0.0035, 0.0036, 0.0033], device='cuda:3'), out_proj_covar=tensor([4.1036e-05, 9.0947e-05, 6.0986e-05, 4.5943e-05, 4.3924e-05, 5.1278e-05, 4.7862e-05, 4.5193e-05], device='cuda:3') 2023-04-27 13:50:51,959 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.736e+02 5.030e+02 6.587e+02 8.143e+02 1.430e+03, threshold=1.317e+03, percent-clipped=9.0 2023-04-27 13:51:48,408 INFO [train.py:904] (3/8) Epoch 1, batch 4450, loss[loss=0.3875, simple_loss=0.422, pruned_loss=0.1765, over 11813.00 frames. ], tot_loss[loss=0.3355, simple_loss=0.391, pruned_loss=0.14, over 3189101.01 frames. ], batch size: 248, lr: 4.32e-02, grad_scale: 16.0 2023-04-27 13:52:57,207 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:53:02,592 INFO [train.py:904] (3/8) Epoch 1, batch 4500, loss[loss=0.3305, simple_loss=0.3881, pruned_loss=0.1364, over 16706.00 frames. ], tot_loss[loss=0.332, simple_loss=0.3888, pruned_loss=0.1376, over 3206480.05 frames. ], batch size: 134, lr: 4.31e-02, grad_scale: 8.0 2023-04-27 13:53:20,027 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.477e+02 4.034e+02 4.961e+02 6.637e+02 1.457e+03, threshold=9.923e+02, percent-clipped=1.0 2023-04-27 13:54:14,092 INFO [train.py:904] (3/8) Epoch 1, batch 4550, loss[loss=0.3611, simple_loss=0.4158, pruned_loss=0.1533, over 16881.00 frames. ], tot_loss[loss=0.3306, simple_loss=0.3877, pruned_loss=0.1368, over 3207436.56 frames. ], batch size: 116, lr: 4.30e-02, grad_scale: 8.0 2023-04-27 13:54:47,112 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0050, 2.9688, 2.0006, 3.0899, 3.0335, 3.1560, 2.9009, 3.0065], device='cuda:3'), covar=tensor([0.0627, 0.0587, 0.2616, 0.0776, 0.0997, 0.0797, 0.0881, 0.0855], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0130, 0.0219, 0.0152, 0.0124, 0.0136, 0.0117, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 13:54:57,994 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:55:25,714 INFO [train.py:904] (3/8) Epoch 1, batch 4600, loss[loss=0.3159, simple_loss=0.3791, pruned_loss=0.1264, over 16870.00 frames. ], tot_loss[loss=0.3293, simple_loss=0.3876, pruned_loss=0.1355, over 3217261.78 frames. ], batch size: 109, lr: 4.29e-02, grad_scale: 8.0 2023-04-27 13:55:39,908 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9281, 4.8597, 4.5849, 5.2624, 5.2374, 4.9649, 5.2975, 5.0772], device='cuda:3'), covar=tensor([0.0303, 0.0194, 0.0867, 0.0221, 0.0206, 0.0227, 0.0160, 0.0188], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0126, 0.0217, 0.0150, 0.0122, 0.0133, 0.0115, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 13:55:43,311 INFO [optim.py:368] (3/8) 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 13:56:07,432 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:28,165 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:36,567 INFO [train.py:904] (3/8) Epoch 1, batch 4650, loss[loss=0.2793, simple_loss=0.3463, pruned_loss=0.1062, over 16819.00 frames. ], tot_loss[loss=0.3256, simple_loss=0.3844, pruned_loss=0.1333, over 3212485.64 frames. ], batch size: 39, lr: 4.28e-02, grad_scale: 8.0 2023-04-27 13:56:43,391 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4031, 3.5731, 1.4107, 3.6380, 2.6044, 3.4647, 1.7431, 2.9937], device='cuda:3'), covar=tensor([0.0104, 0.0081, 0.1339, 0.0080, 0.0403, 0.0146, 0.1022, 0.0300], device='cuda:3'), in_proj_covar=tensor([0.0040, 0.0037, 0.0076, 0.0039, 0.0061, 0.0036, 0.0080, 0.0049], device='cuda:3'), out_proj_covar=tensor([4.9127e-05, 5.0116e-05, 1.0559e-04, 4.7151e-05, 7.5541e-05, 5.2782e-05, 1.0495e-04, 6.7730e-05], device='cuda:3') 2023-04-27 13:57:50,180 INFO [train.py:904] (3/8) Epoch 1, batch 4700, loss[loss=0.2979, simple_loss=0.3654, pruned_loss=0.1152, over 16767.00 frames. ], tot_loss[loss=0.3207, simple_loss=0.3803, pruned_loss=0.1306, over 3222689.20 frames. ], batch size: 83, lr: 4.27e-02, grad_scale: 8.0 2023-04-27 13:58:07,986 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.460e+02 4.210e+02 5.632e+02 6.697e+02 1.082e+03, threshold=1.126e+03, percent-clipped=2.0 2023-04-27 13:58:39,553 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2666, 3.2033, 3.0412, 3.4429, 3.3579, 3.3523, 3.2176, 3.2914], device='cuda:3'), covar=tensor([0.0290, 0.0290, 0.1016, 0.0328, 0.0433, 0.0412, 0.0461, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0132, 0.0224, 0.0156, 0.0130, 0.0137, 0.0118, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 13:59:02,500 INFO [train.py:904] (3/8) Epoch 1, batch 4750, loss[loss=0.2927, simple_loss=0.353, pruned_loss=0.1162, over 16178.00 frames. ], tot_loss[loss=0.3173, simple_loss=0.3766, pruned_loss=0.129, over 3214368.96 frames. ], batch size: 35, lr: 4.26e-02, grad_scale: 8.0 2023-04-27 13:59:06,947 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=2.05 vs. limit=2.0 2023-04-27 13:59:16,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5570, 3.4999, 3.1886, 2.7680, 3.5810, 3.3749, 3.3130, 3.5989], device='cuda:3'), covar=tensor([0.0194, 0.0122, 0.0148, 0.0494, 0.0095, 0.0209, 0.0147, 0.0151], device='cuda:3'), in_proj_covar=tensor([0.0049, 0.0038, 0.0056, 0.0072, 0.0043, 0.0054, 0.0056, 0.0052], device='cuda:3'), out_proj_covar=tensor([9.4752e-05, 7.2099e-05, 1.1450e-04, 1.3120e-04, 7.3209e-05, 1.0170e-04, 1.1230e-04, 1.0919e-04], device='cuda:3') 2023-04-27 13:59:58,108 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:00:11,488 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:00:16,558 INFO [train.py:904] (3/8) Epoch 1, batch 4800, loss[loss=0.2938, simple_loss=0.3579, pruned_loss=0.1148, over 16533.00 frames. ], tot_loss[loss=0.3131, simple_loss=0.3726, pruned_loss=0.1268, over 3216650.45 frames. ], batch size: 75, lr: 4.25e-02, grad_scale: 8.0 2023-04-27 14:00:28,438 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 14:00:34,036 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.251e+02 4.322e+02 5.200e+02 6.651e+02 1.076e+03, threshold=1.040e+03, percent-clipped=0.0 2023-04-27 14:01:01,680 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 14:01:22,950 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:31,748 INFO [train.py:904] (3/8) Epoch 1, batch 4850, loss[loss=0.3926, simple_loss=0.4263, pruned_loss=0.1795, over 12023.00 frames. ], tot_loss[loss=0.3143, simple_loss=0.3742, pruned_loss=0.1272, over 3188517.58 frames. ], batch size: 246, lr: 4.24e-02, grad_scale: 8.0 2023-04-27 14:02:49,096 INFO [train.py:904] (3/8) Epoch 1, batch 4900, loss[loss=0.2522, simple_loss=0.3311, pruned_loss=0.08666, over 16697.00 frames. ], tot_loss[loss=0.3117, simple_loss=0.3731, pruned_loss=0.1252, over 3192285.54 frames. ], batch size: 89, lr: 4.23e-02, grad_scale: 8.0 2023-04-27 14:03:07,683 INFO [optim.py:368] (3/8) 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 14:03:11,154 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3874, 4.1547, 3.8627, 3.6016, 4.3304, 4.2590, 3.7752, 4.1271], device='cuda:3'), covar=tensor([0.0157, 0.0126, 0.0116, 0.0389, 0.0055, 0.0133, 0.0120, 0.0150], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0036, 0.0053, 0.0068, 0.0040, 0.0052, 0.0053, 0.0050], device='cuda:3'), out_proj_covar=tensor([9.1738e-05, 6.9602e-05, 1.1188e-04, 1.2588e-04, 7.0272e-05, 1.0052e-04, 1.0781e-04, 1.0738e-04], device='cuda:3') 2023-04-27 14:03:20,898 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8634, 1.9127, 1.9350, 1.9125, 2.3526, 2.4301, 2.5201, 2.3076], device='cuda:3'), covar=tensor([0.0126, 0.0645, 0.0299, 0.0306, 0.0144, 0.0171, 0.0107, 0.0174], device='cuda:3'), in_proj_covar=tensor([0.0033, 0.0065, 0.0044, 0.0040, 0.0033, 0.0037, 0.0038, 0.0035], device='cuda:3'), out_proj_covar=tensor([4.9299e-05, 1.1364e-04, 7.4378e-05, 5.6002e-05, 4.7533e-05, 5.6635e-05, 5.3528e-05, 5.1444e-05], device='cuda:3') 2023-04-27 14:03:52,386 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7616, 1.5996, 1.5845, 1.4397, 1.8320, 1.8159, 1.9024, 1.7499], device='cuda:3'), covar=tensor([0.0124, 0.0608, 0.0242, 0.0260, 0.0152, 0.0184, 0.0137, 0.0171], device='cuda:3'), in_proj_covar=tensor([0.0033, 0.0066, 0.0046, 0.0040, 0.0033, 0.0037, 0.0038, 0.0035], device='cuda:3'), out_proj_covar=tensor([4.9517e-05, 1.1516e-04, 7.7101e-05, 5.7142e-05, 4.8457e-05, 5.7232e-05, 5.3438e-05, 5.1315e-05], device='cuda:3') 2023-04-27 14:03:54,933 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:04,716 INFO [train.py:904] (3/8) Epoch 1, batch 4950, loss[loss=0.3025, simple_loss=0.3686, pruned_loss=0.1183, over 16876.00 frames. ], tot_loss[loss=0.3126, simple_loss=0.3738, pruned_loss=0.1257, over 3193505.68 frames. ], batch size: 109, lr: 4.21e-02, grad_scale: 8.0 2023-04-27 14:05:04,811 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:05:17,359 INFO [train.py:904] (3/8) Epoch 1, batch 5000, loss[loss=0.3046, simple_loss=0.3722, pruned_loss=0.1185, over 16561.00 frames. ], tot_loss[loss=0.315, simple_loss=0.3765, pruned_loss=0.1267, over 3185290.89 frames. ], batch size: 75, lr: 4.20e-02, grad_scale: 8.0 2023-04-27 14:05:23,541 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 14:05:25,623 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7207, 3.3014, 2.0346, 3.7903, 3.9885, 3.8412, 2.6657, 3.5723], device='cuda:3'), covar=tensor([0.2350, 0.0221, 0.1529, 0.0062, 0.0072, 0.0173, 0.0549, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0074, 0.0132, 0.0045, 0.0045, 0.0062, 0.0098, 0.0074], device='cuda:3'), out_proj_covar=tensor([1.7778e-04, 9.4334e-05, 1.5767e-04, 6.4160e-05, 6.6002e-05, 9.7249e-05, 1.2405e-04, 9.6395e-05], device='cuda:3') 2023-04-27 14:05:35,407 INFO [optim.py:368] (3/8) 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 14:06:31,099 INFO [train.py:904] (3/8) Epoch 1, batch 5050, loss[loss=0.2819, simple_loss=0.3592, pruned_loss=0.1023, over 16473.00 frames. ], tot_loss[loss=0.3127, simple_loss=0.3753, pruned_loss=0.125, over 3207757.03 frames. ], batch size: 75, lr: 4.19e-02, grad_scale: 8.0 2023-04-27 14:07:00,078 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3154, 3.5070, 1.5784, 3.5244, 2.2879, 3.4772, 1.8393, 2.7440], device='cuda:3'), covar=tensor([0.0159, 0.0074, 0.1210, 0.0071, 0.0443, 0.0077, 0.0992, 0.0326], device='cuda:3'), in_proj_covar=tensor([0.0045, 0.0042, 0.0088, 0.0046, 0.0071, 0.0039, 0.0095, 0.0062], device='cuda:3'), out_proj_covar=tensor([5.9448e-05, 5.9398e-05, 1.2501e-04, 5.9580e-05, 9.3918e-05, 6.1583e-05, 1.2827e-04, 9.0219e-05], device='cuda:3') 2023-04-27 14:07:42,590 INFO [train.py:904] (3/8) Epoch 1, batch 5100, loss[loss=0.3096, simple_loss=0.3656, pruned_loss=0.1267, over 16527.00 frames. ], tot_loss[loss=0.3094, simple_loss=0.3726, pruned_loss=0.1231, over 3199853.28 frames. ], batch size: 57, lr: 4.18e-02, grad_scale: 8.0 2023-04-27 14:07:59,829 INFO [optim.py:368] (3/8) 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 14:08:25,387 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3675, 5.4217, 5.3378, 5.5425, 5.0458, 5.0745, 5.2341, 5.6944], device='cuda:3'), covar=tensor([0.0268, 0.0445, 0.0444, 0.0194, 0.0478, 0.0224, 0.0313, 0.0201], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0171, 0.0162, 0.0115, 0.0147, 0.0115, 0.0155, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:08:58,107 INFO [train.py:904] (3/8) Epoch 1, batch 5150, loss[loss=0.2582, simple_loss=0.3428, pruned_loss=0.0868, over 16704.00 frames. ], tot_loss[loss=0.3082, simple_loss=0.3723, pruned_loss=0.1221, over 3194464.62 frames. ], batch size: 89, lr: 4.17e-02, grad_scale: 8.0 2023-04-27 14:09:51,066 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9014, 4.0157, 4.0295, 1.7610, 3.8528, 4.1158, 3.7987, 3.7993], device='cuda:3'), covar=tensor([0.0076, 0.0137, 0.0122, 0.1794, 0.0165, 0.0135, 0.0103, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0059, 0.0065, 0.0057, 0.0132, 0.0057, 0.0058, 0.0059, 0.0075], device='cuda:3'), out_proj_covar=tensor([7.9546e-05, 8.9261e-05, 8.4039e-05, 1.8365e-04, 8.6666e-05, 8.3092e-05, 9.3287e-05, 9.8866e-05], device='cuda:3') 2023-04-27 14:10:12,903 INFO [train.py:904] (3/8) Epoch 1, batch 5200, loss[loss=0.3058, simple_loss=0.3601, pruned_loss=0.1257, over 16888.00 frames. ], tot_loss[loss=0.3074, simple_loss=0.3707, pruned_loss=0.122, over 3207157.33 frames. ], batch size: 109, lr: 4.16e-02, grad_scale: 8.0 2023-04-27 14:10:30,245 INFO [optim.py:368] (3/8) 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 14:10:53,670 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6101, 4.1700, 4.5236, 4.6943, 3.8488, 4.4535, 4.2750, 4.1539], device='cuda:3'), covar=tensor([0.0233, 0.0177, 0.0175, 0.0090, 0.0838, 0.0194, 0.0182, 0.0200], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0065, 0.0116, 0.0087, 0.0143, 0.0092, 0.0078, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-27 14:11:26,081 INFO [train.py:904] (3/8) Epoch 1, batch 5250, loss[loss=0.268, simple_loss=0.3371, pruned_loss=0.0995, over 16724.00 frames. ], tot_loss[loss=0.3044, simple_loss=0.3671, pruned_loss=0.1209, over 3219592.98 frames. ], batch size: 89, lr: 4.15e-02, grad_scale: 8.0 2023-04-27 14:11:50,372 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3846, 2.1608, 2.0279, 2.5623, 2.6943, 2.7919, 2.8824, 2.6591], device='cuda:3'), covar=tensor([0.0123, 0.0604, 0.0387, 0.0160, 0.0190, 0.0167, 0.0104, 0.0239], device='cuda:3'), in_proj_covar=tensor([0.0035, 0.0071, 0.0051, 0.0042, 0.0035, 0.0037, 0.0041, 0.0037], device='cuda:3'), out_proj_covar=tensor([5.4654e-05, 1.2576e-04, 8.7851e-05, 6.3125e-05, 5.3997e-05, 5.9006e-05, 5.7622e-05, 5.7189e-05], device='cuda:3') 2023-04-27 14:12:26,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4776, 3.5253, 3.4393, 1.7061, 3.2899, 3.6049, 3.3677, 3.2880], device='cuda:3'), covar=tensor([0.0104, 0.0130, 0.0127, 0.2045, 0.0171, 0.0090, 0.0135, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0058, 0.0064, 0.0057, 0.0134, 0.0058, 0.0058, 0.0060, 0.0076], device='cuda:3'), out_proj_covar=tensor([7.9789e-05, 8.9771e-05, 8.4883e-05, 1.8615e-04, 8.8160e-05, 8.3862e-05, 9.6494e-05, 1.0321e-04], device='cuda:3') 2023-04-27 14:12:37,169 INFO [train.py:904] (3/8) Epoch 1, batch 5300, loss[loss=0.256, simple_loss=0.3331, pruned_loss=0.08948, over 16390.00 frames. ], tot_loss[loss=0.2995, simple_loss=0.3622, pruned_loss=0.1184, over 3229740.96 frames. ], batch size: 68, lr: 4.14e-02, grad_scale: 8.0 2023-04-27 14:12:54,710 INFO [optim.py:368] (3/8) 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 14:13:00,367 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1955, 5.3889, 5.1226, 5.4387, 4.8624, 4.8560, 5.0522, 5.4924], device='cuda:3'), covar=tensor([0.0300, 0.0510, 0.0584, 0.0220, 0.0460, 0.0306, 0.0382, 0.0271], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0179, 0.0169, 0.0120, 0.0147, 0.0121, 0.0162, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:13:49,535 INFO [train.py:904] (3/8) Epoch 1, batch 5350, loss[loss=0.3197, simple_loss=0.3711, pruned_loss=0.1342, over 17020.00 frames. ], tot_loss[loss=0.2963, simple_loss=0.3596, pruned_loss=0.1165, over 3241349.39 frames. ], batch size: 53, lr: 4.13e-02, grad_scale: 8.0 2023-04-27 14:15:00,987 INFO [train.py:904] (3/8) Epoch 1, batch 5400, loss[loss=0.2953, simple_loss=0.372, pruned_loss=0.1093, over 16723.00 frames. ], tot_loss[loss=0.3001, simple_loss=0.3638, pruned_loss=0.1182, over 3241979.62 frames. ], batch size: 89, lr: 4.12e-02, grad_scale: 8.0 2023-04-27 14:15:18,319 INFO [optim.py:368] (3/8) 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 14:15:32,436 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 14:16:19,545 INFO [train.py:904] (3/8) Epoch 1, batch 5450, loss[loss=0.3905, simple_loss=0.4407, pruned_loss=0.1702, over 16498.00 frames. ], tot_loss[loss=0.3059, simple_loss=0.3685, pruned_loss=0.1217, over 3242152.54 frames. ], batch size: 75, lr: 4.11e-02, grad_scale: 8.0 2023-04-27 14:16:59,894 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 14:17:37,159 INFO [train.py:904] (3/8) Epoch 1, batch 5500, loss[loss=0.4052, simple_loss=0.4322, pruned_loss=0.1891, over 15266.00 frames. ], tot_loss[loss=0.3239, simple_loss=0.3812, pruned_loss=0.1333, over 3189597.66 frames. ], batch size: 190, lr: 4.10e-02, grad_scale: 8.0 2023-04-27 14:17:56,197 INFO [optim.py:368] (3/8) 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 14:18:57,386 INFO [train.py:904] (3/8) Epoch 1, batch 5550, loss[loss=0.4933, simple_loss=0.487, pruned_loss=0.2498, over 11197.00 frames. ], tot_loss[loss=0.3386, simple_loss=0.3914, pruned_loss=0.1429, over 3156999.58 frames. ], batch size: 247, lr: 4.09e-02, grad_scale: 8.0 2023-04-27 14:19:13,033 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9497, 3.7168, 4.1088, 4.2841, 4.4802, 3.9884, 3.9910, 4.3023], device='cuda:3'), covar=tensor([0.0276, 0.0335, 0.0691, 0.0451, 0.0375, 0.0315, 0.0626, 0.0232], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0100, 0.0128, 0.0124, 0.0138, 0.0108, 0.0135, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:20:17,850 INFO [train.py:904] (3/8) Epoch 1, batch 5600, loss[loss=0.3331, simple_loss=0.3916, pruned_loss=0.1373, over 16619.00 frames. ], tot_loss[loss=0.351, simple_loss=0.3993, pruned_loss=0.1513, over 3111931.58 frames. ], batch size: 62, lr: 4.08e-02, grad_scale: 8.0 2023-04-27 14:20:37,625 INFO [optim.py:368] (3/8) 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 14:21:31,838 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3572, 2.9212, 3.0537, 3.7206, 2.7481, 3.5129, 2.9977, 2.9472], device='cuda:3'), covar=tensor([0.0300, 0.0369, 0.0288, 0.0223, 0.1160, 0.0239, 0.0502, 0.0714], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0081, 0.0068, 0.0080, 0.0147, 0.0079, 0.0100, 0.0089], device='cuda:3'), out_proj_covar=tensor([9.5191e-05, 9.7583e-05, 8.0867e-05, 1.0262e-04, 1.7920e-04, 9.5992e-05, 1.1021e-04, 1.1731e-04], device='cuda:3') 2023-04-27 14:21:39,870 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 5650, loss[loss=0.369, simple_loss=0.4305, pruned_loss=0.1537, over 16692.00 frames. ], tot_loss[loss=0.364, simple_loss=0.4084, pruned_loss=0.1598, over 3079625.03 frames. ], batch size: 89, lr: 4.07e-02, grad_scale: 8.0 2023-04-27 14:22:59,479 INFO [train.py:904] (3/8) Epoch 1, batch 5700, loss[loss=0.3663, simple_loss=0.4142, pruned_loss=0.1592, over 16945.00 frames. ], tot_loss[loss=0.3662, simple_loss=0.4097, pruned_loss=0.1614, over 3076000.00 frames. ], batch size: 109, lr: 4.06e-02, grad_scale: 8.0 2023-04-27 14:23:15,451 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:23:17,958 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.964e+02 6.202e+02 7.607e+02 9.700e+02 2.079e+03, threshold=1.521e+03, percent-clipped=5.0 2023-04-27 14:23:29,859 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1078, 4.2076, 1.6890, 4.3603, 2.3480, 3.9722, 2.1414, 2.8481], device='cuda:3'), covar=tensor([0.0054, 0.0073, 0.1398, 0.0037, 0.0652, 0.0140, 0.1106, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0050, 0.0049, 0.0102, 0.0051, 0.0087, 0.0047, 0.0111, 0.0078], device='cuda:3'), out_proj_covar=tensor([6.9208e-05, 7.4091e-05, 1.4827e-04, 6.8902e-05, 1.2042e-04, 7.8493e-05, 1.5973e-04, 1.2151e-04], device='cuda:3') 2023-04-27 14:23:36,033 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9329, 2.4730, 2.2748, 3.1351, 2.5400, 2.8953, 2.6349, 2.4469], device='cuda:3'), covar=tensor([0.0274, 0.0367, 0.0301, 0.0274, 0.0815, 0.0272, 0.0484, 0.0723], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0083, 0.0071, 0.0083, 0.0153, 0.0081, 0.0103, 0.0094], device='cuda:3'), out_proj_covar=tensor([9.9831e-05, 1.0171e-04, 8.4572e-05, 1.0758e-04, 1.8882e-04, 9.8154e-05, 1.1367e-04, 1.2467e-04], device='cuda:3') 2023-04-27 14:23:54,497 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-27 14:24:21,183 INFO [train.py:904] (3/8) Epoch 1, batch 5750, loss[loss=0.3822, simple_loss=0.4315, pruned_loss=0.1665, over 16843.00 frames. ], tot_loss[loss=0.3687, simple_loss=0.4124, pruned_loss=0.1625, over 3075616.11 frames. ], batch size: 116, lr: 4.05e-02, grad_scale: 8.0 2023-04-27 14:24:28,907 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:42,931 INFO [train.py:904] (3/8) Epoch 1, batch 5800, loss[loss=0.3852, simple_loss=0.4115, pruned_loss=0.1795, over 12040.00 frames. ], tot_loss[loss=0.3674, simple_loss=0.4123, pruned_loss=0.1612, over 3074372.21 frames. ], batch size: 248, lr: 4.04e-02, grad_scale: 8.0 2023-04-27 14:26:01,838 INFO [optim.py:368] (3/8) 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 14:26:07,037 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 5850, loss[loss=0.3315, simple_loss=0.3914, pruned_loss=0.1358, over 16807.00 frames. ], tot_loss[loss=0.3618, simple_loss=0.4085, pruned_loss=0.1576, over 3068628.43 frames. ], batch size: 83, lr: 4.03e-02, grad_scale: 8.0 2023-04-27 14:27:14,377 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:25,812 INFO [train.py:904] (3/8) Epoch 1, batch 5900, loss[loss=0.3185, simple_loss=0.385, pruned_loss=0.126, over 16512.00 frames. ], tot_loss[loss=0.3598, simple_loss=0.4073, pruned_loss=0.1562, over 3069888.13 frames. ], batch size: 68, lr: 4.02e-02, grad_scale: 8.0 2023-04-27 14:28:48,065 INFO [optim.py:368] (3/8) 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 14:28:58,928 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:29:31,339 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-04-27 14:29:49,226 INFO [train.py:904] (3/8) Epoch 1, batch 5950, loss[loss=0.3866, simple_loss=0.4264, pruned_loss=0.1734, over 11807.00 frames. ], tot_loss[loss=0.3584, simple_loss=0.4077, pruned_loss=0.1545, over 3068148.26 frames. ], batch size: 247, lr: 4.01e-02, grad_scale: 8.0 2023-04-27 14:30:39,232 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 14:30:52,720 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:31:14,068 INFO [train.py:904] (3/8) Epoch 1, batch 6000, loss[loss=0.3352, simple_loss=0.3917, pruned_loss=0.1393, over 16891.00 frames. ], tot_loss[loss=0.3587, simple_loss=0.4078, pruned_loss=0.1548, over 3078275.80 frames. ], batch size: 90, lr: 4.00e-02, grad_scale: 8.0 2023-04-27 14:31:14,068 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 14:31:23,950 INFO [train.py:938] (3/8) Epoch 1, validation: loss=0.2762, simple_loss=0.3752, pruned_loss=0.08863, over 944034.00 frames. 2023-04-27 14:31:23,951 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17414MB 2023-04-27 14:31:31,561 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:31:41,456 INFO [optim.py:368] (3/8) 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 14:32:01,250 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.74 vs. limit=5.0 2023-04-27 14:32:30,007 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:42,910 INFO [train.py:904] (3/8) Epoch 1, batch 6050, loss[loss=0.3349, simple_loss=0.4024, pruned_loss=0.1336, over 17289.00 frames. ], tot_loss[loss=0.3565, simple_loss=0.4057, pruned_loss=0.1536, over 3092664.97 frames. ], batch size: 52, lr: 3.99e-02, grad_scale: 8.0 2023-04-27 14:32:44,285 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:03,529 INFO [train.py:904] (3/8) Epoch 1, batch 6100, loss[loss=0.3126, simple_loss=0.3659, pruned_loss=0.1296, over 16632.00 frames. ], tot_loss[loss=0.3518, simple_loss=0.4028, pruned_loss=0.1504, over 3105775.31 frames. ], batch size: 57, lr: 3.98e-02, grad_scale: 8.0 2023-04-27 14:34:08,709 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:16,486 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9927, 3.8064, 3.4183, 3.1868, 3.8091, 3.1213, 3.5692, 3.7327], device='cuda:3'), covar=tensor([0.0086, 0.0086, 0.0111, 0.0352, 0.0070, 0.0375, 0.0084, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0036, 0.0051, 0.0071, 0.0040, 0.0062, 0.0050, 0.0049], device='cuda:3'), out_proj_covar=tensor([1.0823e-04, 8.3495e-05, 1.2498e-04, 1.5449e-04, 8.9447e-05, 1.3970e-04, 1.2091e-04, 1.2691e-04], device='cuda:3') 2023-04-27 14:34:21,377 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:24,016 INFO [optim.py:368] (3/8) 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 14:34:59,877 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 6150, loss[loss=0.3296, simple_loss=0.389, pruned_loss=0.1351, over 16883.00 frames. ], tot_loss[loss=0.3483, simple_loss=0.3998, pruned_loss=0.1485, over 3122038.59 frames. ], batch size: 116, lr: 3.97e-02, grad_scale: 8.0 2023-04-27 14:35:57,118 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:46,262 INFO [train.py:904] (3/8) Epoch 1, batch 6200, loss[loss=0.3636, simple_loss=0.4135, pruned_loss=0.1569, over 16401.00 frames. ], tot_loss[loss=0.3458, simple_loss=0.3972, pruned_loss=0.1472, over 3127715.12 frames. ], batch size: 146, lr: 3.96e-02, grad_scale: 8.0 2023-04-27 14:36:49,253 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:03,225 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:05,930 INFO [optim.py:368] (3/8) 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 14:37:06,541 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3289, 4.3013, 4.8346, 4.8556, 4.9994, 4.4239, 4.6641, 4.7686], device='cuda:3'), covar=tensor([0.0245, 0.0224, 0.0315, 0.0296, 0.0283, 0.0240, 0.0435, 0.0208], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0107, 0.0131, 0.0128, 0.0138, 0.0115, 0.0149, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:37:08,080 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:28,885 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8564, 3.9948, 3.5345, 1.8078, 3.0425, 1.9425, 3.2962, 4.0419], device='cuda:3'), covar=tensor([0.0335, 0.0308, 0.0344, 0.2106, 0.0949, 0.1519, 0.1122, 0.0205], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0063, 0.0101, 0.0141, 0.0138, 0.0131, 0.0122, 0.0058], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:37:33,293 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:56,078 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0162, 3.1899, 2.6738, 2.0200, 2.7072, 2.0676, 2.7308, 3.0437], device='cuda:3'), covar=tensor([0.0228, 0.0189, 0.0260, 0.1387, 0.0765, 0.1051, 0.0598, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0062, 0.0100, 0.0138, 0.0137, 0.0129, 0.0119, 0.0057], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:38:02,370 INFO [train.py:904] (3/8) Epoch 1, batch 6250, loss[loss=0.3195, simple_loss=0.3886, pruned_loss=0.1252, over 16783.00 frames. ], tot_loss[loss=0.3436, simple_loss=0.3958, pruned_loss=0.1457, over 3139222.19 frames. ], batch size: 83, lr: 3.95e-02, grad_scale: 8.0 2023-04-27 14:38:11,007 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4788, 3.3676, 3.8035, 3.7997, 3.9202, 3.4327, 3.5978, 3.7912], device='cuda:3'), covar=tensor([0.0307, 0.0330, 0.0467, 0.0432, 0.0378, 0.0379, 0.0613, 0.0275], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0108, 0.0134, 0.0130, 0.0139, 0.0117, 0.0152, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 14:38:20,035 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 14:38:21,002 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:32,085 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:37,881 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 14:39:04,019 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-27 14:39:15,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8007, 3.6963, 4.1898, 4.1970, 4.3331, 3.8654, 3.9522, 4.1281], device='cuda:3'), covar=tensor([0.0292, 0.0345, 0.0561, 0.0513, 0.0365, 0.0328, 0.0600, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0113, 0.0135, 0.0133, 0.0143, 0.0118, 0.0155, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 14:39:17,017 INFO [train.py:904] (3/8) Epoch 1, batch 6300, loss[loss=0.3419, simple_loss=0.3927, pruned_loss=0.1455, over 16754.00 frames. ], tot_loss[loss=0.3428, simple_loss=0.3954, pruned_loss=0.1451, over 3140190.78 frames. ], batch size: 134, lr: 3.94e-02, grad_scale: 8.0 2023-04-27 14:39:25,747 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:39:36,720 INFO [optim.py:368] (3/8) 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 14:39:42,794 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 14:39:59,130 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:28,662 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:36,333 INFO [train.py:904] (3/8) Epoch 1, batch 6350, loss[loss=0.3863, simple_loss=0.4109, pruned_loss=0.1809, over 11554.00 frames. ], tot_loss[loss=0.3478, simple_loss=0.3983, pruned_loss=0.1486, over 3130748.70 frames. ], batch size: 248, lr: 3.93e-02, grad_scale: 8.0 2023-04-27 14:40:40,270 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:41:26,677 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5851, 3.4716, 3.6127, 3.4468, 3.6091, 3.9800, 3.9512, 3.5080], device='cuda:3'), covar=tensor([0.1860, 0.1947, 0.1159, 0.2104, 0.2079, 0.0983, 0.0861, 0.2343], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0194, 0.0158, 0.0168, 0.0210, 0.0159, 0.0151, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-27 14:41:50,569 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:53,486 INFO [train.py:904] (3/8) Epoch 1, batch 6400, loss[loss=0.3132, simple_loss=0.3751, pruned_loss=0.1256, over 16856.00 frames. ], tot_loss[loss=0.3508, simple_loss=0.3994, pruned_loss=0.1511, over 3097495.52 frames. ], batch size: 116, lr: 3.92e-02, grad_scale: 8.0 2023-04-27 14:42:08,584 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 4.039e+02 5.999e+02 7.400e+02 9.043e+02 1.587e+03, threshold=1.480e+03, percent-clipped=2.0 2023-04-27 14:43:09,618 INFO [train.py:904] (3/8) Epoch 1, batch 6450, loss[loss=0.31, simple_loss=0.3809, pruned_loss=0.1195, over 16700.00 frames. ], tot_loss[loss=0.3478, simple_loss=0.3976, pruned_loss=0.149, over 3066304.23 frames. ], batch size: 89, lr: 3.91e-02, grad_scale: 8.0 2023-04-27 14:43:22,744 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:26,518 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:33,813 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:43,730 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5055, 3.7964, 3.6527, 1.5588, 3.6964, 3.8354, 3.5411, 3.4148], device='cuda:3'), covar=tensor([0.0151, 0.0189, 0.0223, 0.2002, 0.0216, 0.0184, 0.0170, 0.0230], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0062, 0.0058, 0.0134, 0.0058, 0.0051, 0.0062, 0.0078], device='cuda:3'), out_proj_covar=tensor([1.0215e-04, 9.2807e-05, 9.3620e-05, 1.9918e-04, 9.2869e-05, 8.2282e-05, 1.0827e-04, 1.1740e-04], device='cuda:3') 2023-04-27 14:43:54,668 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8851, 4.0201, 3.2743, 1.5897, 3.0761, 2.0996, 3.3143, 3.9395], device='cuda:3'), covar=tensor([0.0312, 0.0223, 0.0361, 0.2138, 0.0838, 0.1255, 0.0831, 0.0206], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0064, 0.0103, 0.0143, 0.0139, 0.0133, 0.0124, 0.0061], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 14:44:12,250 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:26,979 INFO [train.py:904] (3/8) Epoch 1, batch 6500, loss[loss=0.3418, simple_loss=0.3914, pruned_loss=0.1461, over 16678.00 frames. ], tot_loss[loss=0.3446, simple_loss=0.3946, pruned_loss=0.1473, over 3078666.69 frames. ], batch size: 134, lr: 3.90e-02, grad_scale: 16.0 2023-04-27 14:44:45,181 INFO [optim.py:368] (3/8) 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 14:44:46,893 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:00,316 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:07,596 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 6550, loss[loss=0.3083, simple_loss=0.3885, pruned_loss=0.114, over 17123.00 frames. ], tot_loss[loss=0.348, simple_loss=0.3987, pruned_loss=0.1487, over 3084809.86 frames. ], batch size: 49, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:45:51,928 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-27 14:45:55,591 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:00,518 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:08,769 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:21,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5262, 2.6955, 2.5784, 1.7747, 2.6450, 2.6328, 2.4165, 2.5639], device='cuda:3'), covar=tensor([0.0187, 0.0099, 0.0168, 0.1349, 0.0148, 0.0101, 0.0245, 0.0207], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0062, 0.0061, 0.0141, 0.0062, 0.0053, 0.0065, 0.0080], device='cuda:3'), out_proj_covar=tensor([1.0954e-04, 9.4264e-05, 9.9844e-05, 2.1022e-04, 9.9923e-05, 8.5402e-05, 1.1359e-04, 1.2123e-04], device='cuda:3') 2023-04-27 14:46:59,693 INFO [train.py:904] (3/8) Epoch 1, batch 6600, loss[loss=0.4124, simple_loss=0.4476, pruned_loss=0.1886, over 15360.00 frames. ], tot_loss[loss=0.351, simple_loss=0.4015, pruned_loss=0.1502, over 3064980.25 frames. ], batch size: 190, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:47:00,497 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-04-27 14:47:18,178 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.822e+02 5.266e+02 6.465e+02 8.044e+02 1.550e+03, threshold=1.293e+03, percent-clipped=3.0 2023-04-27 14:47:32,783 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:47:38,690 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5775, 4.8966, 4.7563, 4.8122, 4.7764, 5.2724, 5.0777, 4.7565], device='cuda:3'), covar=tensor([0.0622, 0.1025, 0.0907, 0.1274, 0.2033, 0.0634, 0.0625, 0.1757], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0202, 0.0167, 0.0173, 0.0215, 0.0163, 0.0157, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 14:47:43,244 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:10,684 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 6650, loss[loss=0.4454, simple_loss=0.4507, pruned_loss=0.22, over 11456.00 frames. ], tot_loss[loss=0.3541, simple_loss=0.403, pruned_loss=0.1526, over 3032107.93 frames. ], batch size: 247, lr: 3.88e-02, grad_scale: 16.0 2023-04-27 14:49:18,700 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:20,249 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 14:49:24,165 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:32,255 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:34,318 INFO [train.py:904] (3/8) Epoch 1, batch 6700, loss[loss=0.3756, simple_loss=0.4167, pruned_loss=0.1672, over 15424.00 frames. ], tot_loss[loss=0.3496, simple_loss=0.3989, pruned_loss=0.1501, over 3057441.91 frames. ], batch size: 190, lr: 3.87e-02, grad_scale: 16.0 2023-04-27 14:49:52,616 INFO [optim.py:368] (3/8) 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 14:49:54,865 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7519, 3.4296, 1.8764, 4.0391, 3.9956, 3.8896, 2.1085, 3.6827], device='cuda:3'), covar=tensor([0.2590, 0.0275, 0.2168, 0.0101, 0.0130, 0.0303, 0.1098, 0.0265], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0084, 0.0156, 0.0054, 0.0058, 0.0069, 0.0121, 0.0099], device='cuda:3'), out_proj_covar=tensor([2.1382e-04, 1.2303e-04, 2.0382e-04, 8.9333e-05, 9.7994e-05, 1.2471e-04, 1.7159e-04, 1.4409e-04], device='cuda:3') 2023-04-27 14:50:23,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5399, 5.2163, 4.9527, 5.1491, 5.0429, 5.4860, 5.3466, 5.0162], device='cuda:3'), covar=tensor([0.0636, 0.0932, 0.0900, 0.1146, 0.2133, 0.0749, 0.0670, 0.1773], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0200, 0.0167, 0.0175, 0.0218, 0.0167, 0.0160, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 14:50:45,161 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:50,989 INFO [train.py:904] (3/8) Epoch 1, batch 6750, loss[loss=0.3373, simple_loss=0.3931, pruned_loss=0.1408, over 16283.00 frames. ], tot_loss[loss=0.3476, simple_loss=0.3976, pruned_loss=0.1488, over 3077947.69 frames. ], batch size: 165, lr: 3.86e-02, grad_scale: 16.0 2023-04-27 14:51:20,616 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-27 14:51:50,790 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:05,969 INFO [train.py:904] (3/8) Epoch 1, batch 6800, loss[loss=0.4008, simple_loss=0.4354, pruned_loss=0.1831, over 15294.00 frames. ], tot_loss[loss=0.3462, simple_loss=0.3966, pruned_loss=0.1479, over 3074058.64 frames. ], batch size: 190, lr: 3.85e-02, grad_scale: 16.0 2023-04-27 14:52:07,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6902, 3.2272, 3.2148, 1.3507, 3.2410, 3.3481, 2.8094, 3.0590], device='cuda:3'), covar=tensor([0.0342, 0.0111, 0.0219, 0.2300, 0.0161, 0.0083, 0.0237, 0.0217], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0066, 0.0063, 0.0145, 0.0064, 0.0056, 0.0068, 0.0082], device='cuda:3'), out_proj_covar=tensor([1.1941e-04, 1.0269e-04, 1.0381e-04, 2.1737e-04, 1.0392e-04, 9.0182e-05, 1.2131e-04, 1.2517e-04], device='cuda:3') 2023-04-27 14:52:24,954 INFO [optim.py:368] (3/8) 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 14:52:33,262 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:52:41,650 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:46,850 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:06,062 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:23,173 INFO [train.py:904] (3/8) Epoch 1, batch 6850, loss[loss=0.3087, simple_loss=0.3901, pruned_loss=0.1136, over 16785.00 frames. ], tot_loss[loss=0.3468, simple_loss=0.3981, pruned_loss=0.1478, over 3102850.77 frames. ], batch size: 39, lr: 3.84e-02, grad_scale: 16.0 2023-04-27 14:53:35,299 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:46,326 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:57,788 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:03,512 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:54:09,396 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-27 14:54:17,822 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-27 14:54:32,616 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-27 14:54:37,392 INFO [train.py:904] (3/8) Epoch 1, batch 6900, loss[loss=0.4749, simple_loss=0.4684, pruned_loss=0.2407, over 11490.00 frames. ], tot_loss[loss=0.3457, simple_loss=0.3993, pruned_loss=0.1461, over 3119123.24 frames. ], batch size: 248, lr: 3.83e-02, grad_scale: 16.0 2023-04-27 14:54:47,053 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:48,611 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7169, 3.3633, 3.1368, 3.9306, 2.8483, 3.6457, 3.0278, 2.7965], device='cuda:3'), covar=tensor([0.0304, 0.0308, 0.0291, 0.0243, 0.1140, 0.0228, 0.0529, 0.0950], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0103, 0.0086, 0.0111, 0.0180, 0.0099, 0.0123, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:3') 2023-04-27 14:54:49,869 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:55,476 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.157e+02 5.195e+02 6.018e+02 7.768e+02 1.319e+03, threshold=1.204e+03, percent-clipped=1.0 2023-04-27 14:54:59,971 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:36,612 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:55:54,244 INFO [train.py:904] (3/8) Epoch 1, batch 6950, loss[loss=0.45, simple_loss=0.4574, pruned_loss=0.2213, over 11267.00 frames. ], tot_loss[loss=0.3496, simple_loss=0.4019, pruned_loss=0.1487, over 3121575.96 frames. ], batch size: 248, lr: 3.82e-02, grad_scale: 16.0 2023-04-27 14:56:25,067 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:25,204 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:47,683 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:11,695 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0139, 2.8928, 2.1811, 3.3272, 3.2854, 3.2760, 2.2472, 2.9732], device='cuda:3'), covar=tensor([0.1804, 0.0200, 0.1467, 0.0103, 0.0146, 0.0216, 0.0735, 0.0314], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0086, 0.0157, 0.0054, 0.0058, 0.0072, 0.0122, 0.0102], device='cuda:3'), out_proj_covar=tensor([2.1786e-04, 1.2802e-04, 2.0927e-04, 9.1047e-05, 1.0084e-04, 1.3175e-04, 1.7608e-04, 1.5144e-04], device='cuda:3') 2023-04-27 14:57:12,232 INFO [train.py:904] (3/8) Epoch 1, batch 7000, loss[loss=0.4665, simple_loss=0.4562, pruned_loss=0.2384, over 11757.00 frames. ], tot_loss[loss=0.3479, simple_loss=0.4012, pruned_loss=0.1473, over 3109086.52 frames. ], batch size: 250, lr: 3.81e-02, grad_scale: 16.0 2023-04-27 14:57:23,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3898, 3.2942, 3.1765, 2.9073, 3.3958, 2.6664, 3.0975, 3.1952], device='cuda:3'), covar=tensor([0.0081, 0.0065, 0.0084, 0.0281, 0.0055, 0.0444, 0.0085, 0.0087], device='cuda:3'), in_proj_covar=tensor([0.0047, 0.0037, 0.0053, 0.0074, 0.0040, 0.0074, 0.0051, 0.0052], device='cuda:3'), out_proj_covar=tensor([1.2392e-04, 9.6883e-05, 1.3714e-04, 1.7125e-04, 9.7482e-05, 1.7493e-04, 1.3301e-04, 1.5216e-04], device='cuda:3') 2023-04-27 14:57:30,859 INFO [optim.py:368] (3/8) 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 14:58:27,021 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:58:31,408 INFO [train.py:904] (3/8) Epoch 1, batch 7050, loss[loss=0.3246, simple_loss=0.3846, pruned_loss=0.1323, over 16775.00 frames. ], tot_loss[loss=0.3474, simple_loss=0.4013, pruned_loss=0.1467, over 3115825.75 frames. ], batch size: 83, lr: 3.80e-02, grad_scale: 16.0 2023-04-27 14:59:30,537 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2627, 1.5110, 1.8454, 1.3277, 1.6709, 1.8772, 2.1274, 1.9840], device='cuda:3'), covar=tensor([0.0053, 0.0344, 0.0150, 0.0277, 0.0127, 0.0214, 0.0092, 0.0140], device='cuda:3'), in_proj_covar=tensor([0.0030, 0.0064, 0.0043, 0.0046, 0.0039, 0.0046, 0.0033, 0.0035], device='cuda:3'), out_proj_covar=tensor([3.6616e-05, 9.4571e-05, 6.0126e-05, 6.4147e-05, 5.3347e-05, 6.3126e-05, 4.8918e-05, 4.9256e-05], device='cuda:3') 2023-04-27 14:59:51,865 INFO [train.py:904] (3/8) Epoch 1, batch 7100, loss[loss=0.3235, simple_loss=0.3815, pruned_loss=0.1328, over 16817.00 frames. ], tot_loss[loss=0.3463, simple_loss=0.3997, pruned_loss=0.1465, over 3109326.82 frames. ], batch size: 124, lr: 3.79e-02, grad_scale: 16.0 2023-04-27 15:00:05,752 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:00:08,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9890, 3.6083, 3.4112, 1.6380, 3.5924, 3.5884, 3.2314, 3.1387], device='cuda:3'), covar=tensor([0.0241, 0.0077, 0.0125, 0.1793, 0.0102, 0.0050, 0.0151, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0064, 0.0061, 0.0145, 0.0063, 0.0055, 0.0067, 0.0080], device='cuda:3'), out_proj_covar=tensor([1.2540e-04, 1.0128e-04, 1.0287e-04, 2.1836e-04, 1.0616e-04, 9.0185e-05, 1.2141e-04, 1.2632e-04], device='cuda:3') 2023-04-27 15:00:11,238 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.812e+02 5.623e+02 6.703e+02 8.193e+02 2.007e+03, threshold=1.341e+03, percent-clipped=3.0 2023-04-27 15:00:20,598 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:00:27,714 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:11,086 INFO [train.py:904] (3/8) Epoch 1, batch 7150, loss[loss=0.3871, simple_loss=0.4124, pruned_loss=0.1809, over 11443.00 frames. ], tot_loss[loss=0.3446, simple_loss=0.3972, pruned_loss=0.146, over 3077689.82 frames. ], batch size: 248, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:01:30,075 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:34,845 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:02:06,321 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7515, 3.6481, 3.6718, 4.1139, 4.0463, 4.0082, 4.0891, 3.9731], device='cuda:3'), covar=tensor([0.0389, 0.0336, 0.1079, 0.0321, 0.0395, 0.0355, 0.0275, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0159, 0.0250, 0.0177, 0.0151, 0.0161, 0.0133, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:02:21,100 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9535, 4.0887, 3.6889, 1.7726, 2.9431, 2.3305, 3.3753, 4.2431], device='cuda:3'), covar=tensor([0.0330, 0.0383, 0.0264, 0.2228, 0.0942, 0.1347, 0.0824, 0.0226], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0073, 0.0111, 0.0155, 0.0148, 0.0143, 0.0137, 0.0068], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 15:02:27,368 INFO [train.py:904] (3/8) Epoch 1, batch 7200, loss[loss=0.3254, simple_loss=0.3774, pruned_loss=0.1367, over 11752.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3934, pruned_loss=0.1423, over 3101043.62 frames. ], batch size: 247, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:02:46,726 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.108e+02 4.516e+02 5.501e+02 7.177e+02 1.508e+03, threshold=1.100e+03, percent-clipped=3.0 2023-04-27 15:03:02,527 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:03:19,571 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:03:42,563 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9672, 1.7437, 1.7533, 2.2039, 2.1869, 2.3579, 1.8072, 2.3769], device='cuda:3'), covar=tensor([0.0094, 0.0623, 0.0292, 0.0180, 0.0115, 0.0130, 0.0229, 0.0080], device='cuda:3'), in_proj_covar=tensor([0.0041, 0.0091, 0.0069, 0.0050, 0.0042, 0.0041, 0.0056, 0.0037], device='cuda:3'), out_proj_covar=tensor([7.6156e-05, 1.6285e-04, 1.2929e-04, 9.2548e-05, 7.4211e-05, 7.5200e-05, 9.4184e-05, 6.7448e-05], device='cuda:3') 2023-04-27 15:03:46,512 INFO [train.py:904] (3/8) Epoch 1, batch 7250, loss[loss=0.2892, simple_loss=0.345, pruned_loss=0.1167, over 16655.00 frames. ], tot_loss[loss=0.3354, simple_loss=0.3905, pruned_loss=0.1402, over 3114149.00 frames. ], batch size: 57, lr: 3.77e-02, grad_scale: 8.0 2023-04-27 15:04:06,709 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:21,273 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:35,838 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:58,720 INFO [train.py:904] (3/8) Epoch 1, batch 7300, loss[loss=0.3555, simple_loss=0.407, pruned_loss=0.152, over 15493.00 frames. ], tot_loss[loss=0.334, simple_loss=0.3892, pruned_loss=0.1394, over 3110495.46 frames. ], batch size: 191, lr: 3.76e-02, grad_scale: 8.0 2023-04-27 15:05:09,583 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6929, 3.6026, 2.2569, 4.3144, 4.2678, 4.1159, 2.5167, 3.7439], device='cuda:3'), covar=tensor([0.2793, 0.0332, 0.1866, 0.0116, 0.0119, 0.0234, 0.1052, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0087, 0.0156, 0.0056, 0.0061, 0.0074, 0.0125, 0.0103], device='cuda:3'), out_proj_covar=tensor([2.1756e-04, 1.3354e-04, 2.1337e-04, 9.5853e-05, 1.0666e-04, 1.3805e-04, 1.8261e-04, 1.5544e-04], device='cuda:3') 2023-04-27 15:05:19,383 INFO [optim.py:368] (3/8) 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 15:05:49,054 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:53,717 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 7350, loss[loss=0.3307, simple_loss=0.3901, pruned_loss=0.1356, over 16844.00 frames. ], tot_loss[loss=0.3315, simple_loss=0.387, pruned_loss=0.138, over 3085847.84 frames. ], batch size: 102, lr: 3.75e-02, grad_scale: 8.0 2023-04-27 15:07:18,983 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3486, 3.2078, 3.1526, 3.3307, 2.9792, 3.2789, 3.1859, 3.0550], device='cuda:3'), covar=tensor([0.0234, 0.0136, 0.0171, 0.0124, 0.0497, 0.0161, 0.0655, 0.0214], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0062, 0.0116, 0.0090, 0.0141, 0.0091, 0.0083, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:07:30,954 INFO [train.py:904] (3/8) Epoch 1, batch 7400, loss[loss=0.3735, simple_loss=0.4301, pruned_loss=0.1584, over 16846.00 frames. ], tot_loss[loss=0.3352, simple_loss=0.3895, pruned_loss=0.1405, over 3070951.42 frames. ], batch size: 42, lr: 3.74e-02, grad_scale: 8.0 2023-04-27 15:07:31,458 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8088, 4.0383, 2.2317, 4.5951, 4.6595, 4.3544, 2.8118, 3.7079], device='cuda:3'), covar=tensor([0.2255, 0.0223, 0.1704, 0.0073, 0.0078, 0.0189, 0.0775, 0.0347], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0086, 0.0156, 0.0056, 0.0060, 0.0074, 0.0125, 0.0103], device='cuda:3'), out_proj_covar=tensor([2.1880e-04, 1.3255e-04, 2.1372e-04, 9.5277e-05, 1.0567e-04, 1.3975e-04, 1.8347e-04, 1.5597e-04], device='cuda:3') 2023-04-27 15:07:32,603 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7308, 4.3490, 4.5124, 4.7304, 3.8768, 4.4446, 4.5893, 4.1313], device='cuda:3'), covar=tensor([0.0240, 0.0147, 0.0158, 0.0099, 0.0835, 0.0232, 0.0144, 0.0231], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0063, 0.0117, 0.0091, 0.0142, 0.0092, 0.0083, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:07:36,106 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:07:40,545 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:08:40,282 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8556, 3.9063, 3.6354, 1.7858, 2.6916, 2.0589, 3.1877, 3.9419], device='cuda:3'), covar=tensor([0.0274, 0.0239, 0.0257, 0.2130, 0.1047, 0.1387, 0.0778, 0.0219], device='cuda:3'), in_proj_covar=tensor([0.0100, 0.0073, 0.0113, 0.0153, 0.0147, 0.0138, 0.0136, 0.0067], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 15:08:52,555 INFO [train.py:904] (3/8) Epoch 1, batch 7450, loss[loss=0.4376, simple_loss=0.4447, pruned_loss=0.2153, over 11063.00 frames. ], tot_loss[loss=0.3378, simple_loss=0.3915, pruned_loss=0.1421, over 3077713.87 frames. ], batch size: 250, lr: 3.73e-02, grad_scale: 8.0 2023-04-27 15:09:22,497 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:10:15,193 INFO [train.py:904] (3/8) Epoch 1, batch 7500, loss[loss=0.3196, simple_loss=0.387, pruned_loss=0.1261, over 16898.00 frames. ], tot_loss[loss=0.3389, simple_loss=0.3931, pruned_loss=0.1424, over 3076130.36 frames. ], batch size: 96, lr: 3.72e-02, grad_scale: 8.0 2023-04-27 15:10:35,038 INFO [optim.py:368] (3/8) 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 15:10:43,189 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:10:44,609 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5579, 3.2635, 2.9590, 3.1282, 2.5945, 2.1001, 3.6602, 4.0680], device='cuda:3'), covar=tensor([0.1753, 0.0747, 0.1005, 0.0404, 0.1795, 0.1256, 0.0293, 0.0068], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0159, 0.0188, 0.0114, 0.0198, 0.0147, 0.0123, 0.0066], device='cuda:3'), out_proj_covar=tensor([2.4046e-04, 1.9057e-04, 2.0618e-04, 1.3407e-04, 2.3811e-04, 1.7537e-04, 1.5044e-04, 8.2996e-05], device='cuda:3') 2023-04-27 15:11:06,850 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:11:31,891 INFO [train.py:904] (3/8) Epoch 1, batch 7550, loss[loss=0.3176, simple_loss=0.3797, pruned_loss=0.1278, over 16770.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3927, pruned_loss=0.1427, over 3064513.25 frames. ], batch size: 102, lr: 3.72e-02, grad_scale: 4.0 2023-04-27 15:11:54,176 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:12:02,247 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7700, 3.0270, 3.2919, 3.2686, 3.3077, 3.0239, 2.8733, 3.1764], device='cuda:3'), covar=tensor([0.0535, 0.0553, 0.0606, 0.0626, 0.0655, 0.0574, 0.1075, 0.0469], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0121, 0.0140, 0.0135, 0.0151, 0.0125, 0.0174, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:12:21,505 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:12:28,301 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6339, 3.5861, 3.5597, 3.9380, 3.9077, 3.7866, 3.8724, 3.8115], device='cuda:3'), covar=tensor([0.0382, 0.0342, 0.0939, 0.0297, 0.0358, 0.0404, 0.0315, 0.0325], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0167, 0.0257, 0.0182, 0.0156, 0.0159, 0.0136, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:12:50,413 INFO [train.py:904] (3/8) Epoch 1, batch 7600, loss[loss=0.3254, simple_loss=0.3941, pruned_loss=0.1283, over 16727.00 frames. ], tot_loss[loss=0.3378, simple_loss=0.3913, pruned_loss=0.1422, over 3077660.42 frames. ], batch size: 39, lr: 3.71e-02, grad_scale: 8.0 2023-04-27 15:13:10,047 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:12,790 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.798e+02 5.452e+02 6.510e+02 8.243e+02 1.443e+03, threshold=1.302e+03, percent-clipped=3.0 2023-04-27 15:13:29,726 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:30,013 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-04-27 15:13:40,334 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:14:13,729 INFO [train.py:904] (3/8) Epoch 1, batch 7650, loss[loss=0.3531, simple_loss=0.4034, pruned_loss=0.1514, over 16358.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3921, pruned_loss=0.143, over 3079239.48 frames. ], batch size: 146, lr: 3.70e-02, grad_scale: 8.0 2023-04-27 15:15:14,477 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 7700, loss[loss=0.3313, simple_loss=0.3872, pruned_loss=0.1377, over 16325.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3918, pruned_loss=0.1432, over 3085562.52 frames. ], batch size: 165, lr: 3.69e-02, grad_scale: 8.0 2023-04-27 15:15:45,480 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:16:00,933 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.630e+02 5.474e+02 6.817e+02 8.632e+02 3.010e+03, threshold=1.363e+03, percent-clipped=1.0 2023-04-27 15:16:45,684 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:57,056 INFO [train.py:904] (3/8) Epoch 1, batch 7750, loss[loss=0.3078, simple_loss=0.3739, pruned_loss=0.1208, over 16722.00 frames. ], tot_loss[loss=0.3387, simple_loss=0.3918, pruned_loss=0.1429, over 3094968.97 frames. ], batch size: 76, lr: 3.68e-02, grad_scale: 8.0 2023-04-27 15:16:59,252 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:17:16,058 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:17:28,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3044, 4.3977, 4.0507, 1.8566, 3.5488, 2.6587, 3.7810, 4.5606], device='cuda:3'), covar=tensor([0.0204, 0.0222, 0.0280, 0.2072, 0.0644, 0.1147, 0.0661, 0.0147], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0076, 0.0115, 0.0152, 0.0147, 0.0139, 0.0136, 0.0067], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 15:18:14,288 INFO [train.py:904] (3/8) Epoch 1, batch 7800, loss[loss=0.431, simple_loss=0.4399, pruned_loss=0.2111, over 11504.00 frames. ], tot_loss[loss=0.3419, simple_loss=0.3937, pruned_loss=0.1451, over 3079214.96 frames. ], batch size: 248, lr: 3.67e-02, grad_scale: 8.0 2023-04-27 15:18:16,667 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:19,959 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:36,278 INFO [optim.py:368] (3/8) 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 15:18:43,903 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:19:20,100 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7306, 3.1625, 3.0688, 2.3102, 2.9843, 3.1094, 3.2532, 1.7151], device='cuda:3'), covar=tensor([0.1282, 0.0116, 0.0090, 0.0462, 0.0123, 0.0102, 0.0101, 0.0821], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0043, 0.0044, 0.0075, 0.0043, 0.0042, 0.0045, 0.0082], device='cuda:3'), out_proj_covar=tensor([2.2163e-04, 9.3268e-05, 1.0002e-04, 1.5827e-04, 9.5303e-05, 9.8887e-05, 9.5864e-05, 1.7100e-04], device='cuda:3') 2023-04-27 15:19:31,859 INFO [train.py:904] (3/8) Epoch 1, batch 7850, loss[loss=0.3542, simple_loss=0.4034, pruned_loss=0.1525, over 16726.00 frames. ], tot_loss[loss=0.3433, simple_loss=0.3951, pruned_loss=0.1458, over 3059276.67 frames. ], batch size: 124, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:19:51,342 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:20:49,083 INFO [train.py:904] (3/8) Epoch 1, batch 7900, loss[loss=0.3322, simple_loss=0.4012, pruned_loss=0.1316, over 16407.00 frames. ], tot_loss[loss=0.3414, simple_loss=0.394, pruned_loss=0.1444, over 3066706.63 frames. ], batch size: 146, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:21:11,994 INFO [optim.py:368] (3/8) 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 15:21:13,855 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1963, 4.1206, 3.8240, 1.7757, 3.0765, 2.4234, 3.4502, 4.3644], device='cuda:3'), covar=tensor([0.0317, 0.0327, 0.0290, 0.2165, 0.0874, 0.1238, 0.0861, 0.0251], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0076, 0.0115, 0.0152, 0.0146, 0.0138, 0.0137, 0.0069], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-27 15:21:39,273 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 7950, loss[loss=0.3886, simple_loss=0.41, pruned_loss=0.1836, over 11712.00 frames. ], tot_loss[loss=0.34, simple_loss=0.393, pruned_loss=0.1435, over 3080104.12 frames. ], batch size: 246, lr: 3.65e-02, grad_scale: 8.0 2023-04-27 15:22:50,379 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-27 15:22:54,877 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:30,885 INFO [train.py:904] (3/8) Epoch 1, batch 8000, loss[loss=0.3431, simple_loss=0.405, pruned_loss=0.1406, over 16277.00 frames. ], tot_loss[loss=0.3393, simple_loss=0.3925, pruned_loss=0.1431, over 3091106.28 frames. ], batch size: 165, lr: 3.64e-02, grad_scale: 8.0 2023-04-27 15:23:51,330 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.520e+02 5.080e+02 6.216e+02 7.580e+02 1.182e+03, threshold=1.243e+03, percent-clipped=0.0 2023-04-27 15:24:45,916 INFO [train.py:904] (3/8) Epoch 1, batch 8050, loss[loss=0.3484, simple_loss=0.3975, pruned_loss=0.1496, over 16691.00 frames. ], tot_loss[loss=0.3384, simple_loss=0.3917, pruned_loss=0.1425, over 3095976.89 frames. ], batch size: 134, lr: 3.63e-02, grad_scale: 8.0 2023-04-27 15:25:04,631 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:08,622 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-04-27 15:25:54,880 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-27 15:25:57,836 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 8100, loss[loss=0.3358, simple_loss=0.3876, pruned_loss=0.142, over 16923.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.39, pruned_loss=0.1406, over 3111859.57 frames. ], batch size: 109, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:26:15,157 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:22,779 INFO [optim.py:368] (3/8) 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 15:26:31,436 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9017, 1.6365, 1.7931, 1.5603, 2.4285, 2.5346, 2.7258, 3.0707], device='cuda:3'), covar=tensor([0.0034, 0.0325, 0.0156, 0.0266, 0.0117, 0.0177, 0.0109, 0.0103], device='cuda:3'), in_proj_covar=tensor([0.0031, 0.0068, 0.0051, 0.0054, 0.0047, 0.0055, 0.0034, 0.0039], device='cuda:3'), out_proj_covar=tensor([4.1845e-05, 1.0078e-04, 7.4248e-05, 8.0889e-05, 6.9954e-05, 7.8378e-05, 5.4155e-05, 6.0412e-05], device='cuda:3') 2023-04-27 15:26:47,995 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-27 15:27:16,440 INFO [train.py:904] (3/8) Epoch 1, batch 8150, loss[loss=0.2947, simple_loss=0.3565, pruned_loss=0.1165, over 17228.00 frames. ], tot_loss[loss=0.3334, simple_loss=0.3876, pruned_loss=0.1396, over 3108457.40 frames. ], batch size: 45, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:27:23,262 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9322, 4.4872, 4.2351, 2.0688, 4.4668, 4.4730, 3.4535, 4.0385], device='cuda:3'), covar=tensor([0.0370, 0.0064, 0.0102, 0.1905, 0.0069, 0.0054, 0.0282, 0.0124], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0064, 0.0067, 0.0148, 0.0066, 0.0056, 0.0074, 0.0087], device='cuda:3'), out_proj_covar=tensor([1.4699e-04, 1.0991e-04, 1.1626e-04, 2.3398e-04, 1.1662e-04, 9.9409e-05, 1.3866e-04, 1.4492e-04], device='cuda:3') 2023-04-27 15:27:27,357 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 8200, loss[loss=0.3977, simple_loss=0.4137, pruned_loss=0.1908, over 11371.00 frames. ], tot_loss[loss=0.3323, simple_loss=0.3857, pruned_loss=0.1394, over 3100475.61 frames. ], batch size: 247, lr: 3.61e-02, grad_scale: 4.0 2023-04-27 15:28:56,620 INFO [optim.py:368] (3/8) 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 15:29:53,197 INFO [train.py:904] (3/8) Epoch 1, batch 8250, loss[loss=0.3073, simple_loss=0.375, pruned_loss=0.1198, over 16771.00 frames. ], tot_loss[loss=0.3291, simple_loss=0.3844, pruned_loss=0.1369, over 3080740.32 frames. ], batch size: 124, lr: 3.60e-02, grad_scale: 4.0 2023-04-27 15:30:15,880 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:19,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6336, 4.4405, 4.4433, 4.8645, 4.8278, 4.5706, 4.8499, 4.7481], device='cuda:3'), covar=tensor([0.0334, 0.0312, 0.0866, 0.0291, 0.0314, 0.0295, 0.0263, 0.0205], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0164, 0.0250, 0.0177, 0.0151, 0.0153, 0.0140, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:30:41,319 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:47,907 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8919, 1.5625, 1.4580, 1.2694, 1.6574, 1.7416, 1.7477, 1.8688], device='cuda:3'), covar=tensor([0.0034, 0.0192, 0.0111, 0.0147, 0.0069, 0.0089, 0.0065, 0.0067], device='cuda:3'), in_proj_covar=tensor([0.0032, 0.0071, 0.0054, 0.0057, 0.0047, 0.0057, 0.0034, 0.0039], device='cuda:3'), out_proj_covar=tensor([4.2578e-05, 1.0545e-04, 7.8850e-05, 8.5432e-05, 7.1161e-05, 8.2099e-05, 5.5842e-05, 6.2321e-05], device='cuda:3') 2023-04-27 15:30:58,376 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 15:31:13,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7593, 4.6340, 4.3262, 4.0071, 4.5830, 2.6928, 4.2734, 4.5402], device='cuda:3'), covar=tensor([0.0071, 0.0050, 0.0062, 0.0248, 0.0049, 0.0830, 0.0061, 0.0078], device='cuda:3'), in_proj_covar=tensor([0.0047, 0.0038, 0.0054, 0.0074, 0.0040, 0.0084, 0.0054, 0.0052], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:31:14,606 INFO [train.py:904] (3/8) Epoch 1, batch 8300, loss[loss=0.2636, simple_loss=0.347, pruned_loss=0.09007, over 16565.00 frames. ], tot_loss[loss=0.3204, simple_loss=0.3789, pruned_loss=0.1309, over 3079770.40 frames. ], batch size: 75, lr: 3.59e-02, grad_scale: 4.0 2023-04-27 15:31:40,147 INFO [optim.py:368] (3/8) 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 15:31:52,868 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0431, 5.3002, 5.0364, 5.1901, 4.6114, 4.7028, 4.8294, 5.3222], device='cuda:3'), covar=tensor([0.0322, 0.0519, 0.0625, 0.0273, 0.0546, 0.0330, 0.0388, 0.0410], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0210, 0.0199, 0.0130, 0.0160, 0.0131, 0.0182, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:31:56,696 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:58,401 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-27 15:31:59,949 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:08,489 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 15:32:12,371 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-27 15:32:36,033 INFO [train.py:904] (3/8) Epoch 1, batch 8350, loss[loss=0.3052, simple_loss=0.3546, pruned_loss=0.1279, over 12147.00 frames. ], tot_loss[loss=0.314, simple_loss=0.3755, pruned_loss=0.1263, over 3075379.33 frames. ], batch size: 246, lr: 3.58e-02, grad_scale: 4.0 2023-04-27 15:33:55,087 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:57,932 INFO [train.py:904] (3/8) Epoch 1, batch 8400, loss[loss=0.3353, simple_loss=0.379, pruned_loss=0.1458, over 12168.00 frames. ], tot_loss[loss=0.3079, simple_loss=0.3708, pruned_loss=0.1225, over 3052600.02 frames. ], batch size: 248, lr: 3.58e-02, grad_scale: 8.0 2023-04-27 15:34:21,544 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.723e+02 4.252e+02 5.243e+02 6.006e+02 1.213e+03, threshold=1.049e+03, percent-clipped=1.0 2023-04-27 15:35:12,364 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:18,229 INFO [train.py:904] (3/8) Epoch 1, batch 8450, loss[loss=0.2721, simple_loss=0.3484, pruned_loss=0.09789, over 16850.00 frames. ], tot_loss[loss=0.302, simple_loss=0.3671, pruned_loss=0.1185, over 3073801.89 frames. ], batch size: 83, lr: 3.57e-02, grad_scale: 8.0 2023-04-27 15:35:30,874 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:36:22,779 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7721, 1.8914, 1.9817, 2.6919, 2.9280, 2.7449, 1.9076, 2.8707], device='cuda:3'), covar=tensor([0.0039, 0.0376, 0.0228, 0.0101, 0.0038, 0.0089, 0.0189, 0.0061], device='cuda:3'), in_proj_covar=tensor([0.0043, 0.0091, 0.0072, 0.0053, 0.0041, 0.0045, 0.0064, 0.0041], device='cuda:3'), out_proj_covar=tensor([8.0999e-05, 1.6472e-04, 1.3922e-04, 1.0124e-04, 7.4259e-05, 8.4931e-05, 1.1002e-04, 7.5196e-05], device='cuda:3') 2023-04-27 15:36:39,471 INFO [train.py:904] (3/8) Epoch 1, batch 8500, loss[loss=0.2356, simple_loss=0.3136, pruned_loss=0.07882, over 16356.00 frames. ], tot_loss[loss=0.2942, simple_loss=0.3608, pruned_loss=0.1138, over 3056778.01 frames. ], batch size: 146, lr: 3.56e-02, grad_scale: 8.0 2023-04-27 15:36:47,417 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4701, 3.5235, 3.1027, 3.2263, 2.7577, 1.9914, 3.6890, 4.1223], device='cuda:3'), covar=tensor([0.1843, 0.0583, 0.0905, 0.0318, 0.1515, 0.1418, 0.0226, 0.0057], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0167, 0.0191, 0.0116, 0.0180, 0.0153, 0.0127, 0.0067], device='cuda:3'), out_proj_covar=tensor([2.4427e-04, 1.9715e-04, 2.0946e-04, 1.3653e-04, 2.1794e-04, 1.8485e-04, 1.5408e-04, 8.6649e-05], device='cuda:3') 2023-04-27 15:36:49,162 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:37:31,570 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3756, 4.0756, 3.9540, 3.6561, 4.0043, 2.2238, 3.8086, 4.1111], device='cuda:3'), covar=tensor([0.0078, 0.0079, 0.0085, 0.0270, 0.0078, 0.1006, 0.0102, 0.0097], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0039, 0.0056, 0.0071, 0.0041, 0.0089, 0.0055, 0.0052], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:38:03,990 INFO [train.py:904] (3/8) Epoch 1, batch 8550, loss[loss=0.2677, simple_loss=0.3306, pruned_loss=0.1023, over 11948.00 frames. ], tot_loss[loss=0.29, simple_loss=0.3573, pruned_loss=0.1114, over 3058849.74 frames. ], batch size: 248, lr: 3.55e-02, grad_scale: 8.0 2023-04-27 15:38:44,828 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7335, 2.6213, 1.5449, 2.7266, 2.1220, 2.6683, 1.9676, 2.4217], device='cuda:3'), covar=tensor([0.0087, 0.0172, 0.1366, 0.0092, 0.0675, 0.0284, 0.1027, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0061, 0.0067, 0.0146, 0.0064, 0.0122, 0.0074, 0.0148, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:39:28,153 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 15:39:44,565 INFO [train.py:904] (3/8) Epoch 1, batch 8600, loss[loss=0.2896, simple_loss=0.3666, pruned_loss=0.1064, over 16208.00 frames. ], tot_loss[loss=0.2892, simple_loss=0.3575, pruned_loss=0.1104, over 3047759.83 frames. ], batch size: 165, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:40:17,874 INFO [optim.py:368] (3/8) 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 15:40:27,931 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:41:26,180 INFO [train.py:904] (3/8) Epoch 1, batch 8650, loss[loss=0.229, simple_loss=0.3195, pruned_loss=0.0693, over 16621.00 frames. ], tot_loss[loss=0.2848, simple_loss=0.3542, pruned_loss=0.1077, over 3030776.37 frames. ], batch size: 134, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:41:54,287 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3965, 4.0303, 3.9541, 3.5794, 4.1255, 1.8731, 3.8080, 3.9750], device='cuda:3'), covar=tensor([0.0082, 0.0089, 0.0072, 0.0290, 0.0062, 0.1175, 0.0091, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0043, 0.0037, 0.0052, 0.0067, 0.0039, 0.0084, 0.0051, 0.0049], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-27 15:42:50,347 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-27 15:43:13,564 INFO [train.py:904] (3/8) Epoch 1, batch 8700, loss[loss=0.2507, simple_loss=0.3327, pruned_loss=0.08436, over 16779.00 frames. ], tot_loss[loss=0.2799, simple_loss=0.3497, pruned_loss=0.105, over 3031886.07 frames. ], batch size: 124, lr: 3.53e-02, grad_scale: 8.0 2023-04-27 15:43:28,813 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8643, 5.3418, 5.3289, 5.2918, 5.3499, 5.7462, 5.6844, 5.2108], device='cuda:3'), covar=tensor([0.0457, 0.1048, 0.0758, 0.1230, 0.1691, 0.0746, 0.0725, 0.1759], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0206, 0.0168, 0.0174, 0.0211, 0.0174, 0.0149, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-27 15:43:41,220 INFO [optim.py:368] (3/8) 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 15:44:49,181 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2790, 4.6174, 4.2967, 4.5883, 4.0717, 4.3082, 4.2953, 4.5515], device='cuda:3'), covar=tensor([0.0466, 0.0593, 0.0751, 0.0247, 0.0528, 0.0438, 0.0448, 0.0524], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0200, 0.0179, 0.0118, 0.0148, 0.0126, 0.0167, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 15:44:49,920 INFO [train.py:904] (3/8) Epoch 1, batch 8750, loss[loss=0.3223, simple_loss=0.3844, pruned_loss=0.1301, over 15405.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3485, pruned_loss=0.1034, over 3038002.27 frames. ], batch size: 191, lr: 3.52e-02, grad_scale: 8.0 2023-04-27 15:45:14,136 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:46:36,144 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 15:46:42,405 INFO [train.py:904] (3/8) Epoch 1, batch 8800, loss[loss=0.2813, simple_loss=0.3382, pruned_loss=0.1122, over 12246.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3461, pruned_loss=0.1016, over 3045221.34 frames. ], batch size: 246, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:46:44,041 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5241, 3.3797, 3.4117, 2.8428, 3.2781, 3.3079, 3.3387, 1.7542], device='cuda:3'), covar=tensor([0.1378, 0.0082, 0.0083, 0.0375, 0.0070, 0.0098, 0.0065, 0.0852], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0044, 0.0049, 0.0082, 0.0045, 0.0048, 0.0047, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:3') 2023-04-27 15:47:13,493 INFO [optim.py:368] (3/8) 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 15:47:25,114 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:47:56,143 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6031, 1.6673, 2.1733, 2.9531, 2.7239, 2.7565, 1.4462, 3.2142], device='cuda:3'), covar=tensor([0.0056, 0.0468, 0.0242, 0.0073, 0.0056, 0.0094, 0.0343, 0.0034], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0091, 0.0070, 0.0053, 0.0042, 0.0043, 0.0068, 0.0039], device='cuda:3'), out_proj_covar=tensor([8.4273e-05, 1.6569e-04, 1.3586e-04, 9.8346e-05, 7.5484e-05, 8.1824e-05, 1.1716e-04, 7.2337e-05], device='cuda:3') 2023-04-27 15:48:27,335 INFO [train.py:904] (3/8) Epoch 1, batch 8850, loss[loss=0.2228, simple_loss=0.3048, pruned_loss=0.07043, over 12646.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3469, pruned_loss=0.09955, over 3037062.95 frames. ], batch size: 248, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:50:13,540 INFO [train.py:904] (3/8) Epoch 1, batch 8900, loss[loss=0.2717, simple_loss=0.3363, pruned_loss=0.1036, over 12094.00 frames. ], tot_loss[loss=0.2704, simple_loss=0.3461, pruned_loss=0.09739, over 3049503.79 frames. ], batch size: 246, lr: 3.50e-02, grad_scale: 8.0 2023-04-27 15:50:31,258 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-27 15:50:33,889 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-27 15:50:42,917 INFO [optim.py:368] (3/8) 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 15:50:56,785 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:07,521 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3453, 4.2019, 1.7162, 4.1728, 2.6365, 4.0099, 2.3509, 3.0971], device='cuda:3'), covar=tensor([0.0030, 0.0095, 0.1746, 0.0032, 0.0852, 0.0234, 0.1365, 0.0575], device='cuda:3'), in_proj_covar=tensor([0.0064, 0.0070, 0.0153, 0.0068, 0.0132, 0.0080, 0.0160, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 15:51:13,329 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:33,164 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-27 15:52:19,950 INFO [train.py:904] (3/8) Epoch 1, batch 8950, loss[loss=0.2501, simple_loss=0.324, pruned_loss=0.08807, over 15246.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3468, pruned_loss=0.09845, over 3061009.01 frames. ], batch size: 190, lr: 3.49e-02, grad_scale: 8.0 2023-04-27 15:53:00,567 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:38,140 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:42,499 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 9000, loss[loss=0.2852, simple_loss=0.3449, pruned_loss=0.1128, over 12387.00 frames. ], tot_loss[loss=0.2679, simple_loss=0.3433, pruned_loss=0.09625, over 3065221.66 frames. ], batch size: 248, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:54:08,282 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 15:54:19,198 INFO [train.py:938] (3/8) Epoch 1, validation: loss=0.2299, simple_loss=0.3267, pruned_loss=0.06658, over 944034.00 frames. 2023-04-27 15:54:19,199 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 15:54:30,507 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-04-27 15:54:46,209 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2115, 3.9627, 4.0037, 4.2068, 3.5081, 4.1504, 4.0254, 3.7935], device='cuda:3'), covar=tensor([0.0242, 0.0166, 0.0181, 0.0110, 0.0685, 0.0154, 0.0266, 0.0208], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0065, 0.0117, 0.0089, 0.0138, 0.0090, 0.0081, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:54:46,238 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:55:56,619 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:56:02,940 INFO [train.py:904] (3/8) Epoch 1, batch 9050, loss[loss=0.2604, simple_loss=0.3304, pruned_loss=0.09518, over 15290.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3456, pruned_loss=0.09811, over 3061327.28 frames. ], batch size: 191, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:56:47,246 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 15:56:48,438 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:20,038 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 15:57:45,072 INFO [train.py:904] (3/8) Epoch 1, batch 9100, loss[loss=0.2797, simple_loss=0.3606, pruned_loss=0.09943, over 15263.00 frames. ], tot_loss[loss=0.2722, simple_loss=0.3458, pruned_loss=0.09936, over 3045436.93 frames. ], batch size: 191, lr: 3.47e-02, grad_scale: 8.0 2023-04-27 15:58:14,291 INFO [optim.py:368] (3/8) 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 15:58:14,875 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:58:17,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4637, 4.7071, 4.7438, 4.6608, 4.7752, 5.1426, 4.9727, 4.6145], device='cuda:3'), covar=tensor([0.0566, 0.1067, 0.0789, 0.1299, 0.1736, 0.0776, 0.0630, 0.1568], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0209, 0.0178, 0.0182, 0.0222, 0.0182, 0.0152, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-27 15:58:20,690 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2802, 3.2027, 3.2253, 3.5096, 3.4144, 3.2401, 3.3838, 3.4194], device='cuda:3'), covar=tensor([0.0347, 0.0293, 0.0853, 0.0304, 0.0440, 0.0828, 0.0492, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0162, 0.0243, 0.0171, 0.0145, 0.0154, 0.0135, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 15:58:54,577 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1046, 3.0958, 1.2507, 2.8814, 2.0575, 3.0223, 1.6368, 2.3720], device='cuda:3'), covar=tensor([0.0089, 0.0135, 0.1818, 0.0135, 0.0973, 0.0216, 0.1580, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0063, 0.0071, 0.0155, 0.0070, 0.0135, 0.0080, 0.0163, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 15:59:42,636 INFO [train.py:904] (3/8) Epoch 1, batch 9150, loss[loss=0.2885, simple_loss=0.3549, pruned_loss=0.1111, over 16173.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3463, pruned_loss=0.099, over 3042554.32 frames. ], batch size: 165, lr: 3.46e-02, grad_scale: 8.0 2023-04-27 16:00:33,108 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6829, 2.7248, 2.5872, 2.1919, 2.6649, 2.5430, 2.6210, 1.8152], device='cuda:3'), covar=tensor([0.0896, 0.0095, 0.0096, 0.0372, 0.0095, 0.0135, 0.0080, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0045, 0.0049, 0.0084, 0.0046, 0.0048, 0.0049, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:3') 2023-04-27 16:01:09,234 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3013, 1.4561, 1.8430, 2.1358, 2.4347, 2.2716, 1.5376, 2.1756], device='cuda:3'), covar=tensor([0.0081, 0.0661, 0.0273, 0.0212, 0.0063, 0.0202, 0.0366, 0.0154], device='cuda:3'), in_proj_covar=tensor([0.0050, 0.0093, 0.0073, 0.0055, 0.0042, 0.0044, 0.0071, 0.0041], device='cuda:3'), out_proj_covar=tensor([9.4323e-05, 1.6781e-04, 1.4001e-04, 1.0267e-04, 7.4394e-05, 8.3527e-05, 1.2228e-04, 7.5503e-05], device='cuda:3') 2023-04-27 16:01:27,877 INFO [train.py:904] (3/8) Epoch 1, batch 9200, loss[loss=0.2685, simple_loss=0.3371, pruned_loss=0.0999, over 16715.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3405, pruned_loss=0.09703, over 3035111.30 frames. ], batch size: 134, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:01:54,927 INFO [optim.py:368] (3/8) 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 16:02:59,671 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7575, 2.0334, 2.2612, 1.5414, 2.4629, 2.4769, 2.9325, 2.9011], device='cuda:3'), covar=tensor([0.0039, 0.0257, 0.0145, 0.0285, 0.0104, 0.0186, 0.0059, 0.0067], device='cuda:3'), in_proj_covar=tensor([0.0033, 0.0072, 0.0056, 0.0063, 0.0050, 0.0061, 0.0033, 0.0043], device='cuda:3'), out_proj_covar=tensor([4.5374e-05, 1.1021e-04, 8.4451e-05, 9.6736e-05, 7.6567e-05, 9.1185e-05, 5.1277e-05, 7.0089e-05], device='cuda:3') 2023-04-27 16:03:04,216 INFO [train.py:904] (3/8) Epoch 1, batch 9250, loss[loss=0.2677, simple_loss=0.3413, pruned_loss=0.09705, over 16628.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3401, pruned_loss=0.09697, over 3055266.93 frames. ], batch size: 134, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:03:41,025 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3737, 3.4374, 2.8819, 3.1483, 2.3024, 1.9354, 3.5028, 3.8931], device='cuda:3'), covar=tensor([0.1916, 0.0684, 0.0998, 0.0420, 0.1820, 0.1664, 0.0328, 0.0082], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0185, 0.0206, 0.0128, 0.0173, 0.0162, 0.0142, 0.0076], device='cuda:3'), out_proj_covar=tensor([2.5632e-04, 2.1702e-04, 2.2586e-04, 1.4955e-04, 2.0897e-04, 1.9579e-04, 1.7116e-04, 9.6710e-05], device='cuda:3') 2023-04-27 16:04:16,908 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:04:58,140 INFO [train.py:904] (3/8) Epoch 1, batch 9300, loss[loss=0.2542, simple_loss=0.3175, pruned_loss=0.09541, over 12527.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3376, pruned_loss=0.09521, over 3054147.03 frames. ], batch size: 250, lr: 3.44e-02, grad_scale: 8.0 2023-04-27 16:05:33,415 INFO [optim.py:368] (3/8) 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 16:06:27,246 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:06:44,326 INFO [train.py:904] (3/8) Epoch 1, batch 9350, loss[loss=0.2663, simple_loss=0.3392, pruned_loss=0.09669, over 16896.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3375, pruned_loss=0.0952, over 3048717.42 frames. ], batch size: 116, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:07:17,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9387, 4.5785, 4.8131, 4.9421, 4.2618, 4.7477, 4.7489, 4.5011], device='cuda:3'), covar=tensor([0.0188, 0.0179, 0.0155, 0.0086, 0.0644, 0.0135, 0.0103, 0.0162], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0064, 0.0114, 0.0089, 0.0138, 0.0088, 0.0080, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:07:24,446 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:07:27,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0963, 4.1897, 4.4456, 4.5244, 4.6841, 4.1197, 4.3097, 4.3941], device='cuda:3'), covar=tensor([0.0250, 0.0274, 0.0436, 0.0425, 0.0285, 0.0268, 0.0524, 0.0206], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0108, 0.0121, 0.0123, 0.0134, 0.0111, 0.0153, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:07:54,038 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4567, 1.6723, 1.8776, 1.5296, 2.3918, 2.2350, 2.7547, 2.6742], device='cuda:3'), covar=tensor([0.0034, 0.0322, 0.0153, 0.0271, 0.0094, 0.0166, 0.0043, 0.0086], device='cuda:3'), in_proj_covar=tensor([0.0032, 0.0074, 0.0058, 0.0065, 0.0050, 0.0063, 0.0033, 0.0043], device='cuda:3'), out_proj_covar=tensor([4.4796e-05, 1.1409e-04, 8.7301e-05, 1.0005e-04, 7.7490e-05, 9.4092e-05, 5.2096e-05, 7.0664e-05], device='cuda:3') 2023-04-27 16:08:26,592 INFO [train.py:904] (3/8) Epoch 1, batch 9400, loss[loss=0.2871, simple_loss=0.3687, pruned_loss=0.1028, over 16692.00 frames. ], tot_loss[loss=0.2623, simple_loss=0.3368, pruned_loss=0.09386, over 3055702.14 frames. ], batch size: 134, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:08:54,863 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:56,911 INFO [optim.py:368] (3/8) 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 16:08:57,987 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:09:58,309 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8781, 3.8384, 3.5485, 1.8459, 2.9100, 2.2112, 3.2593, 3.8509], device='cuda:3'), covar=tensor([0.0267, 0.0388, 0.0304, 0.1915, 0.0886, 0.1167, 0.0825, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0083, 0.0125, 0.0155, 0.0148, 0.0138, 0.0138, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 16:10:08,169 INFO [train.py:904] (3/8) Epoch 1, batch 9450, loss[loss=0.2567, simple_loss=0.3337, pruned_loss=0.08979, over 15146.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3388, pruned_loss=0.09449, over 3051235.08 frames. ], batch size: 190, lr: 3.42e-02, grad_scale: 8.0 2023-04-27 16:10:33,965 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:58,530 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:32,514 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6626, 4.8857, 4.5749, 4.7415, 4.4119, 4.3053, 4.4837, 4.9074], device='cuda:3'), covar=tensor([0.0326, 0.0611, 0.0658, 0.0290, 0.0464, 0.0421, 0.0400, 0.0453], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0215, 0.0187, 0.0131, 0.0157, 0.0127, 0.0178, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 16:11:37,529 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-04-27 16:11:50,036 INFO [train.py:904] (3/8) Epoch 1, batch 9500, loss[loss=0.2519, simple_loss=0.3347, pruned_loss=0.08458, over 16347.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3379, pruned_loss=0.09387, over 3056847.76 frames. ], batch size: 146, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:12:21,414 INFO [optim.py:368] (3/8) 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 16:12:24,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4242, 3.2687, 3.1219, 3.0053, 3.3324, 2.0855, 3.0657, 3.0680], device='cuda:3'), covar=tensor([0.0084, 0.0074, 0.0104, 0.0222, 0.0067, 0.1142, 0.0091, 0.0122], device='cuda:3'), in_proj_covar=tensor([0.0046, 0.0040, 0.0054, 0.0066, 0.0041, 0.0095, 0.0053, 0.0053], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-27 16:13:32,517 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-27 16:13:37,009 INFO [train.py:904] (3/8) Epoch 1, batch 9550, loss[loss=0.2101, simple_loss=0.291, pruned_loss=0.06458, over 16287.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3368, pruned_loss=0.09323, over 3063573.75 frames. ], batch size: 35, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:14:46,666 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:14:48,703 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:15:18,626 INFO [train.py:904] (3/8) Epoch 1, batch 9600, loss[loss=0.2433, simple_loss=0.331, pruned_loss=0.07778, over 16754.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3391, pruned_loss=0.09492, over 3059120.77 frames. ], batch size: 83, lr: 3.40e-02, grad_scale: 8.0 2023-04-27 16:15:48,621 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 4.688e+02 5.707e+02 6.655e+02 1.542e+03, threshold=1.141e+03, percent-clipped=4.0 2023-04-27 16:16:20,294 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:45,946 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:53,272 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 9650, loss[loss=0.2437, simple_loss=0.3286, pruned_loss=0.07939, over 16683.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.342, pruned_loss=0.09578, over 3077805.02 frames. ], batch size: 134, lr: 3.39e-02, grad_scale: 8.0 2023-04-27 16:17:26,833 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:52,716 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:11,320 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-27 16:18:35,170 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 9700, loss[loss=0.2845, simple_loss=0.3572, pruned_loss=0.1058, over 16256.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3402, pruned_loss=0.09497, over 3075707.93 frames. ], batch size: 146, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:19:24,736 INFO [optim.py:368] (3/8) 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 16:19:28,113 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:33,558 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:43,503 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 16:20:27,216 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 1, batch 9750, loss[loss=0.2826, simple_loss=0.3616, pruned_loss=0.1018, over 16895.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3394, pruned_loss=0.09595, over 3050386.30 frames. ], batch size: 116, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:21:17,350 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:19,291 INFO [train.py:904] (3/8) Epoch 1, batch 9800, loss[loss=0.266, simple_loss=0.3477, pruned_loss=0.09212, over 16866.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3393, pruned_loss=0.09476, over 3042257.76 frames. ], batch size: 116, lr: 3.37e-02, grad_scale: 8.0 2023-04-27 16:22:26,148 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:47,905 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 1, batch 9850, loss[loss=0.2881, simple_loss=0.346, pruned_loss=0.1151, over 12456.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3403, pruned_loss=0.09439, over 3031708.56 frames. ], batch size: 248, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:25:45,203 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=2.07 vs. limit=2.0 2023-04-27 16:25:58,729 INFO [train.py:904] (3/8) Epoch 1, batch 9900, loss[loss=0.2574, simple_loss=0.3412, pruned_loss=0.08678, over 15348.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.34, pruned_loss=0.09356, over 3036161.07 frames. ], batch size: 190, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:26:31,897 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.967e+02 4.199e+02 5.185e+02 6.475e+02 1.120e+03, threshold=1.037e+03, percent-clipped=1.0 2023-04-27 16:27:32,796 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:27:57,274 INFO [train.py:904] (3/8) Epoch 1, batch 9950, loss[loss=0.28, simple_loss=0.3418, pruned_loss=0.109, over 12383.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3414, pruned_loss=0.0934, over 3043115.47 frames. ], batch size: 248, lr: 3.35e-02, grad_scale: 8.0 2023-04-27 16:28:53,734 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0633, 4.5833, 3.9358, 4.1491, 3.3457, 2.4897, 4.8196, 5.3255], device='cuda:3'), covar=tensor([0.1784, 0.0457, 0.0829, 0.0379, 0.1304, 0.1361, 0.0216, 0.0027], device='cuda:3'), in_proj_covar=tensor([0.0214, 0.0190, 0.0210, 0.0131, 0.0170, 0.0163, 0.0148, 0.0079], device='cuda:3'), out_proj_covar=tensor([2.4973e-04, 2.2057e-04, 2.2949e-04, 1.5537e-04, 2.0255e-04, 1.9664e-04, 1.7560e-04, 9.7724e-05], device='cuda:3') 2023-04-27 16:30:02,393 INFO [train.py:904] (3/8) Epoch 1, batch 10000, loss[loss=0.2658, simple_loss=0.3336, pruned_loss=0.09898, over 12973.00 frames. ], tot_loss[loss=0.2613, simple_loss=0.3391, pruned_loss=0.09175, over 3069391.08 frames. ], batch size: 248, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:30:30,090 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:30:30,649 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-27 16:30:32,719 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 4.345e+02 5.531e+02 6.703e+02 1.367e+03, threshold=1.106e+03, percent-clipped=3.0 2023-04-27 16:31:22,489 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:31:43,684 INFO [train.py:904] (3/8) Epoch 1, batch 10050, loss[loss=0.2986, simple_loss=0.3692, pruned_loss=0.114, over 16678.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.339, pruned_loss=0.09159, over 3069225.58 frames. ], batch size: 134, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:32:06,790 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:32:23,048 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:16,682 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:18,928 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:33:19,581 INFO [train.py:904] (3/8) Epoch 1, batch 10100, loss[loss=0.254, simple_loss=0.3241, pruned_loss=0.09189, over 16868.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3407, pruned_loss=0.09339, over 3069466.65 frames. ], batch size: 116, lr: 3.33e-02, grad_scale: 16.0 2023-04-27 16:33:49,432 INFO [optim.py:368] (3/8) 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 16:33:53,662 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:34:06,024 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:35:05,145 INFO [train.py:904] (3/8) Epoch 2, batch 0, loss[loss=0.2889, simple_loss=0.3479, pruned_loss=0.115, over 17214.00 frames. ], tot_loss[loss=0.2889, simple_loss=0.3479, pruned_loss=0.115, over 17214.00 frames. ], batch size: 44, lr: 3.26e-02, grad_scale: 8.0 2023-04-27 16:35:05,145 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 16:35:12,741 INFO [train.py:938] (3/8) Epoch 2, validation: loss=0.2235, simple_loss=0.3221, pruned_loss=0.06242, over 944034.00 frames. 2023-04-27 16:35:12,742 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 16:35:36,801 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 16:35:46,409 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5171, 1.8116, 1.9511, 1.5358, 2.2805, 2.2534, 2.2209, 2.5981], device='cuda:3'), covar=tensor([0.0030, 0.0203, 0.0114, 0.0180, 0.0076, 0.0110, 0.0046, 0.0054], device='cuda:3'), in_proj_covar=tensor([0.0036, 0.0077, 0.0065, 0.0071, 0.0062, 0.0069, 0.0039, 0.0046], device='cuda:3'), out_proj_covar=tensor([5.0514e-05, 1.1865e-04, 9.9956e-05, 1.1038e-04, 9.5770e-05, 1.0476e-04, 5.9811e-05, 7.5628e-05], device='cuda:3') 2023-04-27 16:36:09,210 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-27 16:36:19,093 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-27 16:36:22,855 INFO [train.py:904] (3/8) Epoch 2, batch 50, loss[loss=0.2639, simple_loss=0.3354, pruned_loss=0.09622, over 17136.00 frames. ], tot_loss[loss=0.3251, simple_loss=0.3724, pruned_loss=0.1388, over 749939.07 frames. ], batch size: 48, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:36:45,678 INFO [optim.py:368] (3/8) 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 16:36:56,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3942, 3.9418, 3.1297, 4.8102, 2.9577, 4.4060, 3.3213, 2.6523], device='cuda:3'), covar=tensor([0.0202, 0.0207, 0.0241, 0.0150, 0.1099, 0.0157, 0.0400, 0.1157], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0132, 0.0106, 0.0153, 0.0210, 0.0126, 0.0148, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:37:13,495 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.23 vs. limit=5.0 2023-04-27 16:37:16,881 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:37:26,984 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-27 16:37:30,897 INFO [train.py:904] (3/8) Epoch 2, batch 100, loss[loss=0.3122, simple_loss=0.3629, pruned_loss=0.1308, over 16494.00 frames. ], tot_loss[loss=0.3101, simple_loss=0.3604, pruned_loss=0.1299, over 1317355.09 frames. ], batch size: 146, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:37:33,211 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=6.94 vs. limit=5.0 2023-04-27 16:38:07,601 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8117, 4.8145, 5.2027, 5.2654, 5.5048, 4.8657, 5.0372, 5.0541], device='cuda:3'), covar=tensor([0.0288, 0.0222, 0.0394, 0.0461, 0.0286, 0.0240, 0.0607, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0124, 0.0140, 0.0146, 0.0154, 0.0131, 0.0194, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 16:38:22,579 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:38:38,636 INFO [train.py:904] (3/8) Epoch 2, batch 150, loss[loss=0.2811, simple_loss=0.3541, pruned_loss=0.1041, over 16730.00 frames. ], tot_loss[loss=0.306, simple_loss=0.3586, pruned_loss=0.1267, over 1756723.54 frames. ], batch size: 57, lr: 3.24e-02, grad_scale: 4.0 2023-04-27 16:38:39,321 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.38 vs. limit=5.0 2023-04-27 16:38:56,079 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:01,491 INFO [optim.py:368] (3/8) 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 16:39:26,839 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9073, 1.6060, 1.5284, 1.3430, 1.7244, 1.6284, 1.6242, 1.8965], device='cuda:3'), covar=tensor([0.0037, 0.0151, 0.0123, 0.0146, 0.0077, 0.0119, 0.0053, 0.0056], device='cuda:3'), in_proj_covar=tensor([0.0041, 0.0080, 0.0068, 0.0074, 0.0065, 0.0073, 0.0040, 0.0048], device='cuda:3'), out_proj_covar=tensor([5.9338e-05, 1.2409e-04, 1.0367e-04, 1.1473e-04, 1.0212e-04, 1.1338e-04, 6.3053e-05, 7.9240e-05], device='cuda:3') 2023-04-27 16:39:47,448 INFO [train.py:904] (3/8) Epoch 2, batch 200, loss[loss=0.2746, simple_loss=0.3451, pruned_loss=0.102, over 17116.00 frames. ], tot_loss[loss=0.3028, simple_loss=0.3565, pruned_loss=0.1245, over 2108848.22 frames. ], batch size: 48, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:40:00,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5613, 4.5069, 5.0264, 5.0649, 5.2480, 4.5594, 4.7945, 4.9515], device='cuda:3'), covar=tensor([0.0331, 0.0248, 0.0382, 0.0342, 0.0330, 0.0268, 0.0721, 0.0243], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0128, 0.0147, 0.0152, 0.0163, 0.0138, 0.0206, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 16:40:03,411 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:35,737 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.36 vs. limit=5.0 2023-04-27 16:40:43,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2502, 1.8774, 2.2314, 2.8351, 3.3171, 2.9897, 1.9617, 3.1696], device='cuda:3'), covar=tensor([0.0052, 0.0472, 0.0279, 0.0194, 0.0054, 0.0196, 0.0288, 0.0086], device='cuda:3'), in_proj_covar=tensor([0.0056, 0.0095, 0.0078, 0.0063, 0.0045, 0.0050, 0.0076, 0.0045], device='cuda:3'), out_proj_covar=tensor([1.0213e-04, 1.7129e-04, 1.4828e-04, 1.1832e-04, 8.1212e-05, 9.6884e-05, 1.3293e-04, 8.1888e-05], device='cuda:3') 2023-04-27 16:40:50,469 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:40:55,056 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:58,056 INFO [train.py:904] (3/8) Epoch 2, batch 250, loss[loss=0.2771, simple_loss=0.3432, pruned_loss=0.1055, over 17020.00 frames. ], tot_loss[loss=0.2981, simple_loss=0.353, pruned_loss=0.1216, over 2387365.38 frames. ], batch size: 50, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:41:03,951 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 16:41:11,032 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:16,378 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5846, 4.7334, 4.1788, 1.5448, 3.1726, 2.2891, 3.8547, 4.5183], device='cuda:3'), covar=tensor([0.0230, 0.0274, 0.0322, 0.2152, 0.0757, 0.1245, 0.0718, 0.0412], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0087, 0.0127, 0.0154, 0.0144, 0.0138, 0.0141, 0.0085], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 16:41:22,650 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.115e+02 4.151e+02 5.077e+02 6.310e+02 1.196e+03, threshold=1.015e+03, percent-clipped=1.0 2023-04-27 16:41:22,933 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:42:03,853 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 300, loss[loss=0.2687, simple_loss=0.3457, pruned_loss=0.09583, over 17052.00 frames. ], tot_loss[loss=0.2934, simple_loss=0.3496, pruned_loss=0.1186, over 2594992.84 frames. ], batch size: 50, lr: 3.22e-02, grad_scale: 4.0 2023-04-27 16:42:14,874 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:23,533 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6764, 2.7500, 1.5431, 2.7649, 2.0606, 2.7722, 1.8748, 2.3873], device='cuda:3'), covar=tensor([0.0096, 0.0157, 0.1607, 0.0080, 0.0742, 0.0322, 0.1170, 0.0586], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0082, 0.0164, 0.0072, 0.0140, 0.0100, 0.0167, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 16:42:36,335 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:43:04,891 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2030, 3.4563, 3.4142, 1.4809, 3.5554, 3.6040, 3.1131, 2.9338], device='cuda:3'), covar=tensor([0.0639, 0.0088, 0.0180, 0.1745, 0.0104, 0.0064, 0.0250, 0.0269], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0074, 0.0074, 0.0151, 0.0067, 0.0064, 0.0087, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:43:11,730 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 16:43:16,311 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-27 16:43:19,861 INFO [train.py:904] (3/8) Epoch 2, batch 350, loss[loss=0.2654, simple_loss=0.3165, pruned_loss=0.1071, over 16804.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3469, pruned_loss=0.1167, over 2750888.46 frames. ], batch size: 96, lr: 3.21e-02, grad_scale: 4.0 2023-04-27 16:43:39,910 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:43:42,893 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.580e+02 4.016e+02 4.975e+02 5.908e+02 1.216e+03, threshold=9.949e+02, percent-clipped=2.0 2023-04-27 16:44:27,071 INFO [train.py:904] (3/8) Epoch 2, batch 400, loss[loss=0.2674, simple_loss=0.3373, pruned_loss=0.09879, over 16631.00 frames. ], tot_loss[loss=0.2859, simple_loss=0.3438, pruned_loss=0.114, over 2875244.23 frames. ], batch size: 62, lr: 3.21e-02, grad_scale: 8.0 2023-04-27 16:45:33,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9831, 3.0438, 2.8214, 4.2858, 2.5000, 3.9865, 2.6926, 2.5990], device='cuda:3'), covar=tensor([0.0274, 0.0380, 0.0324, 0.0214, 0.1337, 0.0178, 0.0615, 0.1253], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0142, 0.0114, 0.0161, 0.0220, 0.0132, 0.0154, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:45:36,486 INFO [train.py:904] (3/8) Epoch 2, batch 450, loss[loss=0.2579, simple_loss=0.3341, pruned_loss=0.09085, over 17084.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3397, pruned_loss=0.1105, over 2985498.25 frames. ], batch size: 53, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:45:40,283 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6704, 2.7242, 1.6317, 2.7062, 2.0300, 2.7617, 1.7793, 2.3525], device='cuda:3'), covar=tensor([0.0079, 0.0159, 0.1194, 0.0100, 0.0647, 0.0309, 0.1160, 0.0481], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0083, 0.0160, 0.0073, 0.0140, 0.0104, 0.0166, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 16:45:59,585 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.527e+02 4.134e+02 5.028e+02 5.858e+02 1.204e+03, threshold=1.006e+03, percent-clipped=2.0 2023-04-27 16:46:44,100 INFO [train.py:904] (3/8) Epoch 2, batch 500, loss[loss=0.2616, simple_loss=0.3281, pruned_loss=0.09749, over 17124.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3376, pruned_loss=0.1092, over 3057303.01 frames. ], batch size: 49, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:46:53,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8909, 3.9069, 1.8261, 3.7858, 2.6804, 3.9033, 1.9069, 2.9776], device='cuda:3'), covar=tensor([0.0054, 0.0122, 0.1537, 0.0070, 0.0663, 0.0210, 0.1403, 0.0546], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0083, 0.0160, 0.0074, 0.0140, 0.0104, 0.0163, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0003, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 16:47:45,010 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:52,718 INFO [train.py:904] (3/8) Epoch 2, batch 550, loss[loss=0.2407, simple_loss=0.3218, pruned_loss=0.0798, over 17114.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3356, pruned_loss=0.1077, over 3117022.10 frames. ], batch size: 49, lr: 3.19e-02, grad_scale: 8.0 2023-04-27 16:48:17,046 INFO [optim.py:368] (3/8) 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 16:48:17,330 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:48:51,486 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:52,810 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 600, loss[loss=0.279, simple_loss=0.3271, pruned_loss=0.1154, over 16311.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3357, pruned_loss=0.1088, over 3155402.40 frames. ], batch size: 165, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:49:22,654 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:24,352 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:49:33,727 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:45,577 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 16:50:11,093 INFO [train.py:904] (3/8) Epoch 2, batch 650, loss[loss=0.288, simple_loss=0.3354, pruned_loss=0.1203, over 16436.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3344, pruned_loss=0.1086, over 3188632.39 frames. ], batch size: 146, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:50:16,453 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:23,982 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:50:33,024 INFO [optim.py:368] (3/8) 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 16:50:55,947 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7167, 4.3022, 4.4136, 1.5850, 3.1825, 2.5339, 3.8468, 4.6853], device='cuda:3'), covar=tensor([0.0246, 0.0422, 0.0289, 0.2127, 0.0797, 0.1101, 0.0828, 0.0320], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0092, 0.0135, 0.0158, 0.0149, 0.0139, 0.0144, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 16:50:57,100 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:59,405 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2918, 1.7363, 2.4964, 2.8843, 3.1996, 3.1843, 2.0502, 3.2223], device='cuda:3'), covar=tensor([0.0061, 0.0456, 0.0202, 0.0170, 0.0050, 0.0138, 0.0277, 0.0089], device='cuda:3'), in_proj_covar=tensor([0.0064, 0.0103, 0.0085, 0.0073, 0.0049, 0.0054, 0.0086, 0.0048], device='cuda:3'), out_proj_covar=tensor([1.1758e-04, 1.8510e-04, 1.6283e-04, 1.3844e-04, 8.8038e-05, 1.0364e-04, 1.4977e-04, 8.8526e-05], device='cuda:3') 2023-04-27 16:51:10,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0230, 3.8700, 2.4107, 4.5938, 4.4006, 4.3551, 2.0879, 3.5625], device='cuda:3'), covar=tensor([0.1616, 0.0277, 0.1551, 0.0057, 0.0154, 0.0223, 0.1102, 0.0404], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0104, 0.0160, 0.0061, 0.0080, 0.0090, 0.0143, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:51:19,283 INFO [train.py:904] (3/8) Epoch 2, batch 700, loss[loss=0.272, simple_loss=0.3481, pruned_loss=0.09796, over 17151.00 frames. ], tot_loss[loss=0.273, simple_loss=0.3327, pruned_loss=0.1066, over 3215040.83 frames. ], batch size: 49, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:51:54,552 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-27 16:52:06,364 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:52:26,661 INFO [train.py:904] (3/8) Epoch 2, batch 750, loss[loss=0.2641, simple_loss=0.3364, pruned_loss=0.09592, over 16758.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3341, pruned_loss=0.1071, over 3240953.73 frames. ], batch size: 57, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:50,206 INFO [optim.py:368] (3/8) 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 16:52:51,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8911, 4.0183, 2.2804, 5.0577, 5.0544, 4.7425, 2.6017, 3.9831], device='cuda:3'), covar=tensor([0.1785, 0.0353, 0.1608, 0.0125, 0.0097, 0.0262, 0.0999, 0.0401], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0107, 0.0165, 0.0063, 0.0083, 0.0092, 0.0145, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:53:28,000 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:53:33,379 INFO [train.py:904] (3/8) Epoch 2, batch 800, loss[loss=0.2731, simple_loss=0.3423, pruned_loss=0.102, over 17083.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3328, pruned_loss=0.1057, over 3259417.37 frames. ], batch size: 53, lr: 3.16e-02, grad_scale: 8.0 2023-04-27 16:53:35,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4369, 4.7043, 4.4100, 4.6263, 4.1370, 4.2402, 4.2023, 4.7245], device='cuda:3'), covar=tensor([0.0498, 0.0675, 0.0799, 0.0348, 0.0586, 0.0556, 0.0567, 0.0542], device='cuda:3'), in_proj_covar=tensor([0.0207, 0.0286, 0.0251, 0.0169, 0.0197, 0.0163, 0.0230, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 16:54:36,337 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 16:54:43,055 INFO [train.py:904] (3/8) Epoch 2, batch 850, loss[loss=0.2648, simple_loss=0.3236, pruned_loss=0.103, over 16879.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3318, pruned_loss=0.105, over 3277034.64 frames. ], batch size: 116, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:55:06,011 INFO [optim.py:368] (3/8) 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 16:55:08,658 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3448, 4.0609, 4.1114, 4.2865, 3.7136, 4.1520, 4.0177, 3.9939], device='cuda:3'), covar=tensor([0.0268, 0.0203, 0.0222, 0.0130, 0.0756, 0.0189, 0.0301, 0.0223], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0086, 0.0158, 0.0127, 0.0193, 0.0122, 0.0109, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 16:55:49,289 INFO [train.py:904] (3/8) Epoch 2, batch 900, loss[loss=0.2764, simple_loss=0.3505, pruned_loss=0.1011, over 17070.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3304, pruned_loss=0.1035, over 3293425.17 frames. ], batch size: 55, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:56:10,052 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:56,837 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 950, loss[loss=0.2285, simple_loss=0.2943, pruned_loss=0.08132, over 16995.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3302, pruned_loss=0.1036, over 3295919.81 frames. ], batch size: 41, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:57:10,415 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:57:13,956 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:57:38,003 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:06,961 INFO [train.py:904] (3/8) Epoch 2, batch 1000, loss[loss=0.2119, simple_loss=0.2853, pruned_loss=0.06926, over 16722.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3291, pruned_loss=0.1035, over 3294422.42 frames. ], batch size: 39, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:58:16,903 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:37,484 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:51,221 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8715, 4.6952, 4.8820, 5.2883, 5.2795, 4.8260, 5.2852, 5.1077], device='cuda:3'), covar=tensor([0.0386, 0.0409, 0.0956, 0.0312, 0.0326, 0.0356, 0.0250, 0.0272], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0240, 0.0362, 0.0255, 0.0204, 0.0206, 0.0186, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 16:59:13,979 INFO [train.py:904] (3/8) Epoch 2, batch 1050, loss[loss=0.2609, simple_loss=0.3182, pruned_loss=0.1018, over 16490.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.3292, pruned_loss=0.1041, over 3307191.87 frames. ], batch size: 146, lr: 3.13e-02, grad_scale: 8.0 2023-04-27 16:59:18,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4795, 3.5429, 3.8424, 3.9380, 3.9147, 3.4698, 3.6142, 3.7631], device='cuda:3'), covar=tensor([0.0313, 0.0376, 0.0417, 0.0382, 0.0462, 0.0345, 0.0663, 0.0321], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0141, 0.0164, 0.0161, 0.0189, 0.0152, 0.0236, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 16:59:36,187 INFO [optim.py:368] (3/8) 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 17:00:00,097 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:08,423 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 1100, loss[loss=0.2499, simple_loss=0.323, pruned_loss=0.08842, over 17191.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3276, pruned_loss=0.1027, over 3315416.35 frames. ], batch size: 46, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:00:38,841 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 17:01:28,364 INFO [train.py:904] (3/8) Epoch 2, batch 1150, loss[loss=0.2689, simple_loss=0.32, pruned_loss=0.1089, over 16771.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3265, pruned_loss=0.1011, over 3319292.77 frames. ], batch size: 124, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:01:52,663 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.647e+02 4.017e+02 4.974e+02 5.619e+02 1.017e+03, threshold=9.949e+02, percent-clipped=1.0 2023-04-27 17:02:06,021 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 17:02:39,328 INFO [train.py:904] (3/8) Epoch 2, batch 1200, loss[loss=0.2135, simple_loss=0.2877, pruned_loss=0.06966, over 16807.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3262, pruned_loss=0.1007, over 3321650.99 frames. ], batch size: 39, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:02:39,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7051, 2.1676, 2.0836, 2.2014, 2.6691, 2.8621, 3.0217, 2.8575], device='cuda:3'), covar=tensor([0.0083, 0.0155, 0.0112, 0.0145, 0.0070, 0.0081, 0.0036, 0.0070], device='cuda:3'), in_proj_covar=tensor([0.0040, 0.0082, 0.0072, 0.0080, 0.0071, 0.0079, 0.0044, 0.0054], device='cuda:3'), out_proj_covar=tensor([6.4304e-05, 1.2653e-04, 1.1020e-04, 1.2705e-04, 1.1209e-04, 1.2831e-04, 6.9885e-05, 9.1075e-05], device='cuda:3') 2023-04-27 17:03:46,864 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 1250, loss[loss=0.3199, simple_loss=0.3554, pruned_loss=0.1422, over 15523.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3271, pruned_loss=0.1024, over 3299342.65 frames. ], batch size: 190, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:04:10,333 INFO [optim.py:368] (3/8) 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 17:04:23,380 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-27 17:04:26,840 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:28,393 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-27 17:04:49,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:53,732 INFO [train.py:904] (3/8) Epoch 2, batch 1300, loss[loss=0.2569, simple_loss=0.3314, pruned_loss=0.09116, over 16721.00 frames. ], tot_loss[loss=0.266, simple_loss=0.3273, pruned_loss=0.1023, over 3299946.96 frames. ], batch size: 62, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:05:30,716 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 1350, loss[loss=0.2685, simple_loss=0.3286, pruned_loss=0.1042, over 16570.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3283, pruned_loss=0.1017, over 3305800.54 frames. ], batch size: 68, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:06:24,753 INFO [optim.py:368] (3/8) 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 17:06:41,096 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:45,040 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:57,576 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 1400, loss[loss=0.2464, simple_loss=0.3197, pruned_loss=0.08656, over 17126.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3278, pruned_loss=0.1018, over 3308273.58 frames. ], batch size: 47, lr: 3.09e-02, grad_scale: 8.0 2023-04-27 17:08:03,053 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:08:08,119 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:19,425 INFO [train.py:904] (3/8) Epoch 2, batch 1450, loss[loss=0.2512, simple_loss=0.319, pruned_loss=0.09172, over 17231.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.3265, pruned_loss=0.1013, over 3308862.62 frames. ], batch size: 43, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:08:43,788 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.475e+02 3.985e+02 4.876e+02 5.970e+02 8.943e+02, threshold=9.752e+02, percent-clipped=0.0 2023-04-27 17:08:56,528 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3161, 5.7299, 5.3582, 5.6163, 4.8830, 4.9708, 5.2018, 5.8225], device='cuda:3'), covar=tensor([0.0425, 0.0642, 0.0743, 0.0353, 0.0616, 0.0389, 0.0429, 0.0586], device='cuda:3'), in_proj_covar=tensor([0.0210, 0.0292, 0.0254, 0.0173, 0.0199, 0.0166, 0.0232, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 17:09:06,440 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:26,110 INFO [train.py:904] (3/8) Epoch 2, batch 1500, loss[loss=0.2727, simple_loss=0.345, pruned_loss=0.1002, over 16694.00 frames. ], tot_loss[loss=0.264, simple_loss=0.326, pruned_loss=0.101, over 3299968.69 frames. ], batch size: 57, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:09:52,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7093, 6.0063, 5.5954, 5.9562, 5.1672, 4.9812, 5.4528, 6.0792], device='cuda:3'), covar=tensor([0.0307, 0.0516, 0.0693, 0.0258, 0.0569, 0.0359, 0.0420, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0212, 0.0297, 0.0260, 0.0176, 0.0200, 0.0173, 0.0237, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 17:10:30,004 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:33,631 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:36,145 INFO [train.py:904] (3/8) Epoch 2, batch 1550, loss[loss=0.325, simple_loss=0.3644, pruned_loss=0.1428, over 16461.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3272, pruned_loss=0.1021, over 3300929.76 frames. ], batch size: 146, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:10:58,964 INFO [optim.py:368] (3/8) 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 17:11:15,285 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0866, 1.6026, 1.8118, 2.1916, 2.2849, 2.1240, 1.5274, 2.1749], device='cuda:3'), covar=tensor([0.0083, 0.0318, 0.0176, 0.0129, 0.0040, 0.0116, 0.0216, 0.0053], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0104, 0.0091, 0.0074, 0.0051, 0.0054, 0.0088, 0.0052], device='cuda:3'), out_proj_covar=tensor([1.2266e-04, 1.8815e-04, 1.7624e-04, 1.4149e-04, 9.1489e-05, 1.0382e-04, 1.5425e-04, 9.8995e-05], device='cuda:3') 2023-04-27 17:11:44,738 INFO [train.py:904] (3/8) Epoch 2, batch 1600, loss[loss=0.31, simple_loss=0.3596, pruned_loss=0.1302, over 16164.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3308, pruned_loss=0.1041, over 3304406.13 frames. ], batch size: 164, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:11:57,764 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:53,468 INFO [train.py:904] (3/8) Epoch 2, batch 1650, loss[loss=0.3178, simple_loss=0.3626, pruned_loss=0.1365, over 16673.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3322, pruned_loss=0.105, over 3316645.89 frames. ], batch size: 134, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:13:04,427 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7041, 3.6561, 4.0665, 4.0634, 4.1428, 3.7432, 3.7759, 3.9480], device='cuda:3'), covar=tensor([0.0285, 0.0332, 0.0384, 0.0444, 0.0323, 0.0317, 0.0831, 0.0333], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0146, 0.0167, 0.0170, 0.0194, 0.0161, 0.0245, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 17:13:15,779 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:17,808 INFO [optim.py:368] (3/8) 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 17:13:35,468 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:56,968 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:03,719 INFO [train.py:904] (3/8) Epoch 2, batch 1700, loss[loss=0.2977, simple_loss=0.3452, pruned_loss=0.125, over 16701.00 frames. ], tot_loss[loss=0.2747, simple_loss=0.3352, pruned_loss=0.1071, over 3313368.00 frames. ], batch size: 124, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:14:06,069 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6890, 1.9533, 1.7553, 1.7323, 2.5322, 2.4445, 2.5107, 2.5859], device='cuda:3'), covar=tensor([0.0061, 0.0164, 0.0117, 0.0199, 0.0068, 0.0119, 0.0062, 0.0064], device='cuda:3'), in_proj_covar=tensor([0.0041, 0.0083, 0.0074, 0.0084, 0.0073, 0.0082, 0.0047, 0.0056], device='cuda:3'), out_proj_covar=tensor([6.6735e-05, 1.2782e-04, 1.1504e-04, 1.3333e-04, 1.1672e-04, 1.3342e-04, 7.4763e-05, 9.6065e-05], device='cuda:3') 2023-04-27 17:14:38,519 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-27 17:14:40,648 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:41,657 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:55,344 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:11,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7635, 4.6417, 4.8863, 1.9761, 3.6611, 2.4849, 4.1134, 4.7241], device='cuda:3'), covar=tensor([0.0263, 0.0426, 0.0240, 0.1916, 0.0617, 0.1173, 0.0814, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0098, 0.0139, 0.0155, 0.0142, 0.0139, 0.0146, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 17:15:13,200 INFO [train.py:904] (3/8) Epoch 2, batch 1750, loss[loss=0.2748, simple_loss=0.3489, pruned_loss=0.1003, over 17047.00 frames. ], tot_loss[loss=0.274, simple_loss=0.3356, pruned_loss=0.1062, over 3314609.12 frames. ], batch size: 53, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:15:21,950 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:15:39,688 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:15,437 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9399, 3.8177, 3.7449, 3.8643, 3.3935, 3.8397, 3.5821, 3.5642], device='cuda:3'), covar=tensor([0.0271, 0.0142, 0.0212, 0.0161, 0.0758, 0.0199, 0.0582, 0.0253], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0087, 0.0164, 0.0133, 0.0202, 0.0132, 0.0116, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:16:17,753 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:21,485 INFO [train.py:904] (3/8) Epoch 2, batch 1800, loss[loss=0.2722, simple_loss=0.3464, pruned_loss=0.09904, over 17031.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3379, pruned_loss=0.1072, over 3315836.92 frames. ], batch size: 50, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:17:03,338 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:07,246 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:08,272 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9177, 4.6851, 4.2800, 1.9091, 4.6938, 4.6213, 3.8147, 3.8757], device='cuda:3'), covar=tensor([0.0677, 0.0057, 0.0255, 0.1643, 0.0097, 0.0091, 0.0235, 0.0204], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0076, 0.0076, 0.0149, 0.0069, 0.0067, 0.0092, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 17:17:16,330 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 1850, loss[loss=0.255, simple_loss=0.3339, pruned_loss=0.08804, over 17186.00 frames. ], tot_loss[loss=0.2759, simple_loss=0.3379, pruned_loss=0.1069, over 3318438.91 frames. ], batch size: 46, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:17:45,597 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:57,888 INFO [optim.py:368] (3/8) 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 17:18:35,071 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:43,199 INFO [train.py:904] (3/8) Epoch 2, batch 1900, loss[loss=0.2911, simple_loss=0.3339, pruned_loss=0.1242, over 16785.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3362, pruned_loss=0.105, over 3321300.30 frames. ], batch size: 83, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:18:49,560 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:53,090 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6221, 5.0265, 4.8911, 4.9526, 4.9040, 5.5476, 5.3292, 4.9543], device='cuda:3'), covar=tensor([0.0777, 0.0951, 0.0860, 0.1323, 0.1882, 0.0581, 0.0720, 0.1750], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0266, 0.0236, 0.0231, 0.0295, 0.0233, 0.0209, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:19:17,238 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:43,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0310, 4.5815, 2.3973, 5.4087, 5.1813, 4.8237, 2.3635, 3.9582], device='cuda:3'), covar=tensor([0.1563, 0.0206, 0.1415, 0.0051, 0.0110, 0.0253, 0.0994, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0109, 0.0167, 0.0067, 0.0094, 0.0102, 0.0149, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 17:19:51,447 INFO [train.py:904] (3/8) Epoch 2, batch 1950, loss[loss=0.2779, simple_loss=0.3516, pruned_loss=0.1021, over 17114.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.3356, pruned_loss=0.1038, over 3323884.51 frames. ], batch size: 49, lr: 3.03e-02, grad_scale: 8.0 2023-04-27 17:20:14,619 INFO [optim.py:368] (3/8) 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 17:20:40,694 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:59,921 INFO [train.py:904] (3/8) Epoch 2, batch 2000, loss[loss=0.3301, simple_loss=0.3791, pruned_loss=0.1405, over 16298.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3354, pruned_loss=0.1037, over 3322558.16 frames. ], batch size: 165, lr: 3.02e-02, grad_scale: 8.0 2023-04-27 17:21:28,808 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:51,918 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 2050, loss[loss=0.2893, simple_loss=0.3415, pruned_loss=0.1185, over 16710.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.3357, pruned_loss=0.1038, over 3314544.15 frames. ], batch size: 89, lr: 3.02e-02, grad_scale: 16.0 2023-04-27 17:22:11,066 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:32,914 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:20,404 INFO [train.py:904] (3/8) Epoch 2, batch 2100, loss[loss=0.2866, simple_loss=0.3415, pruned_loss=0.1159, over 16722.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.3388, pruned_loss=0.1074, over 3303640.38 frames. ], batch size: 124, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:23:28,410 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1484, 4.5773, 4.8399, 4.9210, 4.2905, 4.8888, 4.9324, 4.4974], device='cuda:3'), covar=tensor([0.0246, 0.0205, 0.0200, 0.0134, 0.0865, 0.0201, 0.0171, 0.0269], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0089, 0.0163, 0.0132, 0.0201, 0.0132, 0.0115, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:23:53,808 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:56,016 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-27 17:24:11,223 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2023-04-27 17:24:15,547 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:28,733 INFO [train.py:904] (3/8) Epoch 2, batch 2150, loss[loss=0.3005, simple_loss=0.3527, pruned_loss=0.1241, over 16793.00 frames. ], tot_loss[loss=0.2775, simple_loss=0.3393, pruned_loss=0.1079, over 3301145.64 frames. ], batch size: 124, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:24:32,705 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-27 17:24:33,514 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:45,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1101, 4.7891, 4.5117, 1.8137, 4.8385, 4.7974, 3.6272, 3.9715], device='cuda:3'), covar=tensor([0.0532, 0.0037, 0.0121, 0.1533, 0.0044, 0.0033, 0.0259, 0.0181], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0075, 0.0076, 0.0147, 0.0069, 0.0066, 0.0093, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 17:24:51,836 INFO [optim.py:368] (3/8) 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 17:25:04,170 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0709, 3.2421, 2.7119, 4.4314, 2.2720, 4.1813, 2.7989, 2.6325], device='cuda:3'), covar=tensor([0.0227, 0.0325, 0.0291, 0.0152, 0.1384, 0.0146, 0.0530, 0.1259], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0153, 0.0128, 0.0177, 0.0232, 0.0143, 0.0160, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:25:22,225 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:23,339 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:38,415 INFO [train.py:904] (3/8) Epoch 2, batch 2200, loss[loss=0.2796, simple_loss=0.3324, pruned_loss=0.1134, over 16795.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3401, pruned_loss=0.1081, over 3294929.70 frames. ], batch size: 102, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:25:44,347 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:56,679 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7977, 3.8702, 2.3898, 4.7345, 4.7093, 4.3101, 2.6047, 3.1172], device='cuda:3'), covar=tensor([0.1860, 0.0343, 0.1487, 0.0097, 0.0157, 0.0361, 0.0965, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0109, 0.0164, 0.0068, 0.0096, 0.0101, 0.0147, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 17:26:12,203 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-27 17:26:24,956 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 17:26:28,674 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-04-27 17:26:48,777 INFO [train.py:904] (3/8) Epoch 2, batch 2250, loss[loss=0.3555, simple_loss=0.3977, pruned_loss=0.1567, over 11911.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3399, pruned_loss=0.1078, over 3307397.36 frames. ], batch size: 246, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:26:51,419 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:53,450 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4926, 3.3195, 1.6444, 3.2957, 2.2355, 3.4253, 1.9018, 2.5906], device='cuda:3'), covar=tensor([0.0063, 0.0260, 0.1677, 0.0064, 0.0900, 0.0465, 0.1295, 0.0682], device='cuda:3'), in_proj_covar=tensor([0.0081, 0.0098, 0.0167, 0.0076, 0.0152, 0.0119, 0.0166, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:26:55,186 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 17:27:12,081 INFO [optim.py:368] (3/8) 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 17:27:19,643 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-27 17:27:32,273 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:51,373 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:57,827 INFO [train.py:904] (3/8) Epoch 2, batch 2300, loss[loss=0.2699, simple_loss=0.3348, pruned_loss=0.1025, over 16709.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3391, pruned_loss=0.1071, over 3307091.05 frames. ], batch size: 124, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:28:01,724 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 17:28:12,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2298, 1.9310, 2.5098, 2.8425, 3.5574, 3.3638, 2.0450, 3.4436], device='cuda:3'), covar=tensor([0.0048, 0.0300, 0.0133, 0.0124, 0.0027, 0.0072, 0.0194, 0.0046], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0107, 0.0090, 0.0078, 0.0053, 0.0054, 0.0088, 0.0054], device='cuda:3'), out_proj_covar=tensor([1.2605e-04, 1.9312e-04, 1.7248e-04, 1.4965e-04, 9.6205e-05, 1.0301e-04, 1.5362e-04, 1.0152e-04], device='cuda:3') 2023-04-27 17:28:26,526 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 2350, loss[loss=0.2944, simple_loss=0.3676, pruned_loss=0.1106, over 16658.00 frames. ], tot_loss[loss=0.2766, simple_loss=0.3395, pruned_loss=0.1069, over 3314484.42 frames. ], batch size: 62, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:29:07,915 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:14,994 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:31,565 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:14,653 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 2400, loss[loss=0.2225, simple_loss=0.2999, pruned_loss=0.07254, over 17210.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.3405, pruned_loss=0.1077, over 3316579.76 frames. ], batch size: 45, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:30:48,897 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 17:30:52,400 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 2450, loss[loss=0.2968, simple_loss=0.363, pruned_loss=0.1153, over 16319.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3404, pruned_loss=0.107, over 3317481.97 frames. ], batch size: 165, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:31:31,779 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:51,021 INFO [optim.py:368] (3/8) 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 17:31:59,229 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:20,257 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:35,554 INFO [train.py:904] (3/8) Epoch 2, batch 2500, loss[loss=0.2712, simple_loss=0.3411, pruned_loss=0.1007, over 16529.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3408, pruned_loss=0.106, over 3318333.14 frames. ], batch size: 68, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:32:36,842 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:27,071 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:43,147 INFO [train.py:904] (3/8) Epoch 2, batch 2550, loss[loss=0.2242, simple_loss=0.2959, pruned_loss=0.07622, over 16974.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3397, pruned_loss=0.105, over 3329563.50 frames. ], batch size: 41, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:34:07,328 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 17:34:08,155 INFO [optim.py:368] (3/8) 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 17:34:27,653 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 2600, loss[loss=0.2574, simple_loss=0.3316, pruned_loss=0.09163, over 17135.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3379, pruned_loss=0.1036, over 3328109.54 frames. ], batch size: 48, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:34:55,755 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-27 17:35:08,880 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:16,533 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 17:35:32,886 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:36:01,522 INFO [train.py:904] (3/8) Epoch 2, batch 2650, loss[loss=0.2988, simple_loss=0.3493, pruned_loss=0.1241, over 15427.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3389, pruned_loss=0.1039, over 3325557.14 frames. ], batch size: 190, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:36:03,295 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:36:30,032 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 17:36:33,937 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 17:37:09,366 INFO [train.py:904] (3/8) Epoch 2, batch 2700, loss[loss=0.2907, simple_loss=0.3432, pruned_loss=0.1191, over 16820.00 frames. ], tot_loss[loss=0.2729, simple_loss=0.3389, pruned_loss=0.1034, over 3323367.31 frames. ], batch size: 102, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:37:55,210 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3606, 2.2664, 1.9375, 1.9998, 2.8833, 2.8083, 3.7513, 3.2144], device='cuda:3'), covar=tensor([0.0019, 0.0143, 0.0134, 0.0163, 0.0071, 0.0120, 0.0055, 0.0068], device='cuda:3'), in_proj_covar=tensor([0.0044, 0.0088, 0.0081, 0.0085, 0.0075, 0.0086, 0.0053, 0.0062], device='cuda:3'), out_proj_covar=tensor([7.4287e-05, 1.3769e-04, 1.2395e-04, 1.3583e-04, 1.2209e-04, 1.4061e-04, 8.4356e-05, 1.0528e-04], device='cuda:3') 2023-04-27 17:38:19,507 INFO [train.py:904] (3/8) Epoch 2, batch 2750, loss[loss=0.2877, simple_loss=0.3485, pruned_loss=0.1134, over 16925.00 frames. ], tot_loss[loss=0.2703, simple_loss=0.3375, pruned_loss=0.1015, over 3330275.95 frames. ], batch size: 109, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:41,654 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.401e+02 4.125e+02 4.810e+02 5.955e+02 1.093e+03, threshold=9.620e+02, percent-clipped=2.0 2023-04-27 17:39:26,263 INFO [train.py:904] (3/8) Epoch 2, batch 2800, loss[loss=0.2731, simple_loss=0.3274, pruned_loss=0.1094, over 16472.00 frames. ], tot_loss[loss=0.2712, simple_loss=0.3371, pruned_loss=0.1026, over 3327045.95 frames. ], batch size: 146, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:33,619 INFO [train.py:904] (3/8) Epoch 2, batch 2850, loss[loss=0.2545, simple_loss=0.3395, pruned_loss=0.08475, over 17259.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3358, pruned_loss=0.1017, over 3326387.11 frames. ], batch size: 52, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:39,281 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-27 17:40:57,335 INFO [optim.py:368] (3/8) 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 17:41:12,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7285, 2.9548, 2.5747, 3.9710, 2.2803, 3.7138, 2.3827, 2.5388], device='cuda:3'), covar=tensor([0.0261, 0.0287, 0.0277, 0.0169, 0.1103, 0.0158, 0.0566, 0.0931], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0159, 0.0136, 0.0189, 0.0238, 0.0150, 0.0166, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:41:41,131 INFO [train.py:904] (3/8) Epoch 2, batch 2900, loss[loss=0.3628, simple_loss=0.3779, pruned_loss=0.1738, over 15419.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3368, pruned_loss=0.1039, over 3316820.74 frames. ], batch size: 190, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:41:59,415 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-27 17:42:14,402 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.26 vs. limit=5.0 2023-04-27 17:42:49,006 INFO [train.py:904] (3/8) Epoch 2, batch 2950, loss[loss=0.2241, simple_loss=0.2915, pruned_loss=0.07836, over 16986.00 frames. ], tot_loss[loss=0.2723, simple_loss=0.3358, pruned_loss=0.1044, over 3313241.84 frames. ], batch size: 41, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:50,404 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:13,805 INFO [optim.py:368] (3/8) 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 17:43:14,144 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 17:43:42,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2562, 2.4039, 2.1538, 2.2139, 2.8354, 2.8006, 3.7162, 3.3244], device='cuda:3'), covar=tensor([0.0021, 0.0111, 0.0106, 0.0131, 0.0061, 0.0098, 0.0032, 0.0043], device='cuda:3'), in_proj_covar=tensor([0.0044, 0.0087, 0.0080, 0.0082, 0.0075, 0.0085, 0.0053, 0.0063], device='cuda:3'), out_proj_covar=tensor([7.4984e-05, 1.3634e-04, 1.2368e-04, 1.3069e-04, 1.2303e-04, 1.3963e-04, 8.6191e-05, 1.0703e-04], device='cuda:3') 2023-04-27 17:43:53,662 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:54,563 INFO [train.py:904] (3/8) Epoch 2, batch 3000, loss[loss=0.2431, simple_loss=0.3277, pruned_loss=0.07922, over 17002.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3353, pruned_loss=0.104, over 3322080.35 frames. ], batch size: 50, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:43:54,563 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 17:44:03,915 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 17:44:13,971 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0026, 4.4948, 2.5313, 5.2140, 5.1482, 4.8145, 2.4740, 3.8860], device='cuda:3'), covar=tensor([0.1647, 0.0245, 0.1383, 0.0080, 0.0158, 0.0353, 0.1040, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0112, 0.0167, 0.0070, 0.0106, 0.0110, 0.0150, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:45:09,406 INFO [train.py:904] (3/8) Epoch 2, batch 3050, loss[loss=0.2924, simple_loss=0.338, pruned_loss=0.1233, over 16786.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3343, pruned_loss=0.1034, over 3317695.35 frames. ], batch size: 102, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:45:33,139 INFO [optim.py:368] (3/8) 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 17:46:03,619 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9968, 4.6941, 4.8908, 5.3198, 5.4237, 4.8268, 5.4130, 5.2033], device='cuda:3'), covar=tensor([0.0482, 0.0512, 0.1099, 0.0289, 0.0279, 0.0343, 0.0190, 0.0248], device='cuda:3'), in_proj_covar=tensor([0.0238, 0.0284, 0.0407, 0.0292, 0.0228, 0.0222, 0.0210, 0.0230], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:46:15,164 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.82 vs. limit=5.0 2023-04-27 17:46:15,658 INFO [train.py:904] (3/8) Epoch 2, batch 3100, loss[loss=0.2836, simple_loss=0.3299, pruned_loss=0.1187, over 16785.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3343, pruned_loss=0.1037, over 3315601.15 frames. ], batch size: 102, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:46:26,903 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2887, 4.1627, 2.1145, 4.3091, 2.9731, 4.2850, 2.2149, 2.9736], device='cuda:3'), covar=tensor([0.0042, 0.0234, 0.1333, 0.0052, 0.0570, 0.0241, 0.1184, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0101, 0.0168, 0.0078, 0.0150, 0.0129, 0.0170, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:47:22,152 INFO [train.py:904] (3/8) Epoch 2, batch 3150, loss[loss=0.2687, simple_loss=0.3277, pruned_loss=0.1048, over 16635.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3341, pruned_loss=0.1035, over 3311655.67 frames. ], batch size: 75, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:39,755 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-04-27 17:47:44,588 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.682e+02 3.986e+02 4.862e+02 6.258e+02 1.253e+03, threshold=9.723e+02, percent-clipped=2.0 2023-04-27 17:48:28,282 INFO [train.py:904] (3/8) Epoch 2, batch 3200, loss[loss=0.2526, simple_loss=0.3335, pruned_loss=0.08584, over 17061.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3338, pruned_loss=0.1026, over 3306341.43 frames. ], batch size: 53, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:48:34,711 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8982, 3.1130, 2.6926, 4.2162, 2.4382, 3.8764, 2.5839, 2.5632], device='cuda:3'), covar=tensor([0.0214, 0.0306, 0.0280, 0.0156, 0.1184, 0.0159, 0.0555, 0.0953], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0162, 0.0138, 0.0192, 0.0245, 0.0153, 0.0171, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:48:50,328 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2817, 4.9793, 5.0511, 4.3131, 4.9330, 2.5302, 4.7792, 5.1735], device='cuda:3'), covar=tensor([0.0059, 0.0059, 0.0057, 0.0346, 0.0056, 0.1059, 0.0072, 0.0083], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0057, 0.0081, 0.0107, 0.0064, 0.0105, 0.0078, 0.0083], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 17:49:34,225 INFO [train.py:904] (3/8) Epoch 2, batch 3250, loss[loss=0.2742, simple_loss=0.3338, pruned_loss=0.1073, over 16846.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3319, pruned_loss=0.1013, over 3319276.34 frames. ], batch size: 96, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:55,441 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 17:49:58,154 INFO [optim.py:368] (3/8) 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 17:49:58,604 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:42,664 INFO [train.py:904] (3/8) Epoch 2, batch 3300, loss[loss=0.2759, simple_loss=0.3376, pruned_loss=0.1072, over 16778.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3341, pruned_loss=0.1029, over 3309643.67 frames. ], batch size: 83, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:51:03,265 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:48,084 INFO [train.py:904] (3/8) Epoch 2, batch 3350, loss[loss=0.3115, simple_loss=0.364, pruned_loss=0.1295, over 16436.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3338, pruned_loss=0.1011, over 3311009.93 frames. ], batch size: 146, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:52:13,285 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.954e+02 4.311e+02 5.188e+02 6.478e+02 1.182e+03, threshold=1.038e+03, percent-clipped=6.0 2023-04-27 17:52:26,045 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-27 17:52:56,206 INFO [train.py:904] (3/8) Epoch 2, batch 3400, loss[loss=0.3019, simple_loss=0.361, pruned_loss=0.1214, over 15657.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3333, pruned_loss=0.1005, over 3317538.26 frames. ], batch size: 191, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:53:04,672 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6754, 4.7105, 4.4353, 1.9567, 4.7054, 4.8448, 3.9491, 3.7549], device='cuda:3'), covar=tensor([0.0727, 0.0077, 0.0183, 0.1477, 0.0141, 0.0090, 0.0226, 0.0248], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0084, 0.0080, 0.0150, 0.0076, 0.0072, 0.0099, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 17:54:05,265 INFO [train.py:904] (3/8) Epoch 2, batch 3450, loss[loss=0.2905, simple_loss=0.3337, pruned_loss=0.1237, over 16867.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3318, pruned_loss=0.0997, over 3311820.22 frames. ], batch size: 96, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:29,793 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.037e+02 4.547e+02 5.342e+02 6.587e+02 1.202e+03, threshold=1.068e+03, percent-clipped=2.0 2023-04-27 17:55:11,512 INFO [train.py:904] (3/8) Epoch 2, batch 3500, loss[loss=0.2919, simple_loss=0.3484, pruned_loss=0.1177, over 15442.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.33, pruned_loss=0.09883, over 3318306.44 frames. ], batch size: 191, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:56:21,450 INFO [train.py:904] (3/8) Epoch 2, batch 3550, loss[loss=0.2886, simple_loss=0.3476, pruned_loss=0.1148, over 16841.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3291, pruned_loss=0.09822, over 3322196.11 frames. ], batch size: 102, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:56:45,024 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.316e+02 4.227e+02 5.059e+02 5.919e+02 1.365e+03, threshold=1.012e+03, percent-clipped=3.0 2023-04-27 17:57:11,461 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:29,798 INFO [train.py:904] (3/8) Epoch 2, batch 3600, loss[loss=0.2693, simple_loss=0.3404, pruned_loss=0.09911, over 16695.00 frames. ], tot_loss[loss=0.2618, simple_loss=0.3279, pruned_loss=0.09783, over 3307607.70 frames. ], batch size: 57, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:57:36,311 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2877, 4.0814, 3.6642, 1.8944, 2.7231, 2.2349, 3.5550, 4.0464], device='cuda:3'), covar=tensor([0.0238, 0.0357, 0.0341, 0.1541, 0.0732, 0.1014, 0.0719, 0.0369], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0112, 0.0146, 0.0154, 0.0147, 0.0140, 0.0151, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:57:41,676 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4300, 4.2524, 3.7412, 1.9182, 2.8274, 2.1587, 3.7416, 4.1727], device='cuda:3'), covar=tensor([0.0216, 0.0287, 0.0344, 0.1507, 0.0742, 0.1099, 0.0624, 0.0306], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0112, 0.0146, 0.0154, 0.0148, 0.0140, 0.0151, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:58:01,694 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2591, 1.9272, 2.4276, 2.8460, 3.1824, 2.8658, 1.7818, 3.0535], device='cuda:3'), covar=tensor([0.0037, 0.0255, 0.0148, 0.0114, 0.0037, 0.0107, 0.0235, 0.0038], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0110, 0.0094, 0.0084, 0.0058, 0.0059, 0.0093, 0.0055], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 17:58:15,114 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8554, 3.8325, 1.9196, 3.8393, 2.4445, 3.8258, 2.0445, 2.7236], device='cuda:3'), covar=tensor([0.0075, 0.0145, 0.1419, 0.0059, 0.0751, 0.0268, 0.1158, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0100, 0.0167, 0.0079, 0.0152, 0.0128, 0.0169, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:58:32,399 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 17:58:37,311 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 17:58:40,990 INFO [train.py:904] (3/8) Epoch 2, batch 3650, loss[loss=0.2439, simple_loss=0.2941, pruned_loss=0.09688, over 16772.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.327, pruned_loss=0.09976, over 3286566.26 frames. ], batch size: 124, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 17:59:08,122 INFO [optim.py:368] (3/8) 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 17:59:38,318 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8964, 3.9807, 2.8473, 4.7846, 4.6675, 4.4244, 2.0989, 3.4338], device='cuda:3'), covar=tensor([0.1555, 0.0291, 0.1128, 0.0057, 0.0178, 0.0202, 0.1134, 0.0539], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0164, 0.0069, 0.0106, 0.0108, 0.0149, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 17:59:54,878 INFO [train.py:904] (3/8) Epoch 2, batch 3700, loss[loss=0.2491, simple_loss=0.3034, pruned_loss=0.09738, over 16815.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3252, pruned_loss=0.1011, over 3275669.05 frames. ], batch size: 90, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 18:01:07,711 INFO [train.py:904] (3/8) Epoch 2, batch 3750, loss[loss=0.2865, simple_loss=0.3377, pruned_loss=0.1177, over 16506.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3255, pruned_loss=0.1029, over 3271584.75 frames. ], batch size: 146, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:01:33,975 INFO [optim.py:368] (3/8) 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 18:02:24,605 INFO [train.py:904] (3/8) Epoch 2, batch 3800, loss[loss=0.2991, simple_loss=0.3466, pruned_loss=0.1258, over 16818.00 frames. ], tot_loss[loss=0.2671, simple_loss=0.3267, pruned_loss=0.1037, over 3264828.09 frames. ], batch size: 96, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:03:12,496 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-27 18:03:40,205 INFO [train.py:904] (3/8) Epoch 2, batch 3850, loss[loss=0.2724, simple_loss=0.3263, pruned_loss=0.1093, over 16855.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3252, pruned_loss=0.1031, over 3268348.55 frames. ], batch size: 96, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:03:43,705 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8323, 3.2385, 3.1339, 1.3457, 3.3060, 3.2951, 2.8982, 2.6463], device='cuda:3'), covar=tensor([0.0849, 0.0114, 0.0196, 0.1502, 0.0120, 0.0106, 0.0286, 0.0355], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0079, 0.0077, 0.0148, 0.0073, 0.0070, 0.0100, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:04:07,046 INFO [optim.py:368] (3/8) 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 18:04:53,515 INFO [train.py:904] (3/8) Epoch 2, batch 3900, loss[loss=0.3356, simple_loss=0.3775, pruned_loss=0.1469, over 12487.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3247, pruned_loss=0.1033, over 3263404.24 frames. ], batch size: 248, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:05:04,305 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7910, 4.5686, 4.5982, 4.6845, 4.1518, 4.5731, 4.5390, 4.3288], device='cuda:3'), covar=tensor([0.0251, 0.0134, 0.0156, 0.0117, 0.0645, 0.0179, 0.0183, 0.0202], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0090, 0.0166, 0.0127, 0.0194, 0.0136, 0.0112, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 18:05:25,204 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:05:55,290 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:06:05,397 INFO [train.py:904] (3/8) Epoch 2, batch 3950, loss[loss=0.2871, simple_loss=0.3427, pruned_loss=0.1157, over 15613.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3241, pruned_loss=0.1035, over 3267446.79 frames. ], batch size: 191, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:06:09,444 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:06:26,040 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0334, 3.1414, 3.2032, 1.5111, 3.3182, 3.3172, 2.9732, 2.7546], device='cuda:3'), covar=tensor([0.0836, 0.0132, 0.0175, 0.1662, 0.0092, 0.0073, 0.0286, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0080, 0.0076, 0.0150, 0.0074, 0.0071, 0.0101, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:06:31,645 INFO [optim.py:368] (3/8) 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 18:06:34,609 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:06:52,487 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:07:10,392 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0821, 4.0767, 2.7746, 5.1607, 5.1544, 4.6198, 2.2987, 3.7990], device='cuda:3'), covar=tensor([0.1524, 0.0306, 0.1348, 0.0040, 0.0067, 0.0217, 0.1231, 0.0510], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0116, 0.0168, 0.0068, 0.0108, 0.0110, 0.0154, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 18:07:12,668 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2061, 3.9926, 3.5470, 1.7373, 2.7079, 2.1530, 3.6059, 4.0776], device='cuda:3'), covar=tensor([0.0244, 0.0364, 0.0383, 0.1744, 0.0890, 0.1150, 0.0628, 0.0277], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0107, 0.0147, 0.0157, 0.0147, 0.0138, 0.0152, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 18:07:14,082 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 18:07:17,633 INFO [train.py:904] (3/8) Epoch 2, batch 4000, loss[loss=0.2185, simple_loss=0.2849, pruned_loss=0.07602, over 16712.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3231, pruned_loss=0.1031, over 3278182.82 frames. ], batch size: 57, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:07:37,748 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:08:02,204 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 18:08:09,008 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6165, 5.3937, 5.2567, 5.1362, 5.3484, 5.6720, 5.4283, 5.2074], device='cuda:3'), covar=tensor([0.0599, 0.0685, 0.0792, 0.1299, 0.1383, 0.0647, 0.0569, 0.1350], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0271, 0.0247, 0.0241, 0.0298, 0.0259, 0.0214, 0.0312], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 18:08:31,014 INFO [train.py:904] (3/8) Epoch 2, batch 4050, loss[loss=0.2233, simple_loss=0.3044, pruned_loss=0.07112, over 16841.00 frames. ], tot_loss[loss=0.2604, simple_loss=0.3213, pruned_loss=0.0998, over 3277666.75 frames. ], batch size: 102, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:08:54,732 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 18:08:56,314 INFO [optim.py:368] (3/8) 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 18:09:17,269 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6610, 4.1174, 3.4589, 2.9689, 2.9519, 2.2746, 4.3760, 5.1285], device='cuda:3'), covar=tensor([0.1955, 0.0480, 0.0842, 0.0645, 0.1838, 0.1238, 0.0255, 0.0056], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0222, 0.0236, 0.0167, 0.0267, 0.0181, 0.0190, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:09:43,227 INFO [train.py:904] (3/8) Epoch 2, batch 4100, loss[loss=0.3437, simple_loss=0.3864, pruned_loss=0.1505, over 12108.00 frames. ], tot_loss[loss=0.259, simple_loss=0.3213, pruned_loss=0.09838, over 3262089.51 frames. ], batch size: 248, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:10:59,418 INFO [train.py:904] (3/8) Epoch 2, batch 4150, loss[loss=0.3018, simple_loss=0.3699, pruned_loss=0.1168, over 16876.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.33, pruned_loss=0.1026, over 3237031.50 frames. ], batch size: 109, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:11:25,251 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.427e+02 3.836e+02 4.498e+02 5.328e+02 1.203e+03, threshold=8.995e+02, percent-clipped=3.0 2023-04-27 18:11:30,421 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6377, 3.8637, 3.2207, 3.1402, 2.9686, 2.2441, 4.0205, 4.5373], device='cuda:3'), covar=tensor([0.1625, 0.0422, 0.0796, 0.0441, 0.1529, 0.1112, 0.0239, 0.0085], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0218, 0.0232, 0.0163, 0.0267, 0.0178, 0.0185, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:12:14,210 INFO [train.py:904] (3/8) Epoch 2, batch 4200, loss[loss=0.3043, simple_loss=0.3704, pruned_loss=0.1191, over 16632.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3396, pruned_loss=0.1059, over 3221643.47 frames. ], batch size: 57, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:15,487 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 4250, loss[loss=0.2323, simple_loss=0.3177, pruned_loss=0.07351, over 15398.00 frames. ], tot_loss[loss=0.2768, simple_loss=0.342, pruned_loss=0.1058, over 3190377.09 frames. ], batch size: 190, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:44,150 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:52,833 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:36,047 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4147, 4.3936, 4.1621, 3.5891, 4.3712, 1.6859, 3.9898, 4.2332], device='cuda:3'), covar=tensor([0.0039, 0.0033, 0.0061, 0.0257, 0.0033, 0.1180, 0.0058, 0.0073], device='cuda:3'), in_proj_covar=tensor([0.0059, 0.0050, 0.0073, 0.0095, 0.0057, 0.0103, 0.0069, 0.0074], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:14:38,545 INFO [train.py:904] (3/8) Epoch 2, batch 4300, loss[loss=0.2912, simple_loss=0.3625, pruned_loss=0.11, over 16224.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3425, pruned_loss=0.1043, over 3186676.17 frames. ], batch size: 165, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:14:49,276 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:10,466 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:15:12,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7061, 3.3599, 1.9999, 4.5721, 4.3174, 3.9199, 1.7744, 2.9570], device='cuda:3'), covar=tensor([0.1707, 0.0392, 0.1607, 0.0090, 0.0155, 0.0288, 0.1369, 0.0719], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0117, 0.0169, 0.0067, 0.0101, 0.0109, 0.0155, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 18:15:29,922 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 4350, loss[loss=0.2554, simple_loss=0.3336, pruned_loss=0.08857, over 16674.00 frames. ], tot_loss[loss=0.2805, simple_loss=0.3475, pruned_loss=0.1067, over 3173142.51 frames. ], batch size: 76, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:16:13,768 INFO [optim.py:368] (3/8) 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 18:16:39,123 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3065, 3.3007, 2.6351, 2.7211, 2.4735, 2.0308, 3.4844, 3.9614], device='cuda:3'), covar=tensor([0.1797, 0.0582, 0.0988, 0.0558, 0.1846, 0.1130, 0.0331, 0.0116], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0221, 0.0237, 0.0167, 0.0277, 0.0182, 0.0190, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:16:56,167 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 4400, loss[loss=0.2545, simple_loss=0.334, pruned_loss=0.0875, over 16749.00 frames. ], tot_loss[loss=0.282, simple_loss=0.3493, pruned_loss=0.1074, over 3174486.98 frames. ], batch size: 83, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:17:26,871 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0849, 3.8659, 3.7006, 1.5956, 3.9217, 3.9704, 3.1920, 3.1624], device='cuda:3'), covar=tensor([0.1042, 0.0105, 0.0276, 0.1740, 0.0097, 0.0051, 0.0277, 0.0356], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0080, 0.0078, 0.0151, 0.0074, 0.0073, 0.0100, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:18:09,487 INFO [train.py:904] (3/8) Epoch 2, batch 4450, loss[loss=0.297, simple_loss=0.3728, pruned_loss=0.1106, over 16692.00 frames. ], tot_loss[loss=0.2832, simple_loss=0.3518, pruned_loss=0.1073, over 3180013.08 frames. ], batch size: 76, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:18:34,420 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.338e+02 3.415e+02 4.231e+02 5.024e+02 9.561e+02, threshold=8.461e+02, percent-clipped=1.0 2023-04-27 18:19:17,771 INFO [train.py:904] (3/8) Epoch 2, batch 4500, loss[loss=0.2627, simple_loss=0.3355, pruned_loss=0.09494, over 16699.00 frames. ], tot_loss[loss=0.2811, simple_loss=0.3506, pruned_loss=0.1058, over 3201035.50 frames. ], batch size: 57, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:20:28,458 INFO [train.py:904] (3/8) Epoch 2, batch 4550, loss[loss=0.2815, simple_loss=0.349, pruned_loss=0.107, over 17203.00 frames. ], tot_loss[loss=0.2807, simple_loss=0.3498, pruned_loss=0.1058, over 3208866.82 frames. ], batch size: 44, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:20:47,459 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0674, 4.1256, 4.1213, 1.6772, 4.3962, 4.2190, 3.3110, 3.3673], device='cuda:3'), covar=tensor([0.0907, 0.0068, 0.0113, 0.1482, 0.0033, 0.0032, 0.0247, 0.0294], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0149, 0.0072, 0.0071, 0.0099, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:20:54,772 INFO [optim.py:368] (3/8) 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 18:21:08,392 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:24,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4122, 3.5101, 2.7331, 2.7897, 2.6451, 2.0330, 3.6896, 4.2849], device='cuda:3'), covar=tensor([0.1968, 0.0609, 0.1121, 0.0563, 0.2024, 0.1281, 0.0314, 0.0085], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0221, 0.0233, 0.0165, 0.0280, 0.0177, 0.0185, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:21:25,017 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:37,294 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 18:21:43,175 INFO [train.py:904] (3/8) Epoch 2, batch 4600, loss[loss=0.3157, simple_loss=0.3609, pruned_loss=0.1353, over 11203.00 frames. ], tot_loss[loss=0.2787, simple_loss=0.3494, pruned_loss=0.104, over 3216420.74 frames. ], batch size: 246, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:21:53,284 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:10,140 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,546 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,647 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:22:54,478 INFO [train.py:904] (3/8) Epoch 2, batch 4650, loss[loss=0.2569, simple_loss=0.3235, pruned_loss=0.09517, over 16991.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3472, pruned_loss=0.1028, over 3233401.05 frames. ], batch size: 41, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:22:54,981 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:59,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0013, 1.5665, 2.1568, 2.8766, 3.0434, 2.9203, 1.8458, 3.1003], device='cuda:3'), covar=tensor([0.0030, 0.0268, 0.0109, 0.0081, 0.0032, 0.0052, 0.0180, 0.0024], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0108, 0.0089, 0.0079, 0.0054, 0.0056, 0.0091, 0.0052], device='cuda:3'), out_proj_covar=tensor([1.2354e-04, 1.9453e-04, 1.6768e-04, 1.4875e-04, 9.4725e-05, 1.0430e-04, 1.5912e-04, 9.4996e-05], device='cuda:3') 2023-04-27 18:23:03,206 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 3.221e+02 3.729e+02 4.286e+02 9.218e+02, threshold=7.459e+02, percent-clipped=2.0 2023-04-27 18:23:27,861 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:23:52,671 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6079, 3.4875, 2.8692, 2.8512, 2.7074, 2.0139, 3.6648, 4.2842], device='cuda:3'), covar=tensor([0.1739, 0.0621, 0.0964, 0.0537, 0.1901, 0.1257, 0.0350, 0.0105], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0215, 0.0232, 0.0164, 0.0275, 0.0177, 0.0183, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:23:54,900 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:58,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7655, 2.7315, 1.6240, 2.7291, 2.1712, 2.6940, 1.9609, 2.3210], device='cuda:3'), covar=tensor([0.0067, 0.0164, 0.1204, 0.0075, 0.0605, 0.0468, 0.0975, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0094, 0.0163, 0.0074, 0.0150, 0.0121, 0.0170, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 18:24:06,745 INFO [train.py:904] (3/8) Epoch 2, batch 4700, loss[loss=0.3101, simple_loss=0.3764, pruned_loss=0.1219, over 16481.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3448, pruned_loss=0.1021, over 3206520.79 frames. ], batch size: 146, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:24:10,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7417, 5.4134, 5.3250, 5.2370, 5.1651, 5.7104, 5.4982, 5.2343], device='cuda:3'), covar=tensor([0.0549, 0.0709, 0.0637, 0.1094, 0.1865, 0.0568, 0.0563, 0.1411], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0257, 0.0235, 0.0223, 0.0296, 0.0244, 0.0190, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 18:24:14,213 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0081, 3.4376, 2.6459, 4.5403, 2.3113, 4.3840, 2.6703, 2.5376], device='cuda:3'), covar=tensor([0.0240, 0.0311, 0.0313, 0.0154, 0.1312, 0.0138, 0.0581, 0.1207], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0171, 0.0143, 0.0200, 0.0256, 0.0157, 0.0176, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 18:24:26,114 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4726, 1.4847, 1.9208, 2.5692, 2.6617, 2.6106, 1.6685, 2.5949], device='cuda:3'), covar=tensor([0.0037, 0.0261, 0.0108, 0.0076, 0.0038, 0.0061, 0.0178, 0.0038], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0105, 0.0087, 0.0076, 0.0053, 0.0055, 0.0089, 0.0051], device='cuda:3'), out_proj_covar=tensor([1.2067e-04, 1.8931e-04, 1.6212e-04, 1.4325e-04, 9.3283e-05, 1.0147e-04, 1.5541e-04, 9.4200e-05], device='cuda:3') 2023-04-27 18:25:20,218 INFO [train.py:904] (3/8) Epoch 2, batch 4750, loss[loss=0.274, simple_loss=0.3381, pruned_loss=0.105, over 15395.00 frames. ], tot_loss[loss=0.2711, simple_loss=0.3413, pruned_loss=0.1004, over 3202921.00 frames. ], batch size: 191, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:45,830 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.318e+02 3.402e+02 4.045e+02 5.011e+02 1.217e+03, threshold=8.089e+02, percent-clipped=2.0 2023-04-27 18:26:30,650 INFO [train.py:904] (3/8) Epoch 2, batch 4800, loss[loss=0.2365, simple_loss=0.311, pruned_loss=0.08098, over 16673.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3358, pruned_loss=0.09673, over 3223429.57 frames. ], batch size: 57, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:27:17,376 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3923, 4.2627, 4.1144, 4.2534, 3.6839, 4.1323, 4.1026, 3.9415], device='cuda:3'), covar=tensor([0.0189, 0.0092, 0.0158, 0.0104, 0.0619, 0.0152, 0.0218, 0.0188], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0085, 0.0147, 0.0113, 0.0173, 0.0121, 0.0102, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:27:43,242 INFO [train.py:904] (3/8) Epoch 2, batch 4850, loss[loss=0.2618, simple_loss=0.3355, pruned_loss=0.09401, over 16930.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3383, pruned_loss=0.09775, over 3199829.27 frames. ], batch size: 109, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:28:11,112 INFO [optim.py:368] (3/8) 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 18:28:58,604 INFO [train.py:904] (3/8) Epoch 2, batch 4900, loss[loss=0.2313, simple_loss=0.3013, pruned_loss=0.08067, over 16708.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3379, pruned_loss=0.09671, over 3194119.31 frames. ], batch size: 57, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:29:24,589 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:04,906 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:10,700 INFO [train.py:904] (3/8) Epoch 2, batch 4950, loss[loss=0.2872, simple_loss=0.3633, pruned_loss=0.1055, over 16863.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3387, pruned_loss=0.09701, over 3180127.29 frames. ], batch size: 116, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:30:34,534 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:37,410 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.417e+02 3.788e+02 4.518e+02 5.433e+02 9.876e+02, threshold=9.036e+02, percent-clipped=3.0 2023-04-27 18:30:54,070 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7854, 3.1103, 2.5921, 4.2330, 2.2577, 4.0591, 2.7241, 2.5968], device='cuda:3'), covar=tensor([0.0263, 0.0336, 0.0303, 0.0172, 0.1209, 0.0164, 0.0551, 0.1028], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0170, 0.0144, 0.0200, 0.0250, 0.0158, 0.0177, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:31:12,913 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9962, 3.8141, 3.8523, 3.1498, 3.8093, 1.7331, 3.6100, 3.7663], device='cuda:3'), covar=tensor([0.0075, 0.0072, 0.0068, 0.0380, 0.0062, 0.1288, 0.0088, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0051, 0.0072, 0.0097, 0.0057, 0.0105, 0.0069, 0.0074], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:31:12,917 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:31:22,567 INFO [train.py:904] (3/8) Epoch 2, batch 5000, loss[loss=0.2398, simple_loss=0.3268, pruned_loss=0.07642, over 16865.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3412, pruned_loss=0.09796, over 3189278.47 frames. ], batch size: 102, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:31:48,921 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9972, 3.9506, 3.2668, 1.6227, 2.8407, 2.0913, 3.4228, 4.0102], device='cuda:3'), covar=tensor([0.0274, 0.0274, 0.0495, 0.1721, 0.0726, 0.1138, 0.0635, 0.0256], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0104, 0.0153, 0.0155, 0.0148, 0.0140, 0.0151, 0.0097], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 18:32:21,460 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:26,395 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:34,517 INFO [train.py:904] (3/8) Epoch 2, batch 5050, loss[loss=0.2909, simple_loss=0.3643, pruned_loss=0.1087, over 15386.00 frames. ], tot_loss[loss=0.2675, simple_loss=0.3411, pruned_loss=0.09695, over 3202279.15 frames. ], batch size: 190, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:00,585 INFO [optim.py:368] (3/8) 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 18:33:24,667 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7455, 3.8701, 3.3008, 3.0874, 3.1466, 2.4023, 4.3016, 4.7850], device='cuda:3'), covar=tensor([0.1760, 0.0625, 0.0924, 0.0549, 0.1532, 0.1082, 0.0244, 0.0074], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0222, 0.0234, 0.0168, 0.0264, 0.0179, 0.0191, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:33:45,520 INFO [train.py:904] (3/8) Epoch 2, batch 5100, loss[loss=0.2454, simple_loss=0.3187, pruned_loss=0.08603, over 16244.00 frames. ], tot_loss[loss=0.2636, simple_loss=0.3379, pruned_loss=0.09468, over 3209840.07 frames. ], batch size: 165, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:53,381 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:34:56,468 INFO [train.py:904] (3/8) Epoch 2, batch 5150, loss[loss=0.2666, simple_loss=0.3328, pruned_loss=0.1002, over 16599.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3373, pruned_loss=0.09357, over 3207579.20 frames. ], batch size: 62, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:35:22,256 INFO [optim.py:368] (3/8) 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 18:35:56,835 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 18:36:07,102 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0051, 2.0666, 1.8128, 2.0988, 2.6418, 2.5780, 3.1606, 3.0022], device='cuda:3'), covar=tensor([0.0017, 0.0151, 0.0175, 0.0159, 0.0073, 0.0117, 0.0025, 0.0048], device='cuda:3'), in_proj_covar=tensor([0.0043, 0.0095, 0.0093, 0.0097, 0.0084, 0.0097, 0.0053, 0.0066], device='cuda:3'), out_proj_covar=tensor([6.6227e-05, 1.4946e-04, 1.4348e-04, 1.5742e-04, 1.3788e-04, 1.5827e-04, 8.7188e-05, 1.1148e-04], device='cuda:3') 2023-04-27 18:36:10,019 INFO [train.py:904] (3/8) Epoch 2, batch 5200, loss[loss=0.2689, simple_loss=0.3388, pruned_loss=0.09948, over 17007.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3363, pruned_loss=0.0937, over 3207673.32 frames. ], batch size: 55, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:37:16,258 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 5250, loss[loss=0.2355, simple_loss=0.3218, pruned_loss=0.07457, over 16714.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3339, pruned_loss=0.09336, over 3199769.18 frames. ], batch size: 124, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:37:48,866 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:38:32,782 INFO [train.py:904] (3/8) Epoch 2, batch 5300, loss[loss=0.2272, simple_loss=0.3068, pruned_loss=0.07384, over 16310.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3302, pruned_loss=0.0919, over 3202209.23 frames. ], batch size: 165, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:39:24,103 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2061, 1.5761, 2.2602, 3.0144, 3.0395, 2.9502, 1.6700, 2.9827], device='cuda:3'), covar=tensor([0.0042, 0.0298, 0.0152, 0.0095, 0.0027, 0.0076, 0.0199, 0.0047], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0108, 0.0091, 0.0082, 0.0058, 0.0059, 0.0091, 0.0053], device='cuda:3'), out_proj_covar=tensor([1.3118e-04, 1.9329e-04, 1.6841e-04, 1.5321e-04, 9.9247e-05, 1.0800e-04, 1.5876e-04, 9.4951e-05], device='cuda:3') 2023-04-27 18:39:43,335 INFO [train.py:904] (3/8) Epoch 2, batch 5350, loss[loss=0.243, simple_loss=0.3306, pruned_loss=0.07763, over 16878.00 frames. ], tot_loss[loss=0.2548, simple_loss=0.328, pruned_loss=0.09077, over 3198287.66 frames. ], batch size: 102, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:40:09,981 INFO [optim.py:368] (3/8) 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 18:40:31,241 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 18:40:56,837 INFO [train.py:904] (3/8) Epoch 2, batch 5400, loss[loss=0.2685, simple_loss=0.3393, pruned_loss=0.0989, over 16488.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3319, pruned_loss=0.09277, over 3191484.96 frames. ], batch size: 68, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:40:57,153 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:42:14,657 INFO [train.py:904] (3/8) Epoch 2, batch 5450, loss[loss=0.3072, simple_loss=0.3578, pruned_loss=0.1283, over 12241.00 frames. ], tot_loss[loss=0.2654, simple_loss=0.3374, pruned_loss=0.09665, over 3201004.13 frames. ], batch size: 248, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:42:21,918 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6885, 3.6532, 4.1575, 4.2215, 4.2306, 3.7387, 3.8437, 3.9166], device='cuda:3'), covar=tensor([0.0270, 0.0356, 0.0301, 0.0327, 0.0345, 0.0309, 0.0680, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0157, 0.0176, 0.0171, 0.0205, 0.0167, 0.0257, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 18:42:43,059 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 4.170e+02 4.882e+02 6.060e+02 1.199e+03, threshold=9.765e+02, percent-clipped=3.0 2023-04-27 18:42:50,481 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 18:43:09,237 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9986, 3.7463, 3.7978, 2.9402, 3.7177, 3.7778, 3.8574, 2.0020], device='cuda:3'), covar=tensor([0.0538, 0.0031, 0.0050, 0.0238, 0.0043, 0.0107, 0.0040, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0050, 0.0057, 0.0106, 0.0051, 0.0055, 0.0056, 0.0101], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 18:43:33,386 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1034, 1.6277, 1.4023, 1.4585, 1.6876, 1.6347, 1.7692, 1.8586], device='cuda:3'), covar=tensor([0.0017, 0.0106, 0.0133, 0.0120, 0.0063, 0.0113, 0.0042, 0.0055], device='cuda:3'), in_proj_covar=tensor([0.0041, 0.0096, 0.0095, 0.0100, 0.0085, 0.0098, 0.0054, 0.0069], device='cuda:3'), out_proj_covar=tensor([6.1220e-05, 1.5160e-04, 1.4589e-04, 1.6007e-04, 1.4036e-04, 1.6146e-04, 8.7170e-05, 1.1568e-04], device='cuda:3') 2023-04-27 18:43:34,026 INFO [train.py:904] (3/8) Epoch 2, batch 5500, loss[loss=0.3806, simple_loss=0.4119, pruned_loss=0.1747, over 11276.00 frames. ], tot_loss[loss=0.2799, simple_loss=0.3484, pruned_loss=0.1057, over 3169649.21 frames. ], batch size: 246, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:43:56,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6214, 3.4105, 3.0373, 1.5885, 2.5888, 2.1272, 3.1616, 3.5696], device='cuda:3'), covar=tensor([0.0332, 0.0381, 0.0437, 0.1715, 0.0713, 0.1002, 0.0682, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0108, 0.0147, 0.0152, 0.0145, 0.0137, 0.0149, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 18:44:50,702 INFO [train.py:904] (3/8) Epoch 2, batch 5550, loss[loss=0.352, simple_loss=0.4002, pruned_loss=0.1519, over 15323.00 frames. ], tot_loss[loss=0.2929, simple_loss=0.3583, pruned_loss=0.1137, over 3159160.35 frames. ], batch size: 190, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:45:11,363 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4358, 1.5011, 2.3412, 3.2063, 3.2335, 3.5453, 1.7852, 3.4469], device='cuda:3'), covar=tensor([0.0039, 0.0265, 0.0141, 0.0083, 0.0030, 0.0052, 0.0195, 0.0040], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0109, 0.0093, 0.0081, 0.0058, 0.0057, 0.0091, 0.0053], device='cuda:3'), out_proj_covar=tensor([1.2453e-04, 1.9345e-04, 1.7116e-04, 1.5204e-04, 1.0014e-04, 1.0303e-04, 1.5728e-04, 9.4620e-05], device='cuda:3') 2023-04-27 18:45:19,404 INFO [optim.py:368] (3/8) 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 18:46:11,045 INFO [train.py:904] (3/8) Epoch 2, batch 5600, loss[loss=0.3213, simple_loss=0.3849, pruned_loss=0.1288, over 16788.00 frames. ], tot_loss[loss=0.3042, simple_loss=0.3658, pruned_loss=0.1213, over 3108823.56 frames. ], batch size: 124, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:47:34,289 INFO [train.py:904] (3/8) Epoch 2, batch 5650, loss[loss=0.4371, simple_loss=0.4429, pruned_loss=0.2157, over 11130.00 frames. ], tot_loss[loss=0.3156, simple_loss=0.3732, pruned_loss=0.129, over 3056408.48 frames. ], batch size: 246, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:01,991 INFO [optim.py:368] (3/8) 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 18:48:53,334 INFO [train.py:904] (3/8) Epoch 2, batch 5700, loss[loss=0.3403, simple_loss=0.4019, pruned_loss=0.1393, over 16886.00 frames. ], tot_loss[loss=0.3196, simple_loss=0.3757, pruned_loss=0.1317, over 3035391.39 frames. ], batch size: 116, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:53,813 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:44,718 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:09,229 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:12,128 INFO [train.py:904] (3/8) Epoch 2, batch 5750, loss[loss=0.2953, simple_loss=0.3591, pruned_loss=0.1157, over 16254.00 frames. ], tot_loss[loss=0.3228, simple_loss=0.3786, pruned_loss=0.1335, over 3015260.92 frames. ], batch size: 165, lr: 2.69e-02, grad_scale: 8.0 2023-04-27 18:50:42,007 INFO [optim.py:368] (3/8) 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 18:51:22,948 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 5800, loss[loss=0.2868, simple_loss=0.3527, pruned_loss=0.1104, over 17046.00 frames. ], tot_loss[loss=0.3204, simple_loss=0.3778, pruned_loss=0.1315, over 3021626.65 frames. ], batch size: 53, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:51:51,403 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8198, 5.5295, 5.5357, 5.4424, 5.4676, 5.9938, 5.6846, 5.4523], device='cuda:3'), covar=tensor([0.0549, 0.1057, 0.0802, 0.1095, 0.1814, 0.0651, 0.0759, 0.1472], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0259, 0.0238, 0.0222, 0.0294, 0.0244, 0.0201, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:52:16,401 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3704, 3.3882, 2.7815, 2.6275, 2.5575, 2.0712, 3.5940, 4.0382], device='cuda:3'), covar=tensor([0.1629, 0.0494, 0.0910, 0.0579, 0.1395, 0.1126, 0.0281, 0.0112], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0219, 0.0233, 0.0169, 0.0262, 0.0179, 0.0188, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:52:16,489 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-04-27 18:52:56,022 INFO [train.py:904] (3/8) Epoch 2, batch 5850, loss[loss=0.2781, simple_loss=0.3502, pruned_loss=0.1029, over 16521.00 frames. ], tot_loss[loss=0.3165, simple_loss=0.3749, pruned_loss=0.129, over 3026678.09 frames. ], batch size: 62, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:57,399 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.722e+02 4.866e+02 5.967e+02 7.152e+02 1.262e+03, threshold=1.193e+03, percent-clipped=0.0 2023-04-27 18:53:26,964 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 18:53:30,705 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 18:53:46,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1006, 4.0254, 1.6953, 3.8511, 2.6587, 4.1259, 2.0692, 2.7582], device='cuda:3'), covar=tensor([0.0039, 0.0122, 0.1560, 0.0052, 0.0757, 0.0156, 0.1324, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0101, 0.0168, 0.0074, 0.0157, 0.0125, 0.0173, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 18:54:02,004 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2646, 4.0702, 4.0704, 4.2118, 3.3312, 4.1179, 4.1717, 3.7601], device='cuda:3'), covar=tensor([0.0317, 0.0222, 0.0250, 0.0161, 0.1059, 0.0236, 0.0257, 0.0362], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0089, 0.0149, 0.0116, 0.0180, 0.0123, 0.0105, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 18:54:18,254 INFO [train.py:904] (3/8) Epoch 2, batch 5900, loss[loss=0.3587, simple_loss=0.3938, pruned_loss=0.1618, over 11628.00 frames. ], tot_loss[loss=0.3141, simple_loss=0.3734, pruned_loss=0.1274, over 3036608.22 frames. ], batch size: 248, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:54:40,532 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:17,985 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:21,265 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6242, 3.3266, 3.1246, 2.2923, 3.1206, 3.1578, 3.2789, 1.5772], device='cuda:3'), covar=tensor([0.0564, 0.0037, 0.0052, 0.0307, 0.0050, 0.0080, 0.0030, 0.0524], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0051, 0.0055, 0.0107, 0.0050, 0.0058, 0.0055, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:3') 2023-04-27 18:55:42,172 INFO [train.py:904] (3/8) Epoch 2, batch 5950, loss[loss=0.3034, simple_loss=0.3651, pruned_loss=0.1208, over 16651.00 frames. ], tot_loss[loss=0.3115, simple_loss=0.3726, pruned_loss=0.1252, over 3036881.88 frames. ], batch size: 134, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:56:13,605 INFO [optim.py:368] (3/8) 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 18:56:56,104 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:01,423 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:04,555 INFO [train.py:904] (3/8) Epoch 2, batch 6000, loss[loss=0.314, simple_loss=0.3693, pruned_loss=0.1294, over 15350.00 frames. ], tot_loss[loss=0.3099, simple_loss=0.3714, pruned_loss=0.1242, over 3050287.48 frames. ], batch size: 190, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:57:04,556 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 18:57:15,927 INFO [train.py:938] (3/8) Epoch 2, validation: loss=0.2302, simple_loss=0.3372, pruned_loss=0.06166, over 944034.00 frames. 2023-04-27 18:57:15,927 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 18:57:17,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6973, 4.8333, 4.1565, 4.8614, 4.4355, 4.1327, 4.6251, 4.7815], device='cuda:3'), covar=tensor([0.0682, 0.0952, 0.1633, 0.0432, 0.0727, 0.0799, 0.0720, 0.1038], device='cuda:3'), in_proj_covar=tensor([0.0213, 0.0292, 0.0275, 0.0187, 0.0202, 0.0184, 0.0241, 0.0211], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 18:58:34,798 INFO [train.py:904] (3/8) Epoch 2, batch 6050, loss[loss=0.3147, simple_loss=0.3569, pruned_loss=0.1363, over 11813.00 frames. ], tot_loss[loss=0.3081, simple_loss=0.3696, pruned_loss=0.1233, over 3046342.97 frames. ], batch size: 246, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 18:58:48,702 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:04,139 INFO [optim.py:368] (3/8) 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 18:59:33,627 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:42,352 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0347, 1.4551, 1.7124, 2.0347, 2.0332, 2.0713, 1.3724, 2.0307], device='cuda:3'), covar=tensor([0.0057, 0.0219, 0.0129, 0.0100, 0.0044, 0.0067, 0.0168, 0.0037], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0108, 0.0093, 0.0080, 0.0059, 0.0058, 0.0091, 0.0053], device='cuda:3'), out_proj_covar=tensor([1.2162e-04, 1.9080e-04, 1.6902e-04, 1.4842e-04, 1.0098e-04, 1.0491e-04, 1.5664e-04, 9.3040e-05], device='cuda:3') 2023-04-27 18:59:52,042 INFO [train.py:904] (3/8) Epoch 2, batch 6100, loss[loss=0.2904, simple_loss=0.3609, pruned_loss=0.11, over 16870.00 frames. ], tot_loss[loss=0.3041, simple_loss=0.3675, pruned_loss=0.1203, over 3069572.04 frames. ], batch size: 96, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:00:19,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6025, 4.3203, 4.4095, 4.5585, 3.8938, 4.4209, 4.4737, 4.1505], device='cuda:3'), covar=tensor([0.0374, 0.0189, 0.0216, 0.0113, 0.0828, 0.0212, 0.0192, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0092, 0.0153, 0.0118, 0.0182, 0.0126, 0.0109, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:01:00,878 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2900, 4.1191, 4.1037, 4.2581, 3.6172, 4.1688, 4.1131, 3.8525], device='cuda:3'), covar=tensor([0.0342, 0.0162, 0.0195, 0.0120, 0.0729, 0.0196, 0.0277, 0.0250], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0091, 0.0151, 0.0117, 0.0180, 0.0123, 0.0107, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:01:10,050 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 19:01:17,144 INFO [train.py:904] (3/8) Epoch 2, batch 6150, loss[loss=0.2427, simple_loss=0.3159, pruned_loss=0.08475, over 16418.00 frames. ], tot_loss[loss=0.3016, simple_loss=0.3652, pruned_loss=0.119, over 3086320.73 frames. ], batch size: 75, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:45,849 INFO [optim.py:368] (3/8) 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 19:02:07,793 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7616, 3.6345, 3.4985, 2.5088, 3.5324, 3.6370, 3.6826, 1.8421], device='cuda:3'), covar=tensor([0.0608, 0.0028, 0.0056, 0.0327, 0.0045, 0.0067, 0.0028, 0.0522], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0049, 0.0057, 0.0107, 0.0050, 0.0055, 0.0057, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:02:13,093 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1734, 2.9217, 2.8603, 2.0072, 2.5478, 2.1354, 2.7796, 3.0658], device='cuda:3'), covar=tensor([0.0260, 0.0431, 0.0331, 0.1199, 0.0585, 0.0793, 0.0487, 0.0303], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0107, 0.0151, 0.0153, 0.0146, 0.0136, 0.0150, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 19:02:34,532 INFO [train.py:904] (3/8) Epoch 2, batch 6200, loss[loss=0.2754, simple_loss=0.3459, pruned_loss=0.1024, over 16370.00 frames. ], tot_loss[loss=0.2996, simple_loss=0.3631, pruned_loss=0.1181, over 3100129.32 frames. ], batch size: 75, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:02:45,546 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:06,791 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:51,369 INFO [train.py:904] (3/8) Epoch 2, batch 6250, loss[loss=0.295, simple_loss=0.3654, pruned_loss=0.1123, over 16674.00 frames. ], tot_loss[loss=0.2997, simple_loss=0.3632, pruned_loss=0.1181, over 3090950.41 frames. ], batch size: 134, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:04:18,705 INFO [optim.py:368] (3/8) 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 19:04:37,344 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:51,244 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:05:05,085 INFO [train.py:904] (3/8) Epoch 2, batch 6300, loss[loss=0.3453, simple_loss=0.4003, pruned_loss=0.1452, over 16653.00 frames. ], tot_loss[loss=0.2989, simple_loss=0.3633, pruned_loss=0.1172, over 3109032.28 frames. ], batch size: 134, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:05:58,786 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7432, 1.3835, 1.5896, 1.6331, 1.6309, 1.7926, 1.3785, 1.8335], device='cuda:3'), covar=tensor([0.0059, 0.0172, 0.0085, 0.0085, 0.0046, 0.0053, 0.0147, 0.0037], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0108, 0.0094, 0.0080, 0.0059, 0.0057, 0.0093, 0.0054], device='cuda:3'), out_proj_covar=tensor([1.2074e-04, 1.9058e-04, 1.7129e-04, 1.4721e-04, 1.0070e-04, 1.0341e-04, 1.5880e-04, 9.3494e-05], device='cuda:3') 2023-04-27 19:06:22,079 INFO [train.py:904] (3/8) Epoch 2, batch 6350, loss[loss=0.3852, simple_loss=0.4284, pruned_loss=0.171, over 15331.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3651, pruned_loss=0.1198, over 3092110.64 frames. ], batch size: 190, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:29,789 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.677e+02 5.217e+02 6.547e+02 8.022e+02 1.954e+03, threshold=1.309e+03, percent-clipped=7.0 2023-04-27 19:07:21,296 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 6400, loss[loss=0.3109, simple_loss=0.3658, pruned_loss=0.128, over 15309.00 frames. ], tot_loss[loss=0.305, simple_loss=0.3662, pruned_loss=0.1219, over 3076507.76 frames. ], batch size: 190, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:08:08,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9097, 3.2845, 2.0483, 4.7212, 4.6177, 4.3891, 2.0733, 3.0922], device='cuda:3'), covar=tensor([0.1494, 0.0476, 0.1532, 0.0055, 0.0134, 0.0184, 0.1195, 0.0712], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0121, 0.0166, 0.0068, 0.0107, 0.0116, 0.0152, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 19:08:34,558 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:08:55,789 INFO [train.py:904] (3/8) Epoch 2, batch 6450, loss[loss=0.2741, simple_loss=0.339, pruned_loss=0.1046, over 15425.00 frames. ], tot_loss[loss=0.3027, simple_loss=0.3651, pruned_loss=0.1201, over 3076791.85 frames. ], batch size: 191, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:09:25,708 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.052e+02 4.859e+02 5.740e+02 6.998e+02 1.216e+03, threshold=1.148e+03, percent-clipped=0.0 2023-04-27 19:10:13,782 INFO [train.py:904] (3/8) Epoch 2, batch 6500, loss[loss=0.2929, simple_loss=0.3589, pruned_loss=0.1134, over 15572.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3612, pruned_loss=0.1177, over 3089966.94 frames. ], batch size: 191, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:10:24,348 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:06,999 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5384, 2.6786, 2.3237, 4.0599, 2.0150, 3.7598, 2.3590, 2.4446], device='cuda:3'), covar=tensor([0.0270, 0.0460, 0.0368, 0.0154, 0.1401, 0.0195, 0.0669, 0.0956], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0180, 0.0150, 0.0208, 0.0259, 0.0164, 0.0182, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:11:13,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4340, 3.3638, 2.7326, 2.5037, 2.5322, 1.9323, 3.3406, 3.7597], device='cuda:3'), covar=tensor([0.1728, 0.0607, 0.1042, 0.0734, 0.1629, 0.1225, 0.0329, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0257, 0.0225, 0.0237, 0.0178, 0.0273, 0.0181, 0.0199, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:11:16,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7177, 1.5268, 2.0027, 2.6342, 2.7658, 2.4789, 1.4243, 2.6562], device='cuda:3'), covar=tensor([0.0031, 0.0235, 0.0130, 0.0082, 0.0033, 0.0078, 0.0193, 0.0036], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0109, 0.0095, 0.0082, 0.0061, 0.0059, 0.0095, 0.0053], device='cuda:3'), out_proj_covar=tensor([1.2089e-04, 1.9172e-04, 1.7309e-04, 1.5082e-04, 1.0473e-04, 1.0547e-04, 1.6319e-04, 9.2472e-05], device='cuda:3') 2023-04-27 19:11:32,121 INFO [train.py:904] (3/8) Epoch 2, batch 6550, loss[loss=0.2914, simple_loss=0.3791, pruned_loss=0.1018, over 16867.00 frames. ], tot_loss[loss=0.301, simple_loss=0.3648, pruned_loss=0.1186, over 3103639.52 frames. ], batch size: 102, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:11:37,689 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:32,923 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:47,616 INFO [train.py:904] (3/8) Epoch 2, batch 6600, loss[loss=0.2942, simple_loss=0.3641, pruned_loss=0.1122, over 16713.00 frames. ], tot_loss[loss=0.3049, simple_loss=0.3689, pruned_loss=0.1205, over 3100685.76 frames. ], batch size: 124, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:13:47,418 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:06,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0980, 3.2593, 2.8278, 4.6729, 2.3325, 4.6068, 2.8061, 2.6435], device='cuda:3'), covar=tensor([0.0250, 0.0345, 0.0296, 0.0126, 0.1314, 0.0107, 0.0506, 0.1087], device='cuda:3'), in_proj_covar=tensor([0.0207, 0.0180, 0.0151, 0.0209, 0.0261, 0.0164, 0.0180, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:14:06,686 INFO [train.py:904] (3/8) Epoch 2, batch 6650, loss[loss=0.2415, simple_loss=0.3195, pruned_loss=0.08173, over 17191.00 frames. ], tot_loss[loss=0.3055, simple_loss=0.3687, pruned_loss=0.1212, over 3121641.34 frames. ], batch size: 46, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:14:13,175 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:13,330 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4371, 1.7574, 1.4759, 1.5915, 2.1870, 2.0037, 2.4793, 2.2323], device='cuda:3'), covar=tensor([0.0015, 0.0152, 0.0161, 0.0166, 0.0082, 0.0126, 0.0029, 0.0067], device='cuda:3'), in_proj_covar=tensor([0.0041, 0.0095, 0.0096, 0.0097, 0.0087, 0.0098, 0.0052, 0.0071], device='cuda:3'), out_proj_covar=tensor([6.0065e-05, 1.4779e-04, 1.4566e-04, 1.5174e-04, 1.4249e-04, 1.5899e-04, 8.2774e-05, 1.1668e-04], device='cuda:3') 2023-04-27 19:14:35,746 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 2, batch 6700, loss[loss=0.3604, simple_loss=0.3883, pruned_loss=0.1662, over 11366.00 frames. ], tot_loss[loss=0.3057, simple_loss=0.3677, pruned_loss=0.1218, over 3094315.32 frames. ], batch size: 247, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:15:26,816 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:15:39,417 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-27 19:15:48,911 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1748, 4.2155, 1.7270, 4.2619, 2.6776, 4.2673, 1.9755, 2.9210], device='cuda:3'), covar=tensor([0.0028, 0.0113, 0.1555, 0.0026, 0.0682, 0.0191, 0.1295, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0072, 0.0104, 0.0166, 0.0071, 0.0158, 0.0132, 0.0173, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 19:15:50,565 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:16:38,378 INFO [train.py:904] (3/8) Epoch 2, batch 6750, loss[loss=0.277, simple_loss=0.3404, pruned_loss=0.1068, over 16880.00 frames. ], tot_loss[loss=0.3045, simple_loss=0.3663, pruned_loss=0.1213, over 3104264.13 frames. ], batch size: 116, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:16:47,163 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:07,556 INFO [optim.py:368] (3/8) 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 19:17:22,328 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:17:53,310 INFO [train.py:904] (3/8) Epoch 2, batch 6800, loss[loss=0.2989, simple_loss=0.3652, pruned_loss=0.1163, over 16781.00 frames. ], tot_loss[loss=0.304, simple_loss=0.366, pruned_loss=0.1209, over 3098017.66 frames. ], batch size: 124, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:18:18,723 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:19:09,603 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8207, 3.3568, 2.3539, 4.3268, 4.2646, 4.1788, 1.7293, 3.1992], device='cuda:3'), covar=tensor([0.1539, 0.0343, 0.1278, 0.0054, 0.0169, 0.0191, 0.1267, 0.0593], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0121, 0.0166, 0.0070, 0.0108, 0.0118, 0.0154, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 19:19:10,216 INFO [train.py:904] (3/8) Epoch 2, batch 6850, loss[loss=0.2921, simple_loss=0.3798, pruned_loss=0.1022, over 16845.00 frames. ], tot_loss[loss=0.3059, simple_loss=0.3676, pruned_loss=0.1221, over 3080351.87 frames. ], batch size: 116, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:19:15,626 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3310, 1.7195, 1.5017, 1.5812, 2.1212, 1.9747, 2.2618, 2.1953], device='cuda:3'), covar=tensor([0.0017, 0.0137, 0.0146, 0.0152, 0.0067, 0.0115, 0.0043, 0.0073], device='cuda:3'), in_proj_covar=tensor([0.0043, 0.0098, 0.0098, 0.0100, 0.0089, 0.0101, 0.0054, 0.0071], device='cuda:3'), out_proj_covar=tensor([6.1895e-05, 1.5270e-04, 1.4942e-04, 1.5627e-04, 1.4445e-04, 1.6250e-04, 8.4960e-05, 1.1636e-04], device='cuda:3') 2023-04-27 19:19:38,485 INFO [optim.py:368] (3/8) 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 19:19:49,821 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:20:24,230 INFO [train.py:904] (3/8) Epoch 2, batch 6900, loss[loss=0.2869, simple_loss=0.3478, pruned_loss=0.113, over 16619.00 frames. ], tot_loss[loss=0.3046, simple_loss=0.3689, pruned_loss=0.1202, over 3103875.74 frames. ], batch size: 57, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:20:50,041 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7229, 2.5574, 2.4407, 1.5559, 2.6127, 2.6243, 2.3968, 2.2322], device='cuda:3'), covar=tensor([0.0845, 0.0138, 0.0207, 0.1194, 0.0133, 0.0121, 0.0337, 0.0411], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0086, 0.0081, 0.0154, 0.0078, 0.0072, 0.0107, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:21:01,218 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:27,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9284, 3.7295, 2.4168, 4.5652, 4.4488, 4.1457, 1.9218, 3.1723], device='cuda:3'), covar=tensor([0.1407, 0.0325, 0.1262, 0.0083, 0.0163, 0.0273, 0.1125, 0.0625], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0120, 0.0167, 0.0070, 0.0109, 0.0119, 0.0154, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 19:21:29,103 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4010, 1.3385, 1.6508, 2.1362, 2.2846, 2.3312, 1.4674, 2.4061], device='cuda:3'), covar=tensor([0.0047, 0.0251, 0.0134, 0.0133, 0.0050, 0.0074, 0.0184, 0.0047], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0108, 0.0094, 0.0082, 0.0061, 0.0059, 0.0095, 0.0054], device='cuda:3'), out_proj_covar=tensor([1.2095e-04, 1.8969e-04, 1.6996e-04, 1.5054e-04, 1.0564e-04, 1.0460e-04, 1.6191e-04, 9.3630e-05], device='cuda:3') 2023-04-27 19:21:40,736 INFO [train.py:904] (3/8) Epoch 2, batch 6950, loss[loss=0.3197, simple_loss=0.3798, pruned_loss=0.1298, over 15227.00 frames. ], tot_loss[loss=0.3094, simple_loss=0.3723, pruned_loss=0.1232, over 3097443.82 frames. ], batch size: 190, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:22:09,949 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.451e+02 5.662e+02 6.855e+02 8.747e+02 1.724e+03, threshold=1.371e+03, percent-clipped=6.0 2023-04-27 19:22:54,960 INFO [train.py:904] (3/8) Epoch 2, batch 7000, loss[loss=0.2915, simple_loss=0.3768, pruned_loss=0.1031, over 17127.00 frames. ], tot_loss[loss=0.3083, simple_loss=0.3721, pruned_loss=0.1222, over 3105705.88 frames. ], batch size: 47, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:23:31,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5584, 3.1613, 3.2074, 2.2130, 3.2466, 3.1467, 3.2851, 1.5966], device='cuda:3'), covar=tensor([0.0593, 0.0042, 0.0052, 0.0320, 0.0041, 0.0085, 0.0037, 0.0479], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0054, 0.0057, 0.0110, 0.0054, 0.0059, 0.0058, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:23:53,099 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:10,870 INFO [train.py:904] (3/8) Epoch 2, batch 7050, loss[loss=0.3222, simple_loss=0.3902, pruned_loss=0.1271, over 16351.00 frames. ], tot_loss[loss=0.3086, simple_loss=0.3731, pruned_loss=0.1221, over 3110734.46 frames. ], batch size: 146, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:24:37,723 INFO [optim.py:368] (3/8) 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 19:24:43,954 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:25:22,996 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:23,674 INFO [train.py:904] (3/8) Epoch 2, batch 7100, loss[loss=0.2736, simple_loss=0.348, pruned_loss=0.09959, over 16712.00 frames. ], tot_loss[loss=0.308, simple_loss=0.3717, pruned_loss=0.1221, over 3104962.48 frames. ], batch size: 124, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:25:40,645 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:26:01,918 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6424, 5.8987, 5.4704, 5.7241, 5.0409, 4.7821, 5.4098, 5.9780], device='cuda:3'), covar=tensor([0.0328, 0.0474, 0.0840, 0.0290, 0.0403, 0.0402, 0.0366, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0295, 0.0276, 0.0187, 0.0204, 0.0186, 0.0248, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:26:38,661 INFO [train.py:904] (3/8) Epoch 2, batch 7150, loss[loss=0.3085, simple_loss=0.3801, pruned_loss=0.1184, over 16765.00 frames. ], tot_loss[loss=0.3039, simple_loss=0.3678, pruned_loss=0.12, over 3126360.38 frames. ], batch size: 89, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:27:07,389 INFO [optim.py:368] (3/8) 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 19:27:16,444 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-27 19:27:52,977 INFO [train.py:904] (3/8) Epoch 2, batch 7200, loss[loss=0.2601, simple_loss=0.3402, pruned_loss=0.09004, over 16393.00 frames. ], tot_loss[loss=0.3002, simple_loss=0.365, pruned_loss=0.1177, over 3121505.62 frames. ], batch size: 146, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:27:53,635 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8755, 2.3775, 2.0927, 3.1183, 2.1404, 2.9534, 2.2721, 2.0403], device='cuda:3'), covar=tensor([0.0319, 0.0461, 0.0345, 0.0220, 0.1202, 0.0219, 0.0666, 0.1034], device='cuda:3'), in_proj_covar=tensor([0.0217, 0.0188, 0.0160, 0.0217, 0.0268, 0.0173, 0.0191, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:28:02,341 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8062, 3.0814, 2.2314, 4.0258, 3.9403, 3.7687, 1.7107, 2.8349], device='cuda:3'), covar=tensor([0.1427, 0.0420, 0.1323, 0.0069, 0.0142, 0.0329, 0.1292, 0.0673], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0122, 0.0167, 0.0071, 0.0109, 0.0122, 0.0156, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 19:28:10,409 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:28:44,453 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 7250, loss[loss=0.24, simple_loss=0.3091, pruned_loss=0.08543, over 16794.00 frames. ], tot_loss[loss=0.2967, simple_loss=0.3617, pruned_loss=0.1158, over 3109851.06 frames. ], batch size: 124, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:29:42,538 INFO [optim.py:368] (3/8) 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 19:29:47,306 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:30:05,832 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8669, 3.7585, 4.3393, 4.3359, 4.3203, 3.8474, 4.0082, 3.9873], device='cuda:3'), covar=tensor([0.0257, 0.0323, 0.0335, 0.0364, 0.0398, 0.0296, 0.0757, 0.0369], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0150, 0.0168, 0.0164, 0.0199, 0.0167, 0.0254, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-27 19:30:19,113 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:30:19,476 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 19:30:29,420 INFO [train.py:904] (3/8) Epoch 2, batch 7300, loss[loss=0.3403, simple_loss=0.3746, pruned_loss=0.153, over 11703.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.3604, pruned_loss=0.1153, over 3109999.90 frames. ], batch size: 248, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:31:46,483 INFO [train.py:904] (3/8) Epoch 2, batch 7350, loss[loss=0.2751, simple_loss=0.3396, pruned_loss=0.1053, over 16727.00 frames. ], tot_loss[loss=0.2953, simple_loss=0.3602, pruned_loss=0.1152, over 3086249.11 frames. ], batch size: 124, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:31:53,426 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0785, 3.7560, 3.4740, 1.6802, 2.7178, 2.2220, 3.3421, 3.9126], device='cuda:3'), covar=tensor([0.0280, 0.0404, 0.0433, 0.1766, 0.0789, 0.0992, 0.0745, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0108, 0.0153, 0.0153, 0.0146, 0.0136, 0.0152, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 19:32:16,421 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.647e+02 4.932e+02 5.634e+02 7.018e+02 1.459e+03, threshold=1.127e+03, percent-clipped=2.0 2023-04-27 19:32:22,398 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:32:49,596 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 19:32:57,423 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:33:06,150 INFO [train.py:904] (3/8) Epoch 2, batch 7400, loss[loss=0.3545, simple_loss=0.4062, pruned_loss=0.1514, over 16295.00 frames. ], tot_loss[loss=0.2973, simple_loss=0.3614, pruned_loss=0.1165, over 3072893.65 frames. ], batch size: 165, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:33:25,371 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:33:41,416 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:33:47,700 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 19:33:55,837 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-27 19:34:03,877 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2607, 1.5463, 2.0300, 2.9946, 3.0656, 3.0778, 1.6885, 3.0605], device='cuda:3'), covar=tensor([0.0035, 0.0239, 0.0142, 0.0068, 0.0030, 0.0050, 0.0186, 0.0032], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0111, 0.0095, 0.0080, 0.0061, 0.0059, 0.0098, 0.0055], device='cuda:3'), out_proj_covar=tensor([1.2147e-04, 1.9370e-04, 1.7242e-04, 1.4584e-04, 1.0405e-04, 1.0425e-04, 1.6661e-04, 9.4163e-05], device='cuda:3') 2023-04-27 19:34:27,189 INFO [train.py:904] (3/8) Epoch 2, batch 7450, loss[loss=0.2803, simple_loss=0.3602, pruned_loss=0.1002, over 16779.00 frames. ], tot_loss[loss=0.2998, simple_loss=0.3633, pruned_loss=0.1181, over 3072135.13 frames. ], batch size: 124, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:34:43,751 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:34:59,851 INFO [optim.py:368] (3/8) 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 19:35:06,748 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 19:35:13,182 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7564, 4.4657, 4.7055, 5.0297, 5.1129, 4.5107, 5.1137, 5.0006], device='cuda:3'), covar=tensor([0.0437, 0.0505, 0.0814, 0.0292, 0.0261, 0.0402, 0.0226, 0.0263], device='cuda:3'), in_proj_covar=tensor([0.0238, 0.0277, 0.0378, 0.0280, 0.0221, 0.0204, 0.0220, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:35:49,109 INFO [train.py:904] (3/8) Epoch 2, batch 7500, loss[loss=0.2616, simple_loss=0.3351, pruned_loss=0.09411, over 16869.00 frames. ], tot_loss[loss=0.2984, simple_loss=0.3627, pruned_loss=0.1171, over 3080191.64 frames. ], batch size: 116, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:36:27,956 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 19:36:32,432 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1768, 5.8870, 5.8419, 5.7950, 5.7842, 6.2784, 6.0152, 5.8073], device='cuda:3'), covar=tensor([0.0545, 0.0902, 0.0770, 0.1215, 0.1989, 0.0636, 0.0694, 0.1491], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0278, 0.0258, 0.0247, 0.0310, 0.0273, 0.0214, 0.0325], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:37:05,042 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:05,673 INFO [train.py:904] (3/8) Epoch 2, batch 7550, loss[loss=0.271, simple_loss=0.3325, pruned_loss=0.1047, over 16568.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3621, pruned_loss=0.1173, over 3094138.10 frames. ], batch size: 62, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:37:10,244 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4261, 5.0518, 5.1901, 5.2021, 4.6488, 5.2327, 5.0829, 4.9300], device='cuda:3'), covar=tensor([0.0173, 0.0141, 0.0115, 0.0088, 0.0576, 0.0121, 0.0098, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0094, 0.0146, 0.0118, 0.0174, 0.0126, 0.0107, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:37:21,042 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 19:37:32,558 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.677e+02 5.137e+02 6.469e+02 8.298e+02 1.927e+03, threshold=1.294e+03, percent-clipped=3.0 2023-04-27 19:37:51,172 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:55,752 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 19:38:04,540 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:38:22,208 INFO [train.py:904] (3/8) Epoch 2, batch 7600, loss[loss=0.3376, simple_loss=0.3737, pruned_loss=0.1507, over 11422.00 frames. ], tot_loss[loss=0.2972, simple_loss=0.3608, pruned_loss=0.1168, over 3103626.27 frames. ], batch size: 248, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:38:37,883 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:40,781 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8234, 3.7483, 3.6240, 2.8668, 3.6873, 1.7013, 3.5099, 3.6024], device='cuda:3'), covar=tensor([0.0085, 0.0068, 0.0096, 0.0402, 0.0070, 0.1536, 0.0088, 0.0132], device='cuda:3'), in_proj_covar=tensor([0.0062, 0.0051, 0.0076, 0.0098, 0.0058, 0.0109, 0.0069, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:38:45,247 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4049, 3.3233, 1.3441, 3.4303, 2.2983, 3.4185, 1.8103, 2.5289], device='cuda:3'), covar=tensor([0.0049, 0.0153, 0.1644, 0.0045, 0.0793, 0.0282, 0.1313, 0.0589], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0111, 0.0174, 0.0073, 0.0161, 0.0139, 0.0179, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 19:39:24,533 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:24,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 19:39:39,705 INFO [train.py:904] (3/8) Epoch 2, batch 7650, loss[loss=0.3455, simple_loss=0.3907, pruned_loss=0.1502, over 15481.00 frames. ], tot_loss[loss=0.3003, simple_loss=0.363, pruned_loss=0.1188, over 3093008.52 frames. ], batch size: 191, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:39:47,262 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 19:40:10,726 INFO [optim.py:368] (3/8) 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 19:40:49,709 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:40:58,861 INFO [train.py:904] (3/8) Epoch 2, batch 7700, loss[loss=0.3071, simple_loss=0.3713, pruned_loss=0.1214, over 15474.00 frames. ], tot_loss[loss=0.3016, simple_loss=0.3638, pruned_loss=0.1197, over 3081112.07 frames. ], batch size: 190, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:42:04,291 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:16,828 INFO [train.py:904] (3/8) Epoch 2, batch 7750, loss[loss=0.2929, simple_loss=0.3607, pruned_loss=0.1125, over 16587.00 frames. ], tot_loss[loss=0.3006, simple_loss=0.3635, pruned_loss=0.1189, over 3083482.73 frames. ], batch size: 62, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:42:39,857 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-04-27 19:42:46,602 INFO [optim.py:368] (3/8) 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 19:43:31,901 INFO [train.py:904] (3/8) Epoch 2, batch 7800, loss[loss=0.3738, simple_loss=0.4031, pruned_loss=0.1722, over 11408.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3647, pruned_loss=0.1199, over 3084192.13 frames. ], batch size: 248, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:43:47,100 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:44:52,421 INFO [train.py:904] (3/8) Epoch 2, batch 7850, loss[loss=0.3213, simple_loss=0.387, pruned_loss=0.1278, over 16691.00 frames. ], tot_loss[loss=0.3048, simple_loss=0.3671, pruned_loss=0.1213, over 3063725.11 frames. ], batch size: 134, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:45:18,593 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 19:45:22,917 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:36,110 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-27 19:45:48,775 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:46:05,623 INFO [train.py:904] (3/8) Epoch 2, batch 7900, loss[loss=0.3381, simple_loss=0.3934, pruned_loss=0.1414, over 17125.00 frames. ], tot_loss[loss=0.3041, simple_loss=0.3659, pruned_loss=0.1211, over 3047858.01 frames. ], batch size: 47, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:46:13,957 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:46:29,401 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:05,057 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 7950, loss[loss=0.2819, simple_loss=0.3511, pruned_loss=0.1064, over 16640.00 frames. ], tot_loss[loss=0.3051, simple_loss=0.3667, pruned_loss=0.1218, over 3048376.22 frames. ], batch size: 134, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:47:48,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2361, 3.1682, 3.2752, 3.4945, 3.4313, 3.2109, 3.4361, 3.4459], device='cuda:3'), covar=tensor([0.0470, 0.0454, 0.0875, 0.0343, 0.0458, 0.0924, 0.0438, 0.0373], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0288, 0.0383, 0.0284, 0.0222, 0.0202, 0.0226, 0.0229], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:47:56,298 INFO [optim.py:368] (3/8) 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 19:48:12,983 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-27 19:48:42,376 INFO [train.py:904] (3/8) Epoch 2, batch 8000, loss[loss=0.3118, simple_loss=0.3773, pruned_loss=0.1232, over 16815.00 frames. ], tot_loss[loss=0.3035, simple_loss=0.3655, pruned_loss=0.1208, over 3054082.34 frames. ], batch size: 83, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:49:16,778 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:18,135 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:31,044 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-04-27 19:49:56,235 INFO [train.py:904] (3/8) Epoch 2, batch 8050, loss[loss=0.2973, simple_loss=0.3586, pruned_loss=0.118, over 16890.00 frames. ], tot_loss[loss=0.302, simple_loss=0.3646, pruned_loss=0.1197, over 3059517.13 frames. ], batch size: 116, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:50:24,924 INFO [optim.py:368] (3/8) 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 19:50:46,738 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,664 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:51:11,141 INFO [train.py:904] (3/8) Epoch 2, batch 8100, loss[loss=0.2866, simple_loss=0.3578, pruned_loss=0.1077, over 16461.00 frames. ], tot_loss[loss=0.3021, simple_loss=0.3646, pruned_loss=0.1199, over 3046646.33 frames. ], batch size: 75, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:52:29,017 INFO [train.py:904] (3/8) Epoch 2, batch 8150, loss[loss=0.2906, simple_loss=0.3517, pruned_loss=0.1148, over 15340.00 frames. ], tot_loss[loss=0.2993, simple_loss=0.3622, pruned_loss=0.1183, over 3059530.16 frames. ], batch size: 190, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:52:52,889 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:53:00,708 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 2, batch 8200, loss[loss=0.2625, simple_loss=0.3348, pruned_loss=0.09508, over 16498.00 frames. ], tot_loss[loss=0.296, simple_loss=0.3588, pruned_loss=0.1166, over 3079583.41 frames. ], batch size: 75, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:53:57,235 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:37,968 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-27 19:54:46,789 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 8250, loss[loss=0.2489, simple_loss=0.3319, pruned_loss=0.08295, over 16433.00 frames. ], tot_loss[loss=0.2929, simple_loss=0.3575, pruned_loss=0.1141, over 3063397.02 frames. ], batch size: 68, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:55:15,137 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:44,642 INFO [optim.py:368] (3/8) 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 19:56:05,977 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:21,630 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6497, 1.8265, 1.6949, 1.7538, 2.3769, 2.1187, 2.7720, 2.6434], device='cuda:3'), covar=tensor([0.0020, 0.0164, 0.0168, 0.0186, 0.0078, 0.0129, 0.0025, 0.0065], device='cuda:3'), in_proj_covar=tensor([0.0047, 0.0099, 0.0103, 0.0105, 0.0092, 0.0102, 0.0056, 0.0074], device='cuda:3'), out_proj_covar=tensor([6.6701e-05, 1.5089e-04, 1.5504e-04, 1.6287e-04, 1.4778e-04, 1.6196e-04, 8.6737e-05, 1.1975e-04], device='cuda:3') 2023-04-27 19:56:21,938 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 19:56:33,571 INFO [train.py:904] (3/8) Epoch 2, batch 8300, loss[loss=0.2531, simple_loss=0.3162, pruned_loss=0.09501, over 11929.00 frames. ], tot_loss[loss=0.2857, simple_loss=0.3528, pruned_loss=0.1093, over 3047642.88 frames. ], batch size: 246, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:57:37,591 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 8350, loss[loss=0.2473, simple_loss=0.3345, pruned_loss=0.08003, over 16873.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3497, pruned_loss=0.1043, over 3076056.13 frames. ], batch size: 96, lr: 2.50e-02, grad_scale: 4.0 2023-04-27 19:58:06,652 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:58:28,816 INFO [optim.py:368] (3/8) 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 19:58:34,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3284, 3.1926, 3.3117, 3.5387, 3.4718, 3.2468, 3.4756, 3.5118], device='cuda:3'), covar=tensor([0.0469, 0.0511, 0.0931, 0.0426, 0.0468, 0.1048, 0.0485, 0.0333], device='cuda:3'), in_proj_covar=tensor([0.0236, 0.0284, 0.0372, 0.0282, 0.0214, 0.0198, 0.0223, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:58:42,168 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:44,100 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:55,557 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8309, 3.5162, 3.3830, 2.5061, 3.3839, 3.2620, 3.4723, 1.6746], device='cuda:3'), covar=tensor([0.0505, 0.0026, 0.0044, 0.0270, 0.0038, 0.0061, 0.0037, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0053, 0.0057, 0.0106, 0.0052, 0.0058, 0.0058, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 19:59:15,561 INFO [train.py:904] (3/8) Epoch 2, batch 8400, loss[loss=0.2733, simple_loss=0.3445, pruned_loss=0.101, over 16675.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3449, pruned_loss=0.101, over 3051011.53 frames. ], batch size: 134, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 19:59:16,061 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:27,410 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4793, 3.3560, 3.4339, 3.1108, 3.4619, 2.0395, 3.1553, 3.1665], device='cuda:3'), covar=tensor([0.0063, 0.0056, 0.0070, 0.0164, 0.0051, 0.1135, 0.0079, 0.0094], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0049, 0.0076, 0.0088, 0.0057, 0.0108, 0.0069, 0.0074], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 19:59:45,486 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:00:06,087 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 20:00:35,047 INFO [train.py:904] (3/8) Epoch 2, batch 8450, loss[loss=0.2481, simple_loss=0.316, pruned_loss=0.09011, over 12077.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3422, pruned_loss=0.09879, over 3036745.07 frames. ], batch size: 246, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 20:01:00,268 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 20:01:29,776 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 20:01:36,640 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8533, 1.4520, 1.9034, 2.7107, 2.6093, 2.5003, 1.7224, 2.5421], device='cuda:3'), covar=tensor([0.0030, 0.0211, 0.0116, 0.0070, 0.0036, 0.0066, 0.0158, 0.0047], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0110, 0.0094, 0.0082, 0.0061, 0.0057, 0.0096, 0.0055], device='cuda:3'), out_proj_covar=tensor([1.1988e-04, 1.8944e-04, 1.6741e-04, 1.4580e-04, 1.0272e-04, 9.7912e-05, 1.6302e-04, 9.2135e-05], device='cuda:3') 2023-04-27 20:01:55,557 INFO [train.py:904] (3/8) Epoch 2, batch 8500, loss[loss=0.2423, simple_loss=0.3207, pruned_loss=0.082, over 16572.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3368, pruned_loss=0.09464, over 3035523.69 frames. ], batch size: 62, lr: 2.49e-02, grad_scale: 8.0 2023-04-27 20:02:17,754 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:33,869 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:03:19,432 INFO [train.py:904] (3/8) Epoch 2, batch 8550, loss[loss=0.2621, simple_loss=0.3454, pruned_loss=0.08941, over 16726.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3328, pruned_loss=0.09244, over 3019146.42 frames. ], batch size: 83, lr: 2.49e-02, grad_scale: 4.0 2023-04-27 20:03:27,088 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 20:03:58,670 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 20:04:03,004 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.327e+02 4.014e+02 4.987e+02 6.688e+02 1.627e+03, threshold=9.973e+02, percent-clipped=3.0 2023-04-27 20:04:26,398 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:04:58,635 INFO [train.py:904] (3/8) Epoch 2, batch 8600, loss[loss=0.265, simple_loss=0.3479, pruned_loss=0.09105, over 16390.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3338, pruned_loss=0.09154, over 3036995.91 frames. ], batch size: 75, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:05:21,872 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9418, 2.4845, 2.2454, 3.3334, 3.2231, 3.2522, 1.8968, 2.7328], device='cuda:3'), covar=tensor([0.1266, 0.0359, 0.1109, 0.0083, 0.0194, 0.0316, 0.1051, 0.0592], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0125, 0.0171, 0.0071, 0.0114, 0.0125, 0.0158, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 20:06:00,971 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8151, 3.8181, 4.2766, 4.2861, 4.2795, 3.8722, 3.9909, 3.9619], device='cuda:3'), covar=tensor([0.0255, 0.0294, 0.0405, 0.0485, 0.0360, 0.0247, 0.0619, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0145, 0.0160, 0.0154, 0.0189, 0.0159, 0.0240, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-27 20:06:33,193 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-04-27 20:06:37,155 INFO [train.py:904] (3/8) Epoch 2, batch 8650, loss[loss=0.2084, simple_loss=0.3057, pruned_loss=0.05558, over 16876.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3318, pruned_loss=0.08919, over 3033574.32 frames. ], batch size: 102, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:07:29,096 INFO [optim.py:368] (3/8) 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 20:07:44,361 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:46,470 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,547 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 8700, loss[loss=0.2093, simple_loss=0.2995, pruned_loss=0.05955, over 16917.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3278, pruned_loss=0.08665, over 3038092.84 frames. ], batch size: 102, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:08:31,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5184, 3.6594, 3.6650, 1.6715, 3.9368, 3.8793, 3.1778, 3.1073], device='cuda:3'), covar=tensor([0.0610, 0.0092, 0.0139, 0.1300, 0.0047, 0.0043, 0.0297, 0.0284], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0082, 0.0082, 0.0152, 0.0072, 0.0071, 0.0108, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:08:49,478 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:09:14,854 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:17,052 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:49,562 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7726, 2.5727, 2.6740, 2.0742, 2.4810, 2.4809, 2.7076, 1.7638], device='cuda:3'), covar=tensor([0.0338, 0.0051, 0.0054, 0.0222, 0.0061, 0.0074, 0.0038, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0053, 0.0056, 0.0103, 0.0052, 0.0058, 0.0056, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:10:01,022 INFO [train.py:904] (3/8) Epoch 2, batch 8750, loss[loss=0.239, simple_loss=0.3151, pruned_loss=0.08146, over 11935.00 frames. ], tot_loss[loss=0.2492, simple_loss=0.3269, pruned_loss=0.08571, over 3032893.64 frames. ], batch size: 247, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:10:57,517 INFO [optim.py:368] (3/8) 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 20:11:10,251 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:11:53,568 INFO [train.py:904] (3/8) Epoch 2, batch 8800, loss[loss=0.2636, simple_loss=0.3439, pruned_loss=0.09161, over 16859.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3258, pruned_loss=0.08471, over 3053681.77 frames. ], batch size: 116, lr: 2.48e-02, grad_scale: 4.0 2023-04-27 20:12:52,462 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-27 20:13:18,182 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:13:38,729 INFO [train.py:904] (3/8) Epoch 2, batch 8850, loss[loss=0.2265, simple_loss=0.319, pruned_loss=0.06698, over 15234.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3269, pruned_loss=0.08321, over 3058843.35 frames. ], batch size: 190, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:14:28,277 INFO [optim.py:368] (3/8) 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 20:14:43,426 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:15:27,232 INFO [train.py:904] (3/8) Epoch 2, batch 8900, loss[loss=0.2285, simple_loss=0.3121, pruned_loss=0.07251, over 16906.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3265, pruned_loss=0.08192, over 3057429.65 frames. ], batch size: 96, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:15:31,406 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-04-27 20:16:27,632 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 20:17:32,216 INFO [train.py:904] (3/8) Epoch 2, batch 8950, loss[loss=0.2911, simple_loss=0.3563, pruned_loss=0.113, over 12785.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3265, pruned_loss=0.08238, over 3073081.33 frames. ], batch size: 247, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:17:59,477 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-27 20:18:00,967 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7370, 2.7213, 1.6514, 2.8817, 2.1815, 2.7593, 1.9323, 2.4548], device='cuda:3'), covar=tensor([0.0079, 0.0226, 0.1359, 0.0060, 0.0763, 0.0433, 0.1224, 0.0560], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0111, 0.0174, 0.0075, 0.0157, 0.0137, 0.0181, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 20:18:08,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9028, 4.6854, 4.5704, 4.6883, 4.3286, 4.5502, 4.5887, 4.3270], device='cuda:3'), covar=tensor([0.0189, 0.0131, 0.0140, 0.0094, 0.0458, 0.0171, 0.0139, 0.0250], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0094, 0.0141, 0.0116, 0.0167, 0.0123, 0.0101, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:18:20,856 INFO [optim.py:368] (3/8) 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 20:18:32,638 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:19:09,088 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0235, 4.1133, 4.2254, 4.1793, 4.2548, 4.6034, 4.4961, 4.1496], device='cuda:3'), covar=tensor([0.1271, 0.1141, 0.0960, 0.1344, 0.1832, 0.0793, 0.0715, 0.1780], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0253, 0.0241, 0.0231, 0.0289, 0.0265, 0.0200, 0.0296], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:19:11,781 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:21,317 INFO [train.py:904] (3/8) Epoch 2, batch 9000, loss[loss=0.2108, simple_loss=0.2961, pruned_loss=0.06274, over 16807.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3226, pruned_loss=0.08025, over 3071775.51 frames. ], batch size: 90, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:19:21,318 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 20:19:31,140 INFO [train.py:938] (3/8) Epoch 2, validation: loss=0.2044, simple_loss=0.3047, pruned_loss=0.05205, over 944034.00 frames. 2023-04-27 20:19:31,141 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 20:20:00,845 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:20:13,331 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-27 20:20:18,327 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1901, 2.4311, 2.1470, 3.5551, 1.8120, 3.3190, 2.3139, 2.1026], device='cuda:3'), covar=tensor([0.0299, 0.0528, 0.0380, 0.0180, 0.1480, 0.0192, 0.0659, 0.1173], device='cuda:3'), in_proj_covar=tensor([0.0214, 0.0194, 0.0161, 0.0216, 0.0266, 0.0173, 0.0196, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:20:52,979 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:21:02,063 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 9050, loss[loss=0.2233, simple_loss=0.3068, pruned_loss=0.06996, over 16205.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3243, pruned_loss=0.08145, over 3079672.97 frames. ], batch size: 165, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:21:37,536 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:22:01,198 INFO [optim.py:368] (3/8) 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 20:22:13,487 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 20:22:58,687 INFO [train.py:904] (3/8) Epoch 2, batch 9100, loss[loss=0.255, simple_loss=0.343, pruned_loss=0.08353, over 16448.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3238, pruned_loss=0.0821, over 3070094.17 frames. ], batch size: 147, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:24:22,919 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:55,899 INFO [train.py:904] (3/8) Epoch 2, batch 9150, loss[loss=0.2385, simple_loss=0.3276, pruned_loss=0.07472, over 16910.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3249, pruned_loss=0.08195, over 3067353.43 frames. ], batch size: 109, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:25:05,955 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 20:25:29,650 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 20:25:49,332 INFO [optim.py:368] (3/8) 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 20:25:52,735 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-27 20:25:59,025 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:26:40,137 INFO [train.py:904] (3/8) Epoch 2, batch 9200, loss[loss=0.2177, simple_loss=0.2896, pruned_loss=0.07294, over 12318.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3192, pruned_loss=0.07973, over 3079074.34 frames. ], batch size: 248, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:27:30,641 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:27:32,188 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0568, 4.2736, 2.4214, 5.0363, 4.9628, 4.6400, 2.3317, 3.2418], device='cuda:3'), covar=tensor([0.1262, 0.0243, 0.1222, 0.0080, 0.0133, 0.0187, 0.1040, 0.0615], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0125, 0.0166, 0.0070, 0.0111, 0.0121, 0.0153, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 20:28:03,164 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 9250, loss[loss=0.2044, simple_loss=0.2823, pruned_loss=0.06325, over 12347.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3193, pruned_loss=0.07996, over 3079589.83 frames. ], batch size: 248, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:28:17,812 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3272, 3.4136, 2.8507, 2.3552, 2.3491, 2.0928, 3.4947, 3.8709], device='cuda:3'), covar=tensor([0.1783, 0.0581, 0.0912, 0.0764, 0.1508, 0.1157, 0.0332, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0227, 0.0238, 0.0180, 0.0214, 0.0184, 0.0198, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:28:35,731 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8566, 4.1334, 3.8810, 3.9140, 3.4796, 3.7796, 3.8311, 4.0859], device='cuda:3'), covar=tensor([0.0474, 0.0693, 0.0672, 0.0353, 0.0557, 0.0620, 0.0478, 0.0590], device='cuda:3'), in_proj_covar=tensor([0.0212, 0.0290, 0.0254, 0.0183, 0.0196, 0.0184, 0.0232, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:29:05,881 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.723e+02 4.011e+02 4.767e+02 6.782e+02 2.707e+03, threshold=9.534e+02, percent-clipped=7.0 2023-04-27 20:29:34,553 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9139, 4.7250, 4.7562, 4.2553, 4.6722, 2.0482, 4.5452, 4.7521], device='cuda:3'), covar=tensor([0.0078, 0.0056, 0.0064, 0.0178, 0.0053, 0.1184, 0.0058, 0.0084], device='cuda:3'), in_proj_covar=tensor([0.0059, 0.0048, 0.0073, 0.0082, 0.0056, 0.0107, 0.0067, 0.0073], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:29:34,677 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9005, 4.1516, 3.6294, 2.8601, 3.0426, 2.3905, 4.3533, 4.7891], device='cuda:3'), covar=tensor([0.1557, 0.0490, 0.0707, 0.0621, 0.1499, 0.1086, 0.0232, 0.0162], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0226, 0.0236, 0.0180, 0.0213, 0.0183, 0.0196, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:30:06,013 INFO [train.py:904] (3/8) Epoch 2, batch 9300, loss[loss=0.2277, simple_loss=0.3006, pruned_loss=0.07738, over 12355.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3165, pruned_loss=0.0784, over 3081869.94 frames. ], batch size: 247, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:30:14,363 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:31:20,840 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:31:40,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8133, 2.7120, 1.5240, 2.7492, 2.0818, 2.7146, 1.9026, 2.4566], device='cuda:3'), covar=tensor([0.0110, 0.0255, 0.1433, 0.0057, 0.0699, 0.0487, 0.1173, 0.0608], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0111, 0.0172, 0.0073, 0.0153, 0.0134, 0.0175, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 20:31:49,020 INFO [train.py:904] (3/8) Epoch 2, batch 9350, loss[loss=0.2516, simple_loss=0.3194, pruned_loss=0.09193, over 16576.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3159, pruned_loss=0.07827, over 3085929.37 frames. ], batch size: 62, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:31:49,931 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2726, 3.9335, 3.8015, 1.5802, 3.9123, 3.9609, 3.2379, 3.1234], device='cuda:3'), covar=tensor([0.0830, 0.0056, 0.0146, 0.1541, 0.0053, 0.0041, 0.0304, 0.0306], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0081, 0.0081, 0.0149, 0.0073, 0.0069, 0.0106, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:32:37,319 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.597e+02 3.883e+02 4.622e+02 6.098e+02 2.142e+03, threshold=9.245e+02, percent-clipped=3.0 2023-04-27 20:33:28,470 INFO [train.py:904] (3/8) Epoch 2, batch 9400, loss[loss=0.2244, simple_loss=0.2995, pruned_loss=0.07465, over 12441.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3159, pruned_loss=0.07785, over 3080632.94 frames. ], batch size: 248, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:33:29,548 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6485, 1.8430, 1.3890, 1.4947, 2.3666, 2.1181, 2.6183, 2.5323], device='cuda:3'), covar=tensor([0.0022, 0.0171, 0.0211, 0.0201, 0.0079, 0.0133, 0.0032, 0.0068], device='cuda:3'), in_proj_covar=tensor([0.0050, 0.0103, 0.0101, 0.0105, 0.0092, 0.0102, 0.0055, 0.0075], device='cuda:3'), out_proj_covar=tensor([6.8681e-05, 1.5780e-04, 1.4982e-04, 1.5965e-04, 1.4326e-04, 1.5921e-04, 7.9984e-05, 1.1775e-04], device='cuda:3') 2023-04-27 20:34:39,252 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 9450, loss[loss=0.2306, simple_loss=0.3196, pruned_loss=0.07085, over 17201.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3175, pruned_loss=0.07857, over 3043913.76 frames. ], batch size: 45, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:35:42,248 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:56,417 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 2, batch 9500, loss[loss=0.2123, simple_loss=0.3055, pruned_loss=0.05958, over 16857.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3164, pruned_loss=0.07765, over 3048268.58 frames. ], batch size: 102, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:37:02,605 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0457, 3.8585, 3.8612, 3.4739, 3.9092, 1.6851, 3.7113, 3.8094], device='cuda:3'), covar=tensor([0.0061, 0.0058, 0.0068, 0.0182, 0.0052, 0.1259, 0.0063, 0.0099], device='cuda:3'), in_proj_covar=tensor([0.0059, 0.0049, 0.0073, 0.0081, 0.0056, 0.0107, 0.0066, 0.0072], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:37:46,333 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:38:34,554 INFO [train.py:904] (3/8) Epoch 2, batch 9550, loss[loss=0.2573, simple_loss=0.3384, pruned_loss=0.08815, over 16388.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3163, pruned_loss=0.07801, over 3065230.54 frames. ], batch size: 146, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:38:37,601 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1737, 3.4816, 3.3634, 1.4591, 3.4825, 3.5112, 3.0603, 2.7474], device='cuda:3'), covar=tensor([0.0936, 0.0092, 0.0156, 0.1628, 0.0117, 0.0065, 0.0322, 0.0431], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0085, 0.0082, 0.0153, 0.0075, 0.0072, 0.0109, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:39:23,643 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.908e+02 4.753e+02 5.956e+02 1.328e+03, threshold=9.507e+02, percent-clipped=2.0 2023-04-27 20:40:12,589 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 9600, loss[loss=0.2433, simple_loss=0.307, pruned_loss=0.08978, over 12133.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3191, pruned_loss=0.07971, over 3057573.11 frames. ], batch size: 247, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:40:18,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3120, 4.5640, 4.2629, 4.3646, 3.9549, 4.0483, 4.0930, 4.5334], device='cuda:3'), covar=tensor([0.0485, 0.0680, 0.0995, 0.0384, 0.0598, 0.0654, 0.0488, 0.0685], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0302, 0.0261, 0.0189, 0.0202, 0.0190, 0.0239, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:40:21,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9796, 3.5755, 3.5579, 2.5382, 3.4963, 3.6477, 3.8041, 2.0029], device='cuda:3'), covar=tensor([0.0469, 0.0025, 0.0042, 0.0257, 0.0033, 0.0038, 0.0019, 0.0387], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0054, 0.0058, 0.0105, 0.0052, 0.0056, 0.0057, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:40:55,352 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8213, 3.6473, 2.4481, 4.5092, 4.4221, 4.1041, 1.5788, 2.8140], device='cuda:3'), covar=tensor([0.1369, 0.0301, 0.1165, 0.0062, 0.0128, 0.0269, 0.1276, 0.0715], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0125, 0.0164, 0.0068, 0.0111, 0.0121, 0.0152, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 20:41:22,128 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:42:00,140 INFO [train.py:904] (3/8) Epoch 2, batch 9650, loss[loss=0.2394, simple_loss=0.3217, pruned_loss=0.07855, over 15265.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3213, pruned_loss=0.08053, over 3053073.66 frames. ], batch size: 191, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:42:54,373 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.395e+02 4.451e+02 5.149e+02 6.499e+02 1.556e+03, threshold=1.030e+03, percent-clipped=6.0 2023-04-27 20:43:09,973 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:43:48,656 INFO [train.py:904] (3/8) Epoch 2, batch 9700, loss[loss=0.2356, simple_loss=0.3137, pruned_loss=0.0787, over 16728.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.32, pruned_loss=0.08026, over 3050485.98 frames. ], batch size: 134, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:44:10,190 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4042, 1.8054, 1.4833, 1.5228, 2.1926, 2.1255, 2.5199, 2.4549], device='cuda:3'), covar=tensor([0.0019, 0.0131, 0.0142, 0.0170, 0.0066, 0.0119, 0.0026, 0.0058], device='cuda:3'), in_proj_covar=tensor([0.0050, 0.0102, 0.0101, 0.0103, 0.0091, 0.0102, 0.0055, 0.0075], device='cuda:3'), out_proj_covar=tensor([6.8464e-05, 1.5519e-04, 1.4771e-04, 1.5463e-04, 1.4053e-04, 1.5854e-04, 7.9206e-05, 1.1672e-04], device='cuda:3') 2023-04-27 20:45:18,949 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0738, 2.7674, 2.7506, 1.6564, 2.8888, 2.8305, 2.6097, 2.4576], device='cuda:3'), covar=tensor([0.0768, 0.0125, 0.0173, 0.1283, 0.0125, 0.0094, 0.0303, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0081, 0.0077, 0.0147, 0.0072, 0.0069, 0.0107, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:45:33,548 INFO [train.py:904] (3/8) Epoch 2, batch 9750, loss[loss=0.2619, simple_loss=0.3276, pruned_loss=0.09811, over 12163.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3192, pruned_loss=0.08013, over 3063304.20 frames. ], batch size: 250, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:46:02,404 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8645, 1.2916, 1.6003, 1.8272, 1.8688, 1.8638, 1.4907, 1.8098], device='cuda:3'), covar=tensor([0.0055, 0.0155, 0.0083, 0.0102, 0.0044, 0.0056, 0.0136, 0.0049], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0110, 0.0096, 0.0085, 0.0064, 0.0059, 0.0098, 0.0056], device='cuda:3'), out_proj_covar=tensor([1.2673e-04, 1.8528e-04, 1.6756e-04, 1.4880e-04, 1.0482e-04, 9.7607e-05, 1.6316e-04, 9.0543e-05], device='cuda:3') 2023-04-27 20:46:21,255 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.927e+02 4.879e+02 5.766e+02 1.284e+03, threshold=9.758e+02, percent-clipped=1.0 2023-04-27 20:47:15,366 INFO [train.py:904] (3/8) Epoch 2, batch 9800, loss[loss=0.223, simple_loss=0.3022, pruned_loss=0.07192, over 12523.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3184, pruned_loss=0.07807, over 3073910.42 frames. ], batch size: 250, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:48:00,420 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:04,161 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6479, 3.4940, 3.2389, 3.8353, 3.8235, 3.5473, 3.9217, 3.9126], device='cuda:3'), covar=tensor([0.0651, 0.0708, 0.1923, 0.0775, 0.0788, 0.1019, 0.0612, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0243, 0.0276, 0.0368, 0.0275, 0.0213, 0.0189, 0.0218, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:49:05,346 INFO [train.py:904] (3/8) Epoch 2, batch 9850, loss[loss=0.231, simple_loss=0.3005, pruned_loss=0.08074, over 12385.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.32, pruned_loss=0.07841, over 3048708.52 frames. ], batch size: 247, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:49:20,994 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8720, 4.1694, 3.9070, 3.9998, 3.5422, 3.7250, 3.8436, 4.0702], device='cuda:3'), covar=tensor([0.0440, 0.0554, 0.0632, 0.0353, 0.0498, 0.0824, 0.0425, 0.0670], device='cuda:3'), in_proj_covar=tensor([0.0210, 0.0290, 0.0253, 0.0181, 0.0195, 0.0186, 0.0230, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:49:21,399 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 20:49:24,278 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 20:49:46,147 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:49:55,601 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.468e+02 4.069e+02 4.622e+02 5.694e+02 1.429e+03, threshold=9.244e+02, percent-clipped=4.0 2023-04-27 20:50:58,119 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 2, batch 9900, loss[loss=0.2377, simple_loss=0.3252, pruned_loss=0.07511, over 15234.00 frames. ], tot_loss[loss=0.2381, simple_loss=0.3203, pruned_loss=0.07791, over 3054284.38 frames. ], batch size: 191, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:51:00,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3285, 3.3332, 1.4205, 3.3939, 2.2504, 3.3746, 1.6827, 2.6320], device='cuda:3'), covar=tensor([0.0051, 0.0194, 0.1780, 0.0044, 0.0839, 0.0313, 0.1593, 0.0645], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0112, 0.0170, 0.0073, 0.0151, 0.0135, 0.0177, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 20:52:04,514 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2550, 5.6450, 5.2734, 5.4399, 4.9137, 4.7716, 5.1061, 5.6475], device='cuda:3'), covar=tensor([0.0360, 0.0503, 0.0668, 0.0290, 0.0413, 0.0412, 0.0396, 0.0504], device='cuda:3'), in_proj_covar=tensor([0.0214, 0.0299, 0.0259, 0.0185, 0.0198, 0.0189, 0.0235, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:52:12,575 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:31,245 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:51,005 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:56,912 INFO [train.py:904] (3/8) Epoch 2, batch 9950, loss[loss=0.2536, simple_loss=0.3371, pruned_loss=0.08503, over 15395.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3219, pruned_loss=0.0784, over 3044102.40 frames. ], batch size: 191, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:53:17,536 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-04-27 20:53:23,715 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:55,569 INFO [optim.py:368] (3/8) 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 20:53:57,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2711, 2.6287, 2.2408, 3.7616, 1.8642, 3.6117, 2.2396, 2.2430], device='cuda:3'), covar=tensor([0.0322, 0.0495, 0.0399, 0.0200, 0.1428, 0.0150, 0.0755, 0.1013], device='cuda:3'), in_proj_covar=tensor([0.0220, 0.0201, 0.0164, 0.0225, 0.0271, 0.0175, 0.0198, 0.0250], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:54:55,893 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:54:56,750 INFO [train.py:904] (3/8) Epoch 2, batch 10000, loss[loss=0.2242, simple_loss=0.3133, pruned_loss=0.06759, over 16196.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3202, pruned_loss=0.07754, over 3070599.58 frames. ], batch size: 165, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:55:41,227 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:55:53,915 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1344, 3.4658, 3.5268, 2.5133, 3.3611, 3.4932, 3.6126, 1.7990], device='cuda:3'), covar=tensor([0.0403, 0.0034, 0.0030, 0.0213, 0.0038, 0.0036, 0.0017, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0057, 0.0058, 0.0108, 0.0055, 0.0057, 0.0059, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:56:37,810 INFO [train.py:904] (3/8) Epoch 2, batch 10050, loss[loss=0.235, simple_loss=0.3225, pruned_loss=0.07376, over 15482.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3196, pruned_loss=0.07722, over 3067664.16 frames. ], batch size: 192, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:56:44,640 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:57:04,521 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4653, 3.7221, 3.2684, 2.5653, 2.7052, 2.0261, 4.0127, 4.3876], device='cuda:3'), covar=tensor([0.1892, 0.0586, 0.0902, 0.0788, 0.1540, 0.1264, 0.0239, 0.0144], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0222, 0.0238, 0.0183, 0.0203, 0.0178, 0.0196, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 20:57:21,069 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8108, 3.2983, 3.4339, 2.3747, 3.2579, 3.2912, 3.4188, 1.7727], device='cuda:3'), covar=tensor([0.0500, 0.0036, 0.0034, 0.0260, 0.0033, 0.0041, 0.0031, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0056, 0.0059, 0.0108, 0.0055, 0.0057, 0.0059, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:57:24,815 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.701e+02 4.201e+02 4.991e+02 6.585e+02 1.392e+03, threshold=9.982e+02, percent-clipped=3.0 2023-04-27 20:58:10,190 INFO [train.py:904] (3/8) Epoch 2, batch 10100, loss[loss=0.2168, simple_loss=0.2907, pruned_loss=0.07144, over 12746.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3201, pruned_loss=0.07731, over 3075315.47 frames. ], batch size: 247, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:58:37,999 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:40,212 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0537, 3.1101, 3.1078, 1.4558, 3.3408, 3.3482, 2.6902, 2.5106], device='cuda:3'), covar=tensor([0.0776, 0.0124, 0.0182, 0.1258, 0.0072, 0.0052, 0.0343, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0078, 0.0074, 0.0140, 0.0071, 0.0066, 0.0102, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0003], device='cuda:3') 2023-04-27 20:58:58,256 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:07,810 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7579, 3.2449, 2.3274, 4.2512, 4.1071, 3.9864, 1.5817, 2.8570], device='cuda:3'), covar=tensor([0.1496, 0.0412, 0.1297, 0.0074, 0.0166, 0.0351, 0.1387, 0.0704], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0123, 0.0160, 0.0063, 0.0110, 0.0117, 0.0150, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 20:59:54,874 INFO [train.py:904] (3/8) Epoch 3, batch 0, loss[loss=0.3829, simple_loss=0.4113, pruned_loss=0.1773, over 16213.00 frames. ], tot_loss[loss=0.3829, simple_loss=0.4113, pruned_loss=0.1773, over 16213.00 frames. ], batch size: 164, lr: 2.28e-02, grad_scale: 8.0 2023-04-27 20:59:54,874 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 21:00:02,299 INFO [train.py:938] (3/8) Epoch 3, validation: loss=0.2012, simple_loss=0.3019, pruned_loss=0.05024, over 944034.00 frames. 2023-04-27 21:00:02,300 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 21:00:17,825 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 21:00:31,871 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:38,800 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.937e+02 4.805e+02 5.776e+02 7.144e+02 1.042e+03, threshold=1.155e+03, percent-clipped=3.0 2023-04-27 21:00:54,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9791, 4.7169, 4.8629, 5.3056, 5.3554, 4.6720, 5.4561, 5.2977], device='cuda:3'), covar=tensor([0.0505, 0.0512, 0.1027, 0.0306, 0.0336, 0.0379, 0.0223, 0.0253], device='cuda:3'), in_proj_covar=tensor([0.0254, 0.0297, 0.0390, 0.0293, 0.0221, 0.0205, 0.0227, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:01:12,895 INFO [train.py:904] (3/8) Epoch 3, batch 50, loss[loss=0.2274, simple_loss=0.3036, pruned_loss=0.07562, over 16990.00 frames. ], tot_loss[loss=0.2908, simple_loss=0.3498, pruned_loss=0.1159, over 740562.42 frames. ], batch size: 41, lr: 2.28e-02, grad_scale: 2.0 2023-04-27 21:01:45,820 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:02,740 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:16,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6084, 2.5899, 2.3940, 4.0349, 2.0369, 3.7674, 2.3934, 2.3733], device='cuda:3'), covar=tensor([0.0292, 0.0525, 0.0349, 0.0162, 0.1389, 0.0155, 0.0726, 0.0936], device='cuda:3'), in_proj_covar=tensor([0.0222, 0.0202, 0.0166, 0.0225, 0.0270, 0.0174, 0.0200, 0.0254], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:02:19,622 INFO [train.py:904] (3/8) Epoch 3, batch 100, loss[loss=0.2927, simple_loss=0.3459, pruned_loss=0.1197, over 16767.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3402, pruned_loss=0.1094, over 1318952.48 frames. ], batch size: 124, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:02:43,271 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:54,676 INFO [optim.py:368] (3/8) 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 21:03:18,759 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:23,384 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2120, 2.7809, 2.4848, 4.5509, 1.8442, 4.6098, 2.4567, 2.4098], device='cuda:3'), covar=tensor([0.0248, 0.0563, 0.0375, 0.0163, 0.1642, 0.0118, 0.0728, 0.1504], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0203, 0.0167, 0.0227, 0.0272, 0.0174, 0.0200, 0.0255], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:03:24,393 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:27,024 INFO [train.py:904] (3/8) Epoch 3, batch 150, loss[loss=0.32, simple_loss=0.3608, pruned_loss=0.1396, over 16299.00 frames. ], tot_loss[loss=0.2735, simple_loss=0.3355, pruned_loss=0.1058, over 1759715.86 frames. ], batch size: 165, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:03:37,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1079, 4.0323, 4.0778, 4.1394, 4.0773, 4.6163, 4.3037, 4.0019], device='cuda:3'), covar=tensor([0.1282, 0.1351, 0.1084, 0.1588, 0.2379, 0.0928, 0.0943, 0.2137], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0297, 0.0276, 0.0263, 0.0332, 0.0292, 0.0222, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:03:49,091 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:05,896 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 200, loss[loss=0.3022, simple_loss=0.3579, pruned_loss=0.1232, over 16481.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3337, pruned_loss=0.1034, over 2094534.47 frames. ], batch size: 146, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:04:39,872 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-27 21:05:08,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9201, 4.5808, 4.8393, 5.2258, 5.3036, 4.6435, 5.3529, 5.1478], device='cuda:3'), covar=tensor([0.0564, 0.0601, 0.1045, 0.0338, 0.0349, 0.0475, 0.0272, 0.0323], device='cuda:3'), in_proj_covar=tensor([0.0278, 0.0322, 0.0435, 0.0324, 0.0247, 0.0228, 0.0243, 0.0256], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:05:09,760 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 250, loss[loss=0.3046, simple_loss=0.3484, pruned_loss=0.1304, over 16822.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3315, pruned_loss=0.1019, over 2366612.16 frames. ], batch size: 102, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:58,361 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:05,594 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9395, 4.7228, 4.6920, 4.1603, 4.7529, 2.0721, 4.4562, 4.8265], device='cuda:3'), covar=tensor([0.0048, 0.0059, 0.0060, 0.0239, 0.0045, 0.1220, 0.0055, 0.0077], device='cuda:3'), in_proj_covar=tensor([0.0067, 0.0055, 0.0086, 0.0095, 0.0062, 0.0116, 0.0075, 0.0083], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:06:26,541 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 21:06:35,583 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:54,934 INFO [train.py:904] (3/8) Epoch 3, batch 300, loss[loss=0.2333, simple_loss=0.3076, pruned_loss=0.07947, over 16718.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3268, pruned_loss=0.09865, over 2583030.15 frames. ], batch size: 62, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:07:28,549 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-27 21:07:29,012 INFO [optim.py:368] (3/8) 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 21:07:40,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9013, 3.1556, 3.6179, 2.7059, 3.4598, 3.5267, 3.5862, 1.8851], device='cuda:3'), covar=tensor([0.0425, 0.0064, 0.0037, 0.0216, 0.0056, 0.0073, 0.0042, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0056, 0.0060, 0.0108, 0.0055, 0.0061, 0.0060, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:07:59,135 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 350, loss[loss=0.2545, simple_loss=0.334, pruned_loss=0.08751, over 17118.00 frames. ], tot_loss[loss=0.2587, simple_loss=0.3241, pruned_loss=0.09663, over 2743739.06 frames. ], batch size: 53, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:08:04,822 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-04-27 21:08:36,896 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:06,941 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5284, 2.4197, 2.3777, 2.6154, 3.0566, 2.8691, 3.6661, 3.3805], device='cuda:3'), covar=tensor([0.0021, 0.0135, 0.0126, 0.0122, 0.0078, 0.0122, 0.0043, 0.0057], device='cuda:3'), in_proj_covar=tensor([0.0055, 0.0109, 0.0106, 0.0108, 0.0100, 0.0108, 0.0062, 0.0080], device='cuda:3'), out_proj_covar=tensor([7.5825e-05, 1.6484e-04, 1.5433e-04, 1.6095e-04, 1.5315e-04, 1.6520e-04, 9.2223e-05, 1.2435e-04], device='cuda:3') 2023-04-27 21:09:09,380 INFO [train.py:904] (3/8) Epoch 3, batch 400, loss[loss=0.2413, simple_loss=0.3068, pruned_loss=0.08791, over 16799.00 frames. ], tot_loss[loss=0.2584, simple_loss=0.323, pruned_loss=0.09693, over 2860807.74 frames. ], batch size: 39, lr: 2.26e-02, grad_scale: 4.0 2023-04-27 21:09:41,230 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.989e+02 4.408e+02 5.242e+02 6.457e+02 1.269e+03, threshold=1.048e+03, percent-clipped=5.0 2023-04-27 21:10:08,712 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:10,765 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:17,851 INFO [train.py:904] (3/8) Epoch 3, batch 450, loss[loss=0.2374, simple_loss=0.3269, pruned_loss=0.07397, over 17233.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3214, pruned_loss=0.09596, over 2964363.65 frames. ], batch size: 52, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:10:41,831 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:49,449 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:14,831 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 500, loss[loss=0.2281, simple_loss=0.3091, pruned_loss=0.0736, over 16726.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3197, pruned_loss=0.0943, over 3047122.58 frames. ], batch size: 57, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:11:46,146 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:52,935 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6243, 4.5599, 4.2649, 2.0808, 3.1052, 2.3808, 3.8457, 4.5021], device='cuda:3'), covar=tensor([0.0253, 0.0531, 0.0350, 0.1623, 0.0727, 0.1015, 0.0798, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0108, 0.0151, 0.0145, 0.0136, 0.0129, 0.0141, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 21:12:00,585 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 4.223e+02 4.982e+02 5.898e+02 1.253e+03, threshold=9.964e+02, percent-clipped=1.0 2023-04-27 21:12:34,379 INFO [train.py:904] (3/8) Epoch 3, batch 550, loss[loss=0.2904, simple_loss=0.3345, pruned_loss=0.1232, over 16344.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3174, pruned_loss=0.09261, over 3106297.69 frames. ], batch size: 165, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:12:45,302 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:06,327 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0528, 3.9942, 3.9150, 3.4478, 3.9656, 1.8038, 3.6796, 3.9316], device='cuda:3'), covar=tensor([0.0080, 0.0058, 0.0084, 0.0270, 0.0062, 0.1234, 0.0082, 0.0106], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0058, 0.0091, 0.0103, 0.0066, 0.0116, 0.0079, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:13:40,833 INFO [train.py:904] (3/8) Epoch 3, batch 600, loss[loss=0.2243, simple_loss=0.292, pruned_loss=0.07831, over 16975.00 frames. ], tot_loss[loss=0.2499, simple_loss=0.3159, pruned_loss=0.09199, over 3162636.66 frames. ], batch size: 41, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:13:50,525 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:55,155 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-04-27 21:14:15,688 INFO [optim.py:368] (3/8) 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 21:14:38,951 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:48,033 INFO [train.py:904] (3/8) Epoch 3, batch 650, loss[loss=0.2388, simple_loss=0.3162, pruned_loss=0.08071, over 17110.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3135, pruned_loss=0.09009, over 3205332.54 frames. ], batch size: 47, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:15:57,313 INFO [train.py:904] (3/8) Epoch 3, batch 700, loss[loss=0.2061, simple_loss=0.2812, pruned_loss=0.06556, over 16813.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3131, pruned_loss=0.08946, over 3228028.91 frames. ], batch size: 42, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:16:30,865 INFO [optim.py:368] (3/8) 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 21:16:37,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0174, 4.2428, 3.4489, 2.8518, 3.1929, 2.3751, 4.6874, 4.8711], device='cuda:3'), covar=tensor([0.1703, 0.0491, 0.0930, 0.0779, 0.1916, 0.1191, 0.0215, 0.0174], device='cuda:3'), in_proj_covar=tensor([0.0260, 0.0236, 0.0250, 0.0194, 0.0263, 0.0187, 0.0208, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:16:47,038 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4452, 2.3728, 2.1003, 2.1566, 2.9481, 2.7837, 3.7920, 3.2840], device='cuda:3'), covar=tensor([0.0019, 0.0141, 0.0171, 0.0180, 0.0084, 0.0129, 0.0037, 0.0074], device='cuda:3'), in_proj_covar=tensor([0.0058, 0.0110, 0.0110, 0.0111, 0.0102, 0.0112, 0.0065, 0.0084], device='cuda:3'), out_proj_covar=tensor([8.0597e-05, 1.6617e-04, 1.5869e-04, 1.6479e-04, 1.5656e-04, 1.7103e-04, 9.7367e-05, 1.3029e-04], device='cuda:3') 2023-04-27 21:16:55,516 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 750, loss[loss=0.2705, simple_loss=0.3246, pruned_loss=0.1082, over 16659.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3141, pruned_loss=0.08954, over 3259001.74 frames. ], batch size: 89, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:17:06,002 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:35,082 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:38,565 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3739, 4.5435, 5.0142, 5.0728, 5.0496, 4.5826, 4.5837, 4.6188], device='cuda:3'), covar=tensor([0.0258, 0.0228, 0.0331, 0.0328, 0.0313, 0.0250, 0.0688, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0168, 0.0194, 0.0183, 0.0221, 0.0193, 0.0284, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 21:17:39,866 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9400, 4.6223, 4.8780, 5.1717, 5.3139, 4.4783, 5.3402, 5.1494], device='cuda:3'), covar=tensor([0.0564, 0.0631, 0.1017, 0.0347, 0.0347, 0.0554, 0.0259, 0.0315], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0338, 0.0458, 0.0346, 0.0267, 0.0242, 0.0259, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:17:57,838 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 800, loss[loss=0.3063, simple_loss=0.3484, pruned_loss=0.1322, over 16762.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3149, pruned_loss=0.09036, over 3272262.96 frames. ], batch size: 124, lr: 2.24e-02, grad_scale: 8.0 2023-04-27 21:18:27,353 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:34,875 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:40,095 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:19:10,613 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3313, 4.4256, 4.0042, 1.8955, 3.0261, 2.2725, 3.7250, 4.2558], device='cuda:3'), covar=tensor([0.0350, 0.0509, 0.0414, 0.1751, 0.0762, 0.1213, 0.0806, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0114, 0.0154, 0.0148, 0.0138, 0.0132, 0.0147, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 21:19:20,165 INFO [train.py:904] (3/8) Epoch 3, batch 850, loss[loss=0.2091, simple_loss=0.2887, pruned_loss=0.06476, over 17204.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3124, pruned_loss=0.08844, over 3286189.47 frames. ], batch size: 44, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:19:58,138 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:20:03,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9807, 1.5734, 1.5459, 1.4592, 1.9168, 1.6890, 1.8431, 1.8844], device='cuda:3'), covar=tensor([0.0035, 0.0123, 0.0126, 0.0127, 0.0074, 0.0122, 0.0057, 0.0068], device='cuda:3'), in_proj_covar=tensor([0.0058, 0.0110, 0.0111, 0.0111, 0.0102, 0.0111, 0.0066, 0.0083], device='cuda:3'), out_proj_covar=tensor([8.1336e-05, 1.6535e-04, 1.6020e-04, 1.6311e-04, 1.5625e-04, 1.6932e-04, 9.7912e-05, 1.2998e-04], device='cuda:3') 2023-04-27 21:20:27,600 INFO [train.py:904] (3/8) Epoch 3, batch 900, loss[loss=0.2413, simple_loss=0.3205, pruned_loss=0.08101, over 17109.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3111, pruned_loss=0.08778, over 3283110.83 frames. ], batch size: 49, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:20:42,486 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-27 21:21:03,014 INFO [optim.py:368] (3/8) 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 21:21:03,408 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9826, 3.9190, 4.4646, 4.4971, 4.4972, 4.0358, 4.1811, 4.0615], device='cuda:3'), covar=tensor([0.0249, 0.0300, 0.0351, 0.0336, 0.0322, 0.0279, 0.0672, 0.0357], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0170, 0.0198, 0.0186, 0.0224, 0.0197, 0.0286, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 21:21:16,137 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9181, 4.0247, 2.8422, 5.0765, 4.9909, 4.4849, 2.1655, 3.4234], device='cuda:3'), covar=tensor([0.1397, 0.0319, 0.1105, 0.0079, 0.0220, 0.0271, 0.1088, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0128, 0.0165, 0.0073, 0.0137, 0.0133, 0.0153, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 21:21:24,795 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4470, 4.3466, 4.3253, 3.7453, 4.3241, 2.0403, 4.0731, 4.3448], device='cuda:3'), covar=tensor([0.0078, 0.0064, 0.0082, 0.0294, 0.0053, 0.1151, 0.0074, 0.0112], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0062, 0.0096, 0.0112, 0.0070, 0.0120, 0.0084, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:21:26,694 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:35,849 INFO [train.py:904] (3/8) Epoch 3, batch 950, loss[loss=0.2604, simple_loss=0.3208, pruned_loss=0.09996, over 16821.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3107, pruned_loss=0.08794, over 3282952.55 frames. ], batch size: 102, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:21:56,751 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 21:22:34,057 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 1000, loss[loss=0.2806, simple_loss=0.3307, pruned_loss=0.1152, over 12012.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.31, pruned_loss=0.08835, over 3286526.71 frames. ], batch size: 248, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:23:01,304 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:18,067 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2310, 5.1613, 4.9603, 5.0955, 4.4226, 5.0542, 5.1404, 4.7376], device='cuda:3'), covar=tensor([0.0370, 0.0163, 0.0182, 0.0134, 0.0829, 0.0190, 0.0173, 0.0277], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0125, 0.0188, 0.0153, 0.0222, 0.0166, 0.0133, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:23:21,765 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.770e+02 3.937e+02 4.742e+02 6.159e+02 9.838e+02, threshold=9.485e+02, percent-clipped=2.0 2023-04-27 21:23:23,973 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:55,374 INFO [train.py:904] (3/8) Epoch 3, batch 1050, loss[loss=0.2433, simple_loss=0.3146, pruned_loss=0.08596, over 16598.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3083, pruned_loss=0.08689, over 3299305.38 frames. ], batch size: 62, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:24:25,342 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:46,812 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:02,813 INFO [train.py:904] (3/8) Epoch 3, batch 1100, loss[loss=0.2605, simple_loss=0.3076, pruned_loss=0.1067, over 16746.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3075, pruned_loss=0.08608, over 3312391.40 frames. ], batch size: 124, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:25:08,106 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.90 vs. limit=5.0 2023-04-27 21:25:12,306 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:38,414 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.639e+02 4.233e+02 5.080e+02 6.558e+02 1.311e+03, threshold=1.016e+03, percent-clipped=3.0 2023-04-27 21:26:10,402 INFO [train.py:904] (3/8) Epoch 3, batch 1150, loss[loss=0.2037, simple_loss=0.2906, pruned_loss=0.05846, over 17102.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3078, pruned_loss=0.08576, over 3315476.87 frames. ], batch size: 49, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:26:42,688 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:27:19,790 INFO [train.py:904] (3/8) Epoch 3, batch 1200, loss[loss=0.2833, simple_loss=0.3338, pruned_loss=0.1164, over 16919.00 frames. ], tot_loss[loss=0.239, simple_loss=0.307, pruned_loss=0.08549, over 3321108.21 frames. ], batch size: 116, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:27:36,527 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 21:27:56,783 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.899e+02 4.622e+02 5.754e+02 9.760e+02, threshold=9.243e+02, percent-clipped=0.0 2023-04-27 21:28:27,750 INFO [train.py:904] (3/8) Epoch 3, batch 1250, loss[loss=0.2167, simple_loss=0.2867, pruned_loss=0.07332, over 16767.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3069, pruned_loss=0.08567, over 3316948.89 frames. ], batch size: 39, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:30,701 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 1300, loss[loss=0.2542, simple_loss=0.3067, pruned_loss=0.1008, over 15348.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3069, pruned_loss=0.0848, over 3310838.48 frames. ], batch size: 191, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:56,068 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:17,379 INFO [optim.py:368] (3/8) 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 21:30:49,953 INFO [train.py:904] (3/8) Epoch 3, batch 1350, loss[loss=0.2302, simple_loss=0.298, pruned_loss=0.08123, over 15594.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3077, pruned_loss=0.08542, over 3316443.29 frames. ], batch size: 191, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:30:56,229 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:13,017 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:20,909 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:58,965 INFO [train.py:904] (3/8) Epoch 3, batch 1400, loss[loss=0.2353, simple_loss=0.3212, pruned_loss=0.07468, over 17143.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3076, pruned_loss=0.08502, over 3324470.09 frames. ], batch size: 48, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:32:09,165 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:33:07,161 INFO [train.py:904] (3/8) Epoch 3, batch 1450, loss[loss=0.2353, simple_loss=0.2982, pruned_loss=0.08614, over 16776.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3064, pruned_loss=0.08418, over 3328994.78 frames. ], batch size: 83, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:33:11,877 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8829, 3.2523, 2.8440, 4.3049, 4.1794, 4.2017, 1.7368, 3.5482], device='cuda:3'), covar=tensor([0.1198, 0.0372, 0.0865, 0.0069, 0.0164, 0.0161, 0.1101, 0.0401], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0126, 0.0161, 0.0073, 0.0139, 0.0130, 0.0149, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 21:33:13,342 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-27 21:33:13,878 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:24,020 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:39,372 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:12,712 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-27 21:34:14,645 INFO [train.py:904] (3/8) Epoch 3, batch 1500, loss[loss=0.2572, simple_loss=0.3189, pruned_loss=0.09775, over 16792.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3062, pruned_loss=0.08482, over 3321477.03 frames. ], batch size: 83, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:34:43,974 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:47,229 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:35:22,094 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 21:35:23,687 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:35:24,397 INFO [train.py:904] (3/8) Epoch 3, batch 1550, loss[loss=0.1962, simple_loss=0.2683, pruned_loss=0.06202, over 15835.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3084, pruned_loss=0.08674, over 3314552.39 frames. ], batch size: 35, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:31,485 INFO [train.py:904] (3/8) Epoch 3, batch 1600, loss[loss=0.2137, simple_loss=0.2931, pruned_loss=0.06715, over 15848.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3095, pruned_loss=0.08769, over 3317375.83 frames. ], batch size: 35, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:32,444 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-27 21:36:46,706 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:07,922 INFO [optim.py:368] (3/8) 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 21:37:23,392 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8448, 3.8712, 1.8576, 3.9081, 2.6598, 3.9335, 1.9684, 2.8636], device='cuda:3'), covar=tensor([0.0058, 0.0229, 0.1347, 0.0054, 0.0661, 0.0285, 0.1152, 0.0556], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0136, 0.0170, 0.0084, 0.0156, 0.0164, 0.0177, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 21:37:38,901 INFO [train.py:904] (3/8) Epoch 3, batch 1650, loss[loss=0.2331, simple_loss=0.3054, pruned_loss=0.08047, over 15869.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3108, pruned_loss=0.08835, over 3327664.70 frames. ], batch size: 35, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:37:39,890 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:01,236 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:03,305 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:03,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1126, 3.7950, 2.8502, 4.7634, 4.6711, 4.5591, 2.0027, 3.2732], device='cuda:3'), covar=tensor([0.1187, 0.0297, 0.0891, 0.0049, 0.0154, 0.0169, 0.1014, 0.0572], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0161, 0.0073, 0.0141, 0.0134, 0.0153, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 21:38:04,703 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:24,292 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:50,891 INFO [train.py:904] (3/8) Epoch 3, batch 1700, loss[loss=0.2489, simple_loss=0.3105, pruned_loss=0.09364, over 16806.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.313, pruned_loss=0.08874, over 3327728.26 frames. ], batch size: 102, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:38:56,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5334, 4.8682, 4.4936, 4.6743, 4.1601, 4.2522, 4.4106, 4.8825], device='cuda:3'), covar=tensor([0.0489, 0.0527, 0.0838, 0.0345, 0.0603, 0.0671, 0.0474, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0381, 0.0330, 0.0232, 0.0247, 0.0224, 0.0298, 0.0254], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:39:11,658 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:23,758 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3700, 1.7631, 2.2194, 2.9485, 3.0161, 3.6949, 1.6314, 3.3639], device='cuda:3'), covar=tensor([0.0040, 0.0167, 0.0112, 0.0085, 0.0043, 0.0032, 0.0182, 0.0036], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0120, 0.0106, 0.0101, 0.0080, 0.0068, 0.0107, 0.0065], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 21:39:26,757 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-27 21:39:28,009 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.555e+02 4.242e+02 5.121e+02 6.013e+02 1.262e+03, threshold=1.024e+03, percent-clipped=2.0 2023-04-27 21:39:30,885 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:33,270 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:57,284 INFO [train.py:904] (3/8) Epoch 3, batch 1750, loss[loss=0.2199, simple_loss=0.297, pruned_loss=0.07141, over 17191.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3139, pruned_loss=0.08839, over 3332419.32 frames. ], batch size: 45, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:41:06,920 INFO [train.py:904] (3/8) Epoch 3, batch 1800, loss[loss=0.2607, simple_loss=0.3259, pruned_loss=0.0977, over 16789.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3154, pruned_loss=0.08866, over 3331528.04 frames. ], batch size: 102, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:41:30,520 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:41:45,811 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:54,177 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8939, 4.6517, 4.0765, 2.0540, 3.2048, 2.6525, 4.1413, 4.8972], device='cuda:3'), covar=tensor([0.0206, 0.0644, 0.0489, 0.1663, 0.0791, 0.0983, 0.0576, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0123, 0.0155, 0.0146, 0.0137, 0.0131, 0.0146, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 21:42:05,674 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3334, 2.0578, 1.6252, 1.9399, 2.8077, 2.6126, 3.7297, 3.1589], device='cuda:3'), covar=tensor([0.0019, 0.0170, 0.0218, 0.0182, 0.0085, 0.0134, 0.0040, 0.0082], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0117, 0.0114, 0.0118, 0.0110, 0.0118, 0.0074, 0.0093], device='cuda:3'), out_proj_covar=tensor([8.6502e-05, 1.7351e-04, 1.6266e-04, 1.7365e-04, 1.6737e-04, 1.7761e-04, 1.0986e-04, 1.4390e-04], device='cuda:3') 2023-04-27 21:42:14,095 INFO [train.py:904] (3/8) Epoch 3, batch 1850, loss[loss=0.258, simple_loss=0.3127, pruned_loss=0.1017, over 16872.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3168, pruned_loss=0.08902, over 3321626.53 frames. ], batch size: 109, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:42:15,592 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5572, 4.3371, 4.5329, 4.8521, 4.8826, 4.3834, 4.8249, 4.8728], device='cuda:3'), covar=tensor([0.0496, 0.0534, 0.0880, 0.0339, 0.0305, 0.0425, 0.0451, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0364, 0.0497, 0.0369, 0.0280, 0.0259, 0.0280, 0.0297], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:42:47,473 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:09,390 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:22,338 INFO [train.py:904] (3/8) Epoch 3, batch 1900, loss[loss=0.2532, simple_loss=0.3114, pruned_loss=0.09754, over 16722.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3151, pruned_loss=0.08789, over 3321786.59 frames. ], batch size: 134, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:43:31,014 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:42,570 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:02,291 INFO [optim.py:368] (3/8) 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 21:44:12,146 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:12,241 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:32,422 INFO [train.py:904] (3/8) Epoch 3, batch 1950, loss[loss=0.2104, simple_loss=0.2865, pruned_loss=0.06721, over 16812.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3147, pruned_loss=0.08721, over 3304032.94 frames. ], batch size: 42, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:44:32,726 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:58,063 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:06,106 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:35,334 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:36,251 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2000, loss[loss=0.2232, simple_loss=0.3139, pruned_loss=0.06626, over 17116.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3146, pruned_loss=0.08754, over 3307916.48 frames. ], batch size: 49, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:45:45,119 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 21:46:00,548 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 21:46:01,055 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:13,065 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:17,541 INFO [optim.py:368] (3/8) 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 21:46:20,970 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2293, 2.6363, 2.3917, 4.6443, 1.8212, 4.5369, 2.3454, 2.5044], device='cuda:3'), covar=tensor([0.0276, 0.0708, 0.0463, 0.0161, 0.1931, 0.0165, 0.0835, 0.1558], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0216, 0.0183, 0.0242, 0.0284, 0.0187, 0.0205, 0.0277], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:46:43,088 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1745, 3.4461, 3.3635, 1.5939, 3.5533, 3.5160, 2.9872, 2.7071], device='cuda:3'), covar=tensor([0.0838, 0.0111, 0.0182, 0.1238, 0.0083, 0.0070, 0.0336, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0082, 0.0083, 0.0146, 0.0078, 0.0076, 0.0114, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 21:46:46,961 INFO [train.py:904] (3/8) Epoch 3, batch 2050, loss[loss=0.2472, simple_loss=0.3121, pruned_loss=0.09116, over 15935.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3138, pruned_loss=0.08724, over 3304463.67 frames. ], batch size: 35, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:46:50,739 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9902, 2.8729, 3.3837, 2.3766, 3.2725, 3.3830, 3.2947, 2.0012], device='cuda:3'), covar=tensor([0.0350, 0.0104, 0.0047, 0.0230, 0.0040, 0.0045, 0.0036, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0057, 0.0061, 0.0110, 0.0055, 0.0062, 0.0063, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:47:11,159 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 21:47:33,617 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:47:55,916 INFO [train.py:904] (3/8) Epoch 3, batch 2100, loss[loss=0.2246, simple_loss=0.3061, pruned_loss=0.07156, over 17214.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3141, pruned_loss=0.08703, over 3307243.12 frames. ], batch size: 44, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:48:20,691 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:48:55,063 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2150, loss[loss=0.2171, simple_loss=0.289, pruned_loss=0.07264, over 16985.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3144, pruned_loss=0.08818, over 3308762.09 frames. ], batch size: 41, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:49:22,487 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:47,730 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2200, loss[loss=0.2589, simple_loss=0.3151, pruned_loss=0.1013, over 16869.00 frames. ], tot_loss[loss=0.247, simple_loss=0.316, pruned_loss=0.08896, over 3305114.61 frames. ], batch size: 96, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:50:14,236 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:18,182 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-27 21:50:31,389 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 21:50:46,033 INFO [optim.py:368] (3/8) 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 21:50:48,907 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2250, loss[loss=0.2486, simple_loss=0.3083, pruned_loss=0.09443, over 16419.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3174, pruned_loss=0.09028, over 3297357.67 frames. ], batch size: 146, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:51:18,566 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:42,425 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:55,747 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 21:52:09,941 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:20,009 INFO [train.py:904] (3/8) Epoch 3, batch 2300, loss[loss=0.248, simple_loss=0.309, pruned_loss=0.09344, over 16819.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3181, pruned_loss=0.09001, over 3308588.93 frames. ], batch size: 116, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:52:57,419 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:01,629 INFO [optim.py:368] (3/8) 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 21:53:05,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5831, 4.8865, 4.2477, 4.7587, 4.2652, 4.3029, 4.4798, 4.8491], device='cuda:3'), covar=tensor([0.1217, 0.1133, 0.2256, 0.0707, 0.1067, 0.1122, 0.0963, 0.1209], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0384, 0.0337, 0.0234, 0.0250, 0.0229, 0.0299, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:53:11,989 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:28,522 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2877, 2.1399, 1.9894, 1.9622, 2.5865, 2.5972, 3.6948, 3.0703], device='cuda:3'), covar=tensor([0.0025, 0.0157, 0.0154, 0.0169, 0.0089, 0.0131, 0.0038, 0.0083], device='cuda:3'), in_proj_covar=tensor([0.0061, 0.0118, 0.0114, 0.0117, 0.0108, 0.0116, 0.0076, 0.0094], device='cuda:3'), out_proj_covar=tensor([8.7957e-05, 1.7533e-04, 1.6329e-04, 1.7138e-04, 1.6396e-04, 1.7367e-04, 1.1210e-04, 1.4452e-04], device='cuda:3') 2023-04-27 21:53:30,352 INFO [train.py:904] (3/8) Epoch 3, batch 2350, loss[loss=0.285, simple_loss=0.3323, pruned_loss=0.1188, over 16696.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3178, pruned_loss=0.08968, over 3314475.60 frames. ], batch size: 89, lr: 2.16e-02, grad_scale: 4.0 2023-04-27 21:54:01,189 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:54:06,357 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-27 21:54:34,989 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 21:54:35,720 INFO [train.py:904] (3/8) Epoch 3, batch 2400, loss[loss=0.3033, simple_loss=0.3555, pruned_loss=0.1255, over 16766.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3183, pruned_loss=0.08986, over 3311575.94 frames. ], batch size: 83, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:55:09,611 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.46 vs. limit=5.0 2023-04-27 21:55:17,391 INFO [optim.py:368] (3/8) 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 21:55:32,304 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2450, loss[loss=0.2389, simple_loss=0.323, pruned_loss=0.07735, over 16690.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3201, pruned_loss=0.0901, over 3305618.07 frames. ], batch size: 62, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:56:32,287 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6533, 2.5785, 2.4272, 4.1152, 1.8729, 3.5880, 2.1853, 2.3098], device='cuda:3'), covar=tensor([0.0297, 0.0614, 0.0429, 0.0168, 0.1635, 0.0230, 0.0898, 0.1131], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0219, 0.0185, 0.0247, 0.0288, 0.0192, 0.0209, 0.0282], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:56:33,240 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:37,032 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2064, 4.9722, 5.0157, 4.3844, 4.8737, 1.8374, 4.7356, 5.2040], device='cuda:3'), covar=tensor([0.0065, 0.0063, 0.0067, 0.0320, 0.0061, 0.1281, 0.0076, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0066, 0.0103, 0.0120, 0.0076, 0.0118, 0.0092, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 21:56:46,137 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2500, loss[loss=0.3254, simple_loss=0.373, pruned_loss=0.1389, over 12094.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3201, pruned_loss=0.09052, over 3299339.32 frames. ], batch size: 246, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:57:27,643 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.850e+02 4.203e+02 4.842e+02 6.402e+02 1.699e+03, threshold=9.683e+02, percent-clipped=7.0 2023-04-27 21:57:36,903 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:45,785 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7039, 2.5957, 2.4995, 4.2214, 1.9820, 3.7606, 2.2804, 2.3170], device='cuda:3'), covar=tensor([0.0317, 0.0612, 0.0382, 0.0160, 0.1551, 0.0211, 0.0841, 0.1119], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0218, 0.0184, 0.0246, 0.0287, 0.0192, 0.0209, 0.0281], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 21:58:03,439 INFO [train.py:904] (3/8) Epoch 3, batch 2550, loss[loss=0.2727, simple_loss=0.3328, pruned_loss=0.1063, over 16712.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.321, pruned_loss=0.09102, over 3285175.92 frames. ], batch size: 134, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:58:10,931 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:31,382 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:44,551 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:52,551 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:01,511 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:13,834 INFO [train.py:904] (3/8) Epoch 3, batch 2600, loss[loss=0.2418, simple_loss=0.3225, pruned_loss=0.0806, over 17135.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3189, pruned_loss=0.08888, over 3305819.61 frames. ], batch size: 49, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:59:24,865 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-27 21:59:38,824 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:49,852 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-27 21:59:53,532 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.483e+02 3.680e+02 4.418e+02 5.171e+02 1.031e+03, threshold=8.837e+02, percent-clipped=1.0 2023-04-27 22:00:09,152 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2650, loss[loss=0.2481, simple_loss=0.3358, pruned_loss=0.08015, over 17063.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3201, pruned_loss=0.08942, over 3312949.35 frames. ], batch size: 50, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:01:22,737 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:01:30,281 INFO [train.py:904] (3/8) Epoch 3, batch 2700, loss[loss=0.2155, simple_loss=0.3022, pruned_loss=0.06437, over 17213.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3208, pruned_loss=0.08891, over 3320286.93 frames. ], batch size: 45, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:01:46,838 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1523, 4.7885, 4.5398, 5.3370, 5.3294, 4.7990, 5.2893, 5.2781], device='cuda:3'), covar=tensor([0.0498, 0.0587, 0.1806, 0.0424, 0.0572, 0.0343, 0.0516, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0373, 0.0496, 0.0381, 0.0282, 0.0266, 0.0289, 0.0305], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:02:09,771 INFO [optim.py:368] (3/8) 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 22:02:14,216 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9004, 4.7883, 4.6471, 4.0562, 4.6686, 1.9394, 4.4426, 4.8257], device='cuda:3'), covar=tensor([0.0055, 0.0053, 0.0073, 0.0308, 0.0057, 0.1272, 0.0078, 0.0089], device='cuda:3'), in_proj_covar=tensor([0.0077, 0.0064, 0.0102, 0.0117, 0.0075, 0.0115, 0.0091, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:02:23,971 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2750, loss[loss=0.2708, simple_loss=0.3322, pruned_loss=0.1047, over 16735.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3201, pruned_loss=0.0876, over 3316740.11 frames. ], batch size: 124, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:02:39,130 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9226, 3.5268, 3.2671, 4.8480, 4.7642, 4.3364, 1.4633, 3.7395], device='cuda:3'), covar=tensor([0.1310, 0.0370, 0.0771, 0.0062, 0.0152, 0.0275, 0.1329, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0130, 0.0163, 0.0077, 0.0144, 0.0138, 0.0154, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 22:02:51,745 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9475, 2.6225, 2.5026, 4.5791, 1.7927, 4.2340, 2.3987, 2.5354], device='cuda:3'), covar=tensor([0.0316, 0.0738, 0.0463, 0.0149, 0.1928, 0.0222, 0.0854, 0.1409], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0222, 0.0187, 0.0251, 0.0292, 0.0197, 0.0213, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:03:16,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8619, 2.7045, 2.4862, 4.4654, 1.8383, 3.9357, 2.4621, 2.2810], device='cuda:3'), covar=tensor([0.0331, 0.0711, 0.0450, 0.0157, 0.1853, 0.0268, 0.0840, 0.1477], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0222, 0.0187, 0.0253, 0.0293, 0.0198, 0.0213, 0.0286], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:03:22,893 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 22:03:28,906 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:45,515 INFO [train.py:904] (3/8) Epoch 3, batch 2800, loss[loss=0.2466, simple_loss=0.3338, pruned_loss=0.07974, over 16724.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3209, pruned_loss=0.08843, over 3305270.49 frames. ], batch size: 57, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:03:55,520 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:25,541 INFO [optim.py:368] (3/8) 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 22:04:40,579 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2520, 1.9100, 1.5870, 1.8560, 2.3717, 2.3065, 2.5487, 2.4565], device='cuda:3'), covar=tensor([0.0026, 0.0115, 0.0136, 0.0135, 0.0060, 0.0099, 0.0058, 0.0066], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0118, 0.0115, 0.0117, 0.0110, 0.0117, 0.0077, 0.0095], device='cuda:3'), out_proj_covar=tensor([8.6580e-05, 1.7275e-04, 1.6273e-04, 1.7039e-04, 1.6548e-04, 1.7542e-04, 1.1522e-04, 1.4649e-04], device='cuda:3') 2023-04-27 22:04:54,743 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 2850, loss[loss=0.2768, simple_loss=0.3272, pruned_loss=0.1132, over 16219.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3204, pruned_loss=0.08836, over 3310084.38 frames. ], batch size: 165, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:05:08,344 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 22:05:20,384 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:36,688 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:03,312 INFO [train.py:904] (3/8) Epoch 3, batch 2900, loss[loss=0.2257, simple_loss=0.2951, pruned_loss=0.07814, over 15892.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3181, pruned_loss=0.08767, over 3312649.58 frames. ], batch size: 35, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:06:25,879 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:31,898 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1416, 3.9395, 2.0763, 4.0235, 2.6646, 4.0947, 2.0294, 2.9175], device='cuda:3'), covar=tensor([0.0039, 0.0203, 0.1223, 0.0055, 0.0626, 0.0362, 0.1233, 0.0601], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0138, 0.0171, 0.0086, 0.0162, 0.0170, 0.0182, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 22:06:42,401 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:43,093 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 2950, loss[loss=0.2737, simple_loss=0.3272, pruned_loss=0.1101, over 16777.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3178, pruned_loss=0.08868, over 3318040.51 frames. ], batch size: 124, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:07:49,173 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:07,100 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:12,632 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3000, loss[loss=0.2546, simple_loss=0.3231, pruned_loss=0.09305, over 16555.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3183, pruned_loss=0.08932, over 3318272.81 frames. ], batch size: 75, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:08:19,857 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 22:08:30,494 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 22:09:10,134 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:37,539 INFO [train.py:904] (3/8) Epoch 3, batch 3050, loss[loss=0.2969, simple_loss=0.3509, pruned_loss=0.1214, over 16719.00 frames. ], tot_loss[loss=0.249, simple_loss=0.3188, pruned_loss=0.08964, over 3322145.49 frames. ], batch size: 134, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:10:18,576 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7234, 3.3851, 2.6483, 3.6854, 3.6064, 3.5881, 3.5912, 3.6592], device='cuda:3'), covar=tensor([0.0701, 0.0942, 0.3077, 0.1095, 0.1266, 0.1306, 0.1257, 0.1260], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0377, 0.0496, 0.0379, 0.0282, 0.0269, 0.0300, 0.0306], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:10:44,544 INFO [train.py:904] (3/8) Epoch 3, batch 3100, loss[loss=0.2443, simple_loss=0.2928, pruned_loss=0.09795, over 16726.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3174, pruned_loss=0.08863, over 3326648.24 frames. ], batch size: 134, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:11:00,304 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9632, 4.9266, 4.7672, 4.1411, 4.7374, 2.0966, 4.4924, 4.8272], device='cuda:3'), covar=tensor([0.0056, 0.0050, 0.0066, 0.0317, 0.0054, 0.1132, 0.0077, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0067, 0.0103, 0.0121, 0.0075, 0.0115, 0.0093, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:11:28,255 INFO [optim.py:368] (3/8) 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 22:11:53,221 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3150, loss[loss=0.1929, simple_loss=0.2759, pruned_loss=0.05496, over 16833.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3156, pruned_loss=0.08756, over 3319202.74 frames. ], batch size: 42, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:11:59,859 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7005, 4.6790, 4.3369, 1.8952, 3.3215, 2.6063, 3.9442, 4.2062], device='cuda:3'), covar=tensor([0.0247, 0.0441, 0.0312, 0.1564, 0.0598, 0.0927, 0.0641, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0123, 0.0154, 0.0144, 0.0135, 0.0130, 0.0148, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 22:12:12,235 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:34,999 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:44,061 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:58,217 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3200, loss[loss=0.2289, simple_loss=0.31, pruned_loss=0.07396, over 17051.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3139, pruned_loss=0.08642, over 3321720.87 frames. ], batch size: 50, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:13:39,249 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:42,227 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.397e+02 3.601e+02 4.370e+02 5.311e+02 9.274e+02, threshold=8.739e+02, percent-clipped=1.0 2023-04-27 22:13:46,605 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9077, 3.9329, 4.4082, 4.3822, 4.4161, 3.9655, 4.0471, 3.9454], device='cuda:3'), covar=tensor([0.0262, 0.0355, 0.0330, 0.0384, 0.0382, 0.0328, 0.0750, 0.0480], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0192, 0.0210, 0.0210, 0.0252, 0.0210, 0.0318, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 22:13:56,172 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9682, 3.7620, 3.9539, 4.2716, 4.2511, 3.8526, 4.0528, 4.2540], device='cuda:3'), covar=tensor([0.0660, 0.0600, 0.1158, 0.0405, 0.0457, 0.1021, 0.1135, 0.0392], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0381, 0.0499, 0.0385, 0.0284, 0.0278, 0.0306, 0.0306], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:14:06,494 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3250, loss[loss=0.2097, simple_loss=0.3036, pruned_loss=0.05792, over 17025.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3142, pruned_loss=0.08628, over 3326112.40 frames. ], batch size: 50, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:14:38,900 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:56,593 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3300, loss[loss=0.2228, simple_loss=0.3129, pruned_loss=0.06636, over 17292.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3154, pruned_loss=0.08676, over 3320541.05 frames. ], batch size: 52, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:15:57,197 INFO [optim.py:368] (3/8) 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 22:16:22,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3688, 3.0672, 2.8669, 4.9339, 2.2356, 4.7299, 2.7435, 2.7236], device='cuda:3'), covar=tensor([0.0260, 0.0689, 0.0398, 0.0142, 0.1756, 0.0159, 0.0837, 0.1430], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0223, 0.0186, 0.0251, 0.0295, 0.0197, 0.0214, 0.0286], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:16:24,693 INFO [train.py:904] (3/8) Epoch 3, batch 3350, loss[loss=0.22, simple_loss=0.3075, pruned_loss=0.0662, over 17268.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3172, pruned_loss=0.0878, over 3315813.70 frames. ], batch size: 52, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:16:40,764 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-27 22:17:04,740 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2464, 5.0956, 5.1011, 4.5098, 4.8723, 2.2196, 4.7357, 5.1475], device='cuda:3'), covar=tensor([0.0059, 0.0050, 0.0065, 0.0282, 0.0068, 0.1082, 0.0086, 0.0101], device='cuda:3'), in_proj_covar=tensor([0.0079, 0.0068, 0.0104, 0.0121, 0.0077, 0.0117, 0.0095, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:17:33,461 INFO [train.py:904] (3/8) Epoch 3, batch 3400, loss[loss=0.2816, simple_loss=0.3347, pruned_loss=0.1143, over 16745.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3174, pruned_loss=0.08784, over 3307718.27 frames. ], batch size: 124, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:18:13,372 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 3.881e+02 4.662e+02 5.457e+02 1.013e+03, threshold=9.324e+02, percent-clipped=1.0 2023-04-27 22:18:40,209 INFO [train.py:904] (3/8) Epoch 3, batch 3450, loss[loss=0.2667, simple_loss=0.3156, pruned_loss=0.1089, over 16300.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.315, pruned_loss=0.08665, over 3313330.78 frames. ], batch size: 165, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:18:58,643 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:47,199 INFO [train.py:904] (3/8) Epoch 3, batch 3500, loss[loss=0.2399, simple_loss=0.3143, pruned_loss=0.08276, over 16469.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3133, pruned_loss=0.08531, over 3317005.37 frames. ], batch size: 68, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:20:04,593 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:26,131 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 22:20:31,310 INFO [optim.py:368] (3/8) 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 22:20:40,691 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0429, 3.3370, 3.8662, 2.7434, 3.6859, 3.9664, 3.7242, 1.8688], device='cuda:3'), covar=tensor([0.0343, 0.0167, 0.0036, 0.0196, 0.0039, 0.0037, 0.0037, 0.0331], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0056, 0.0060, 0.0112, 0.0055, 0.0064, 0.0061, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:20:48,795 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 22:20:59,079 INFO [train.py:904] (3/8) Epoch 3, batch 3550, loss[loss=0.237, simple_loss=0.32, pruned_loss=0.07703, over 16745.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3126, pruned_loss=0.08519, over 3311387.50 frames. ], batch size: 57, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:21:29,386 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:34,654 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-27 22:21:45,161 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:05,764 INFO [train.py:904] (3/8) Epoch 3, batch 3600, loss[loss=0.2084, simple_loss=0.287, pruned_loss=0.06489, over 17205.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3115, pruned_loss=0.08424, over 3314988.33 frames. ], batch size: 44, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:22:08,549 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7937, 4.8898, 5.3900, 5.3466, 5.3465, 5.0003, 4.8884, 4.8050], device='cuda:3'), covar=tensor([0.0231, 0.0230, 0.0296, 0.0349, 0.0327, 0.0194, 0.0682, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0195, 0.0209, 0.0210, 0.0255, 0.0212, 0.0317, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 22:22:33,415 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:50,840 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-27 22:23:14,442 INFO [train.py:904] (3/8) Epoch 3, batch 3650, loss[loss=0.2051, simple_loss=0.2623, pruned_loss=0.07392, over 16816.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3097, pruned_loss=0.08504, over 3307867.35 frames. ], batch size: 83, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:24:17,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1874, 1.3696, 1.8747, 2.1125, 2.2218, 2.3260, 1.6254, 2.2500], device='cuda:3'), covar=tensor([0.0049, 0.0174, 0.0098, 0.0083, 0.0050, 0.0053, 0.0132, 0.0033], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0123, 0.0110, 0.0102, 0.0085, 0.0069, 0.0112, 0.0066], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 22:24:29,887 INFO [train.py:904] (3/8) Epoch 3, batch 3700, loss[loss=0.2218, simple_loss=0.2874, pruned_loss=0.07808, over 16678.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3082, pruned_loss=0.08637, over 3276191.50 frames. ], batch size: 76, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:25:13,796 INFO [optim.py:368] (3/8) 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 22:25:14,326 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9671, 4.9446, 4.7830, 4.0771, 4.8248, 2.1076, 4.5907, 4.8087], device='cuda:3'), covar=tensor([0.0053, 0.0045, 0.0062, 0.0305, 0.0049, 0.1303, 0.0080, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0067, 0.0100, 0.0115, 0.0073, 0.0113, 0.0090, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:25:22,518 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:42,963 INFO [train.py:904] (3/8) Epoch 3, batch 3750, loss[loss=0.2486, simple_loss=0.3051, pruned_loss=0.09607, over 16725.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3093, pruned_loss=0.08789, over 3272307.90 frames. ], batch size: 134, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:26:46,897 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:26:50,668 INFO [train.py:904] (3/8) Epoch 3, batch 3800, loss[loss=0.2563, simple_loss=0.3261, pruned_loss=0.0932, over 15507.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3101, pruned_loss=0.08967, over 3273483.64 frames. ], batch size: 190, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:27:34,020 INFO [optim.py:368] (3/8) 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 22:27:52,861 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3850, loss[loss=0.2582, simple_loss=0.3064, pruned_loss=0.105, over 16895.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3103, pruned_loss=0.0906, over 3264628.24 frames. ], batch size: 116, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:01,359 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 3900, loss[loss=0.205, simple_loss=0.2768, pruned_loss=0.06659, over 16638.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3082, pruned_loss=0.08985, over 3268213.26 frames. ], batch size: 57, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:36,313 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8466, 3.7847, 1.8735, 3.6847, 2.7180, 3.8724, 1.8332, 2.9245], device='cuda:3'), covar=tensor([0.0046, 0.0204, 0.1353, 0.0059, 0.0572, 0.0290, 0.1288, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0137, 0.0170, 0.0082, 0.0161, 0.0168, 0.0181, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 22:29:47,391 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5670, 4.3505, 4.5183, 4.8103, 4.8620, 4.2153, 4.7203, 4.8361], device='cuda:3'), covar=tensor([0.0568, 0.0546, 0.0928, 0.0350, 0.0335, 0.0542, 0.0488, 0.0294], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0350, 0.0454, 0.0362, 0.0266, 0.0250, 0.0284, 0.0284], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:29:50,321 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5693, 2.3906, 1.8170, 2.3483, 3.0236, 2.8406, 3.2267, 3.1028], device='cuda:3'), covar=tensor([0.0047, 0.0129, 0.0164, 0.0138, 0.0069, 0.0114, 0.0039, 0.0068], device='cuda:3'), in_proj_covar=tensor([0.0060, 0.0118, 0.0118, 0.0117, 0.0111, 0.0120, 0.0079, 0.0096], device='cuda:3'), out_proj_covar=tensor([8.8751e-05, 1.7129e-04, 1.6629e-04, 1.6986e-04, 1.6384e-04, 1.7595e-04, 1.1735e-04, 1.4640e-04], device='cuda:3') 2023-04-27 22:29:56,949 INFO [optim.py:368] (3/8) 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 22:29:57,704 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 22:29:57,733 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-27 22:30:25,198 INFO [train.py:904] (3/8) Epoch 3, batch 3950, loss[loss=0.2386, simple_loss=0.3025, pruned_loss=0.08735, over 16416.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3068, pruned_loss=0.08934, over 3283750.01 frames. ], batch size: 68, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:30:44,705 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8492, 2.7725, 2.7443, 4.2282, 3.9848, 3.9645, 1.3470, 3.4011], device='cuda:3'), covar=tensor([0.1289, 0.0518, 0.0920, 0.0062, 0.0176, 0.0253, 0.1349, 0.0420], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0131, 0.0164, 0.0075, 0.0143, 0.0138, 0.0154, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 22:31:34,849 INFO [train.py:904] (3/8) Epoch 3, batch 4000, loss[loss=0.2596, simple_loss=0.317, pruned_loss=0.1011, over 16437.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3068, pruned_loss=0.08955, over 3268780.61 frames. ], batch size: 146, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:32:17,081 INFO [optim.py:368] (3/8) 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 22:32:45,198 INFO [train.py:904] (3/8) Epoch 3, batch 4050, loss[loss=0.225, simple_loss=0.3019, pruned_loss=0.07404, over 16450.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3047, pruned_loss=0.08628, over 3281774.13 frames. ], batch size: 68, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:33:28,054 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6639, 3.9403, 3.0800, 2.4427, 2.8840, 2.2139, 3.9728, 4.3361], device='cuda:3'), covar=tensor([0.2027, 0.0552, 0.1094, 0.1066, 0.1922, 0.1238, 0.0354, 0.0227], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0245, 0.0261, 0.0214, 0.0305, 0.0197, 0.0226, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:33:46,331 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:33:55,705 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8229, 3.6304, 2.9854, 1.5446, 2.4426, 2.0402, 3.1051, 3.6655], device='cuda:3'), covar=tensor([0.0262, 0.0366, 0.0520, 0.1767, 0.0882, 0.1043, 0.0643, 0.0365], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0119, 0.0158, 0.0146, 0.0140, 0.0131, 0.0145, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-27 22:33:58,267 INFO [train.py:904] (3/8) Epoch 3, batch 4100, loss[loss=0.2489, simple_loss=0.3182, pruned_loss=0.08984, over 16618.00 frames. ], tot_loss[loss=0.2375, simple_loss=0.3051, pruned_loss=0.08493, over 3274209.05 frames. ], batch size: 62, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:34:33,762 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9549, 4.8563, 4.6760, 4.0615, 4.8622, 1.7158, 4.5890, 4.7549], device='cuda:3'), covar=tensor([0.0045, 0.0039, 0.0056, 0.0271, 0.0035, 0.1419, 0.0053, 0.0075], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0066, 0.0097, 0.0116, 0.0075, 0.0116, 0.0089, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:34:42,994 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.218e+02 2.972e+02 3.823e+02 4.924e+02 9.607e+02, threshold=7.646e+02, percent-clipped=2.0 2023-04-27 22:35:13,089 INFO [train.py:904] (3/8) Epoch 3, batch 4150, loss[loss=0.3078, simple_loss=0.36, pruned_loss=0.1278, over 11787.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3146, pruned_loss=0.08981, over 3231652.68 frames. ], batch size: 247, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:35:37,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5205, 2.5129, 2.2432, 3.8272, 1.8617, 3.5693, 2.2842, 2.2293], device='cuda:3'), covar=tensor([0.0314, 0.0739, 0.0472, 0.0196, 0.1756, 0.0225, 0.0888, 0.1371], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0227, 0.0191, 0.0252, 0.0298, 0.0202, 0.0217, 0.0292], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:36:09,390 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-04-27 22:36:22,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7725, 3.6988, 3.6974, 3.7513, 3.6907, 4.1077, 4.0052, 3.8467], device='cuda:3'), covar=tensor([0.1862, 0.1161, 0.1075, 0.1671, 0.2242, 0.1132, 0.0836, 0.1943], device='cuda:3'), in_proj_covar=tensor([0.0227, 0.0307, 0.0284, 0.0272, 0.0350, 0.0304, 0.0250, 0.0365], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:36:27,612 INFO [train.py:904] (3/8) Epoch 3, batch 4200, loss[loss=0.3011, simple_loss=0.3803, pruned_loss=0.1109, over 16696.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3232, pruned_loss=0.09309, over 3195476.72 frames. ], batch size: 134, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:37:10,953 INFO [optim.py:368] (3/8) 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 22:37:35,700 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:37:35,819 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7880, 3.5609, 2.7604, 5.0174, 4.8448, 4.4373, 2.1110, 3.6597], device='cuda:3'), covar=tensor([0.1359, 0.0451, 0.1087, 0.0059, 0.0154, 0.0310, 0.1143, 0.0524], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0160, 0.0071, 0.0130, 0.0132, 0.0150, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 22:37:40,125 INFO [train.py:904] (3/8) Epoch 3, batch 4250, loss[loss=0.2128, simple_loss=0.3016, pruned_loss=0.06203, over 16384.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3259, pruned_loss=0.09338, over 3167427.28 frames. ], batch size: 146, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:38:06,868 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:18,447 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8492, 2.9083, 2.4013, 4.4036, 1.9449, 4.0565, 2.4674, 2.5308], device='cuda:3'), covar=tensor([0.0333, 0.0706, 0.0506, 0.0167, 0.1865, 0.0229, 0.0897, 0.1421], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0229, 0.0193, 0.0248, 0.0301, 0.0201, 0.0219, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:38:53,671 INFO [train.py:904] (3/8) Epoch 3, batch 4300, loss[loss=0.2406, simple_loss=0.3201, pruned_loss=0.08055, over 16843.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3262, pruned_loss=0.09126, over 3176335.57 frames. ], batch size: 42, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:39:05,008 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:34,439 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7848, 3.8838, 3.7576, 2.3711, 3.6030, 3.5838, 3.7417, 1.5969], device='cuda:3'), covar=tensor([0.0410, 0.0013, 0.0023, 0.0248, 0.0026, 0.0047, 0.0015, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0049, 0.0056, 0.0108, 0.0052, 0.0060, 0.0055, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:39:37,583 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:38,217 INFO [optim.py:368] (3/8) 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 22:40:06,092 INFO [train.py:904] (3/8) Epoch 3, batch 4350, loss[loss=0.2674, simple_loss=0.3425, pruned_loss=0.09617, over 17215.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3304, pruned_loss=0.09306, over 3176528.28 frames. ], batch size: 45, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:41:10,562 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 4400, loss[loss=0.2472, simple_loss=0.3317, pruned_loss=0.08138, over 15252.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3328, pruned_loss=0.09466, over 3157727.73 frames. ], batch size: 190, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:41:47,890 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6589, 3.5713, 3.6508, 3.5786, 3.7369, 4.0937, 3.9192, 3.5679], device='cuda:3'), covar=tensor([0.1695, 0.1465, 0.1192, 0.1746, 0.2176, 0.1259, 0.0900, 0.1855], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0307, 0.0285, 0.0270, 0.0350, 0.0303, 0.0246, 0.0361], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 22:41:59,849 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0778, 1.5507, 1.9478, 2.8398, 2.8866, 3.2703, 1.6042, 2.9747], device='cuda:3'), covar=tensor([0.0033, 0.0194, 0.0132, 0.0076, 0.0044, 0.0031, 0.0182, 0.0035], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0121, 0.0108, 0.0101, 0.0086, 0.0066, 0.0113, 0.0063], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 22:42:05,388 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.377e+02 3.698e+02 4.447e+02 5.386e+02 9.920e+02, threshold=8.895e+02, percent-clipped=4.0 2023-04-27 22:42:21,153 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 4450, loss[loss=0.2687, simple_loss=0.3475, pruned_loss=0.09494, over 15313.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3361, pruned_loss=0.09519, over 3158805.30 frames. ], batch size: 190, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:42:55,233 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:47,147 INFO [train.py:904] (3/8) Epoch 3, batch 4500, loss[loss=0.2642, simple_loss=0.3347, pruned_loss=0.09685, over 15477.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3342, pruned_loss=0.09318, over 3181029.70 frames. ], batch size: 191, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:44:22,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9659, 3.5568, 3.6525, 1.5194, 3.9780, 3.8918, 2.9306, 2.8932], device='cuda:3'), covar=tensor([0.0998, 0.0158, 0.0227, 0.1346, 0.0047, 0.0036, 0.0350, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0080, 0.0080, 0.0143, 0.0073, 0.0072, 0.0113, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 22:44:22,413 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:29,087 INFO [optim.py:368] (3/8) 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 22:44:52,902 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:56,550 INFO [train.py:904] (3/8) Epoch 3, batch 4550, loss[loss=0.2509, simple_loss=0.3218, pruned_loss=0.08997, over 17062.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.3337, pruned_loss=0.09285, over 3199732.82 frames. ], batch size: 53, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:07,317 INFO [train.py:904] (3/8) Epoch 3, batch 4600, loss[loss=0.318, simple_loss=0.3816, pruned_loss=0.1272, over 15358.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3352, pruned_loss=0.09314, over 3206436.22 frames. ], batch size: 190, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:10,655 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:20,460 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:44,279 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 3, batch 4650, loss[loss=0.2826, simple_loss=0.3368, pruned_loss=0.1142, over 11712.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.3336, pruned_loss=0.09229, over 3207377.84 frames. ], batch size: 246, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:47:37,779 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9275, 2.6980, 2.4034, 4.5864, 1.9388, 3.9453, 2.3631, 2.3211], device='cuda:3'), covar=tensor([0.0331, 0.0797, 0.0536, 0.0165, 0.2046, 0.0250, 0.0988, 0.1725], device='cuda:3'), in_proj_covar=tensor([0.0254, 0.0234, 0.0195, 0.0256, 0.0309, 0.0202, 0.0221, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:48:24,053 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 4700, loss[loss=0.2798, simple_loss=0.3512, pruned_loss=0.1042, over 15460.00 frames. ], tot_loss[loss=0.2556, simple_loss=0.3303, pruned_loss=0.09044, over 3210875.21 frames. ], batch size: 190, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:48:40,748 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6608, 2.5678, 2.3272, 4.2930, 1.8540, 3.8720, 2.3244, 2.3898], device='cuda:3'), covar=tensor([0.0348, 0.0757, 0.0487, 0.0150, 0.1856, 0.0226, 0.0924, 0.1364], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0236, 0.0196, 0.0257, 0.0312, 0.0205, 0.0221, 0.0305], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:49:17,308 INFO [optim.py:368] (3/8) 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 22:49:41,336 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3653, 2.3208, 1.9063, 2.0097, 2.8111, 2.5997, 3.3170, 3.1843], device='cuda:3'), covar=tensor([0.0011, 0.0109, 0.0160, 0.0153, 0.0073, 0.0110, 0.0023, 0.0042], device='cuda:3'), in_proj_covar=tensor([0.0052, 0.0114, 0.0121, 0.0118, 0.0112, 0.0122, 0.0073, 0.0095], device='cuda:3'), out_proj_covar=tensor([7.4661e-05, 1.6384e-04, 1.7075e-04, 1.7086e-04, 1.6590e-04, 1.7866e-04, 1.0517e-04, 1.4273e-04], device='cuda:3') 2023-04-27 22:49:44,461 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3703, 4.2021, 1.6418, 4.5232, 2.7634, 4.3756, 1.8917, 2.9922], device='cuda:3'), covar=tensor([0.0039, 0.0139, 0.1883, 0.0023, 0.0748, 0.0222, 0.1577, 0.0612], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0128, 0.0171, 0.0078, 0.0160, 0.0158, 0.0179, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 22:49:45,103 INFO [train.py:904] (3/8) Epoch 3, batch 4750, loss[loss=0.2317, simple_loss=0.3094, pruned_loss=0.07698, over 16688.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3262, pruned_loss=0.08842, over 3218974.55 frames. ], batch size: 57, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:52,649 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:50:58,609 INFO [train.py:904] (3/8) Epoch 3, batch 4800, loss[loss=0.2823, simple_loss=0.3543, pruned_loss=0.1052, over 15456.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3229, pruned_loss=0.08664, over 3208604.83 frames. ], batch size: 191, lr: 2.06e-02, grad_scale: 8.0 2023-04-27 22:51:28,791 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:42,398 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8038, 1.0901, 1.2324, 1.6790, 1.7238, 1.8854, 1.3693, 1.7990], device='cuda:3'), covar=tensor([0.0049, 0.0178, 0.0086, 0.0095, 0.0064, 0.0037, 0.0125, 0.0044], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0122, 0.0107, 0.0101, 0.0088, 0.0066, 0.0112, 0.0063], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-27 22:51:47,435 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 4850, loss[loss=0.2732, simple_loss=0.3493, pruned_loss=0.09857, over 15400.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3244, pruned_loss=0.08641, over 3204299.33 frames. ], batch size: 190, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:25,028 INFO [train.py:904] (3/8) Epoch 3, batch 4900, loss[loss=0.2647, simple_loss=0.3269, pruned_loss=0.1013, over 16645.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.324, pruned_loss=0.08556, over 3192360.19 frames. ], batch size: 57, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:28,505 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:29,489 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:57,994 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-27 22:53:58,784 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:54:18,171 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9223, 2.7144, 2.6817, 1.7353, 2.8918, 2.9085, 2.5325, 2.3712], device='cuda:3'), covar=tensor([0.0746, 0.0119, 0.0131, 0.1063, 0.0077, 0.0083, 0.0261, 0.0426], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0082, 0.0078, 0.0145, 0.0072, 0.0073, 0.0110, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-27 22:54:18,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3004, 3.0002, 2.5355, 2.3532, 2.2057, 1.9900, 2.9301, 3.0695], device='cuda:3'), covar=tensor([0.1595, 0.0454, 0.0982, 0.1045, 0.1479, 0.1187, 0.0373, 0.0353], device='cuda:3'), in_proj_covar=tensor([0.0262, 0.0233, 0.0249, 0.0206, 0.0281, 0.0186, 0.0214, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:54:35,307 INFO [train.py:904] (3/8) Epoch 3, batch 4950, loss[loss=0.2838, simple_loss=0.3663, pruned_loss=0.1006, over 17162.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3241, pruned_loss=0.08542, over 3202290.84 frames. ], batch size: 49, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:54:36,194 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:08,064 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:28,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3684, 1.4236, 1.7980, 2.6444, 2.5395, 2.7151, 1.5521, 2.7100], device='cuda:3'), covar=tensor([0.0060, 0.0226, 0.0136, 0.0096, 0.0056, 0.0050, 0.0175, 0.0028], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0123, 0.0108, 0.0101, 0.0087, 0.0067, 0.0110, 0.0062], device='cuda:3'), out_proj_covar=tensor([1.3282e-04, 1.9598e-04, 1.7871e-04, 1.6696e-04, 1.3877e-04, 1.0488e-04, 1.7383e-04, 9.6925e-05], device='cuda:3') 2023-04-27 22:55:48,336 INFO [train.py:904] (3/8) Epoch 3, batch 5000, loss[loss=0.2111, simple_loss=0.3013, pruned_loss=0.06043, over 17259.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.325, pruned_loss=0.08514, over 3211932.71 frames. ], batch size: 52, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:02,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5735, 4.3947, 4.5961, 4.8669, 4.9348, 4.3869, 4.9225, 4.8991], device='cuda:3'), covar=tensor([0.0613, 0.0530, 0.0863, 0.0295, 0.0283, 0.0425, 0.0268, 0.0228], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0347, 0.0455, 0.0345, 0.0257, 0.0244, 0.0275, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 22:56:35,248 INFO [optim.py:368] (3/8) 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 22:56:59,649 INFO [train.py:904] (3/8) Epoch 3, batch 5050, loss[loss=0.2501, simple_loss=0.3264, pruned_loss=0.08694, over 16768.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3255, pruned_loss=0.08468, over 3211215.07 frames. ], batch size: 83, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:59,955 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:57:41,650 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-27 22:58:08,624 INFO [train.py:904] (3/8) Epoch 3, batch 5100, loss[loss=0.2273, simple_loss=0.298, pruned_loss=0.07829, over 17035.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3227, pruned_loss=0.08313, over 3215844.26 frames. ], batch size: 55, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:58:38,786 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:58:48,450 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:58:57,532 INFO [optim.py:368] (3/8) 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 22:59:14,341 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8469, 1.9911, 1.6104, 1.5956, 2.6329, 2.3146, 2.7630, 2.8059], device='cuda:3'), covar=tensor([0.0016, 0.0165, 0.0191, 0.0210, 0.0075, 0.0126, 0.0036, 0.0063], device='cuda:3'), in_proj_covar=tensor([0.0053, 0.0119, 0.0122, 0.0121, 0.0114, 0.0123, 0.0074, 0.0097], device='cuda:3'), out_proj_covar=tensor([7.5768e-05, 1.7137e-04, 1.7121e-04, 1.7400e-04, 1.6661e-04, 1.7960e-04, 1.0512e-04, 1.4561e-04], device='cuda:3') 2023-04-27 22:59:23,205 INFO [train.py:904] (3/8) Epoch 3, batch 5150, loss[loss=0.2849, simple_loss=0.366, pruned_loss=0.1019, over 16820.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3231, pruned_loss=0.08281, over 3199184.24 frames. ], batch size: 116, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:59:35,418 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7116, 3.1938, 2.6409, 4.6231, 4.3652, 4.2621, 2.1008, 3.0876], device='cuda:3'), covar=tensor([0.1688, 0.0470, 0.1131, 0.0078, 0.0158, 0.0211, 0.1287, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0129, 0.0164, 0.0071, 0.0130, 0.0138, 0.0155, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 22:59:50,322 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:00:19,380 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:00:35,982 INFO [train.py:904] (3/8) Epoch 3, batch 5200, loss[loss=0.2288, simple_loss=0.2948, pruned_loss=0.08147, over 16249.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3218, pruned_loss=0.0824, over 3210540.16 frames. ], batch size: 35, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:00:40,368 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:00:51,384 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8020, 2.7489, 2.2225, 3.8470, 3.7543, 3.6402, 1.4524, 2.6157], device='cuda:3'), covar=tensor([0.1227, 0.0391, 0.1103, 0.0058, 0.0141, 0.0225, 0.1259, 0.0715], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0127, 0.0162, 0.0070, 0.0128, 0.0137, 0.0153, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 23:01:22,143 INFO [optim.py:368] (3/8) 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 23:01:45,800 INFO [train.py:904] (3/8) Epoch 3, batch 5250, loss[loss=0.2131, simple_loss=0.2879, pruned_loss=0.0691, over 17180.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3197, pruned_loss=0.08224, over 3217158.00 frames. ], batch size: 46, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:01:47,881 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:56,098 INFO [train.py:904] (3/8) Epoch 3, batch 5300, loss[loss=0.2744, simple_loss=0.3353, pruned_loss=0.1067, over 12079.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.315, pruned_loss=0.08029, over 3218232.38 frames. ], batch size: 246, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:03:43,233 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 5350, loss[loss=0.2366, simple_loss=0.317, pruned_loss=0.07812, over 16538.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3138, pruned_loss=0.07973, over 3221719.58 frames. ], batch size: 68, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:04:08,379 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:04:17,457 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-27 23:05:17,182 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 5400, loss[loss=0.2671, simple_loss=0.3433, pruned_loss=0.09547, over 16919.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3169, pruned_loss=0.08073, over 3227371.41 frames. ], batch size: 109, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:06:01,106 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4452, 3.5963, 2.6463, 2.1812, 2.8187, 2.2383, 3.6514, 4.1050], device='cuda:3'), covar=tensor([0.2033, 0.0699, 0.1253, 0.1190, 0.1791, 0.1196, 0.0420, 0.0303], device='cuda:3'), in_proj_covar=tensor([0.0262, 0.0236, 0.0251, 0.0209, 0.0286, 0.0186, 0.0214, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:06:07,786 INFO [optim.py:368] (3/8) 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 23:06:32,026 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-27 23:06:34,320 INFO [train.py:904] (3/8) Epoch 3, batch 5450, loss[loss=0.3753, simple_loss=0.4193, pruned_loss=0.1656, over 15392.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3214, pruned_loss=0.08374, over 3211743.31 frames. ], batch size: 191, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:06:38,473 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9460, 5.4082, 5.3531, 5.4818, 5.2786, 5.9739, 5.5691, 5.4161], device='cuda:3'), covar=tensor([0.0558, 0.1104, 0.1032, 0.1457, 0.2142, 0.0743, 0.0789, 0.1660], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0297, 0.0279, 0.0271, 0.0344, 0.0304, 0.0243, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:06:52,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7983, 2.8646, 2.4258, 3.9667, 3.7577, 3.7383, 1.5133, 2.8089], device='cuda:3'), covar=tensor([0.1462, 0.0471, 0.1161, 0.0102, 0.0241, 0.0322, 0.1399, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0132, 0.0167, 0.0072, 0.0136, 0.0142, 0.0157, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 23:07:24,678 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:07:28,827 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 23:07:48,601 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8457, 3.8452, 1.5248, 4.0187, 2.4889, 4.0386, 1.8931, 2.6278], device='cuda:3'), covar=tensor([0.0050, 0.0231, 0.1725, 0.0037, 0.0769, 0.0259, 0.1408, 0.0713], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0134, 0.0175, 0.0082, 0.0165, 0.0163, 0.0180, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 23:07:49,275 INFO [train.py:904] (3/8) Epoch 3, batch 5500, loss[loss=0.288, simple_loss=0.3631, pruned_loss=0.1064, over 16727.00 frames. ], tot_loss[loss=0.2569, simple_loss=0.3314, pruned_loss=0.09123, over 3192760.57 frames. ], batch size: 89, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:08:17,353 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 23:08:19,375 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0159, 1.5873, 1.4707, 1.4547, 1.8177, 1.5676, 1.7252, 1.8961], device='cuda:3'), covar=tensor([0.0023, 0.0088, 0.0125, 0.0127, 0.0070, 0.0104, 0.0046, 0.0074], device='cuda:3'), in_proj_covar=tensor([0.0053, 0.0115, 0.0119, 0.0120, 0.0111, 0.0123, 0.0074, 0.0097], device='cuda:3'), out_proj_covar=tensor([7.4981e-05, 1.6546e-04, 1.6544e-04, 1.7302e-04, 1.6223e-04, 1.7969e-04, 1.0599e-04, 1.4521e-04], device='cuda:3') 2023-04-27 23:08:34,609 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9371, 3.3427, 3.5708, 2.4876, 3.4234, 3.5959, 3.5574, 1.8501], device='cuda:3'), covar=tensor([0.0336, 0.0032, 0.0028, 0.0201, 0.0037, 0.0052, 0.0027, 0.0304], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0052, 0.0059, 0.0106, 0.0051, 0.0061, 0.0057, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:08:39,221 INFO [optim.py:368] (3/8) 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 23:09:06,199 INFO [train.py:904] (3/8) Epoch 3, batch 5550, loss[loss=0.4728, simple_loss=0.4722, pruned_loss=0.2366, over 11201.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.341, pruned_loss=0.09919, over 3167040.28 frames. ], batch size: 248, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:09:18,193 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:10:10,511 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-27 23:10:20,719 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0514, 2.9810, 2.5066, 4.2878, 1.9885, 4.1615, 2.5712, 2.5665], device='cuda:3'), covar=tensor([0.0269, 0.0681, 0.0444, 0.0181, 0.1731, 0.0191, 0.0801, 0.1298], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0230, 0.0193, 0.0250, 0.0302, 0.0204, 0.0222, 0.0292], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:10:25,184 INFO [train.py:904] (3/8) Epoch 3, batch 5600, loss[loss=0.3064, simple_loss=0.3689, pruned_loss=0.1219, over 16793.00 frames. ], tot_loss[loss=0.2797, simple_loss=0.3478, pruned_loss=0.1058, over 3132143.37 frames. ], batch size: 124, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:10:56,234 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:11:21,465 INFO [optim.py:368] (3/8) 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 23:11:44,571 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9135, 3.6467, 2.5431, 4.8268, 4.6240, 4.2836, 2.0261, 3.3755], device='cuda:3'), covar=tensor([0.1425, 0.0357, 0.1223, 0.0059, 0.0137, 0.0242, 0.1159, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0127, 0.0164, 0.0070, 0.0133, 0.0140, 0.0153, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 23:11:48,513 INFO [train.py:904] (3/8) Epoch 3, batch 5650, loss[loss=0.2968, simple_loss=0.3653, pruned_loss=0.1142, over 16553.00 frames. ], tot_loss[loss=0.2902, simple_loss=0.3549, pruned_loss=0.1128, over 3098661.12 frames. ], batch size: 75, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:12:03,968 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8079, 2.5646, 2.5622, 1.8345, 2.4071, 2.4333, 2.4204, 1.7555], device='cuda:3'), covar=tensor([0.0280, 0.0032, 0.0051, 0.0225, 0.0045, 0.0075, 0.0033, 0.0281], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0051, 0.0058, 0.0107, 0.0050, 0.0062, 0.0056, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:12:17,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 23:12:39,989 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8405, 4.5050, 2.0381, 4.8615, 2.8484, 4.9246, 2.3996, 3.2086], device='cuda:3'), covar=tensor([0.0039, 0.0152, 0.1710, 0.0024, 0.0789, 0.0140, 0.1367, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0131, 0.0173, 0.0079, 0.0161, 0.0162, 0.0180, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 23:13:10,043 INFO [train.py:904] (3/8) Epoch 3, batch 5700, loss[loss=0.3607, simple_loss=0.394, pruned_loss=0.1637, over 11405.00 frames. ], tot_loss[loss=0.2946, simple_loss=0.3579, pruned_loss=0.1157, over 3068535.65 frames. ], batch size: 248, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:13:36,294 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 23:14:00,537 INFO [optim.py:368] (3/8) 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 23:14:15,786 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 23:14:21,075 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6991, 2.6453, 1.6097, 2.6526, 2.0023, 2.6870, 1.7688, 2.2609], device='cuda:3'), covar=tensor([0.0080, 0.0218, 0.1125, 0.0068, 0.0591, 0.0384, 0.1040, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0131, 0.0172, 0.0079, 0.0161, 0.0161, 0.0180, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 23:14:27,438 INFO [train.py:904] (3/8) Epoch 3, batch 5750, loss[loss=0.3752, simple_loss=0.3983, pruned_loss=0.176, over 11569.00 frames. ], tot_loss[loss=0.2962, simple_loss=0.3596, pruned_loss=0.1164, over 3062794.56 frames. ], batch size: 248, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:21,998 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:15:47,176 INFO [train.py:904] (3/8) Epoch 3, batch 5800, loss[loss=0.2638, simple_loss=0.3398, pruned_loss=0.09387, over 16781.00 frames. ], tot_loss[loss=0.2955, simple_loss=0.36, pruned_loss=0.1155, over 3059394.98 frames. ], batch size: 124, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:48,974 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:16:22,367 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-27 23:16:36,164 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.751e+02 5.049e+02 6.109e+02 8.114e+02 1.629e+03, threshold=1.222e+03, percent-clipped=2.0 2023-04-27 23:16:48,880 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3978, 4.6469, 4.7226, 4.7659, 4.7929, 5.1920, 4.8273, 4.6358], device='cuda:3'), covar=tensor([0.0804, 0.1167, 0.0802, 0.1314, 0.1600, 0.0645, 0.0822, 0.1632], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0302, 0.0283, 0.0270, 0.0346, 0.0303, 0.0250, 0.0361], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:17:05,081 INFO [train.py:904] (3/8) Epoch 3, batch 5850, loss[loss=0.3363, simple_loss=0.3729, pruned_loss=0.1498, over 11431.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3574, pruned_loss=0.1132, over 3058170.38 frames. ], batch size: 246, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:17:24,108 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:31,355 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:18:26,955 INFO [train.py:904] (3/8) Epoch 3, batch 5900, loss[loss=0.2639, simple_loss=0.3465, pruned_loss=0.09062, over 16494.00 frames. ], tot_loss[loss=0.2894, simple_loss=0.3557, pruned_loss=0.1116, over 3071903.54 frames. ], batch size: 75, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:18:51,852 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:19:15,341 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:19:47,796 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:48,573 INFO [train.py:904] (3/8) Epoch 3, batch 5950, loss[loss=0.2754, simple_loss=0.3456, pruned_loss=0.1026, over 16663.00 frames. ], tot_loss[loss=0.2874, simple_loss=0.356, pruned_loss=0.1094, over 3075096.07 frames. ], batch size: 134, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:20:40,881 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1025, 2.2060, 1.8473, 1.8903, 2.8345, 2.5129, 3.2283, 3.0685], device='cuda:3'), covar=tensor([0.0014, 0.0132, 0.0169, 0.0167, 0.0068, 0.0111, 0.0036, 0.0059], device='cuda:3'), in_proj_covar=tensor([0.0052, 0.0117, 0.0121, 0.0122, 0.0110, 0.0123, 0.0077, 0.0098], device='cuda:3'), out_proj_covar=tensor([7.1134e-05, 1.6685e-04, 1.6758e-04, 1.7429e-04, 1.6094e-04, 1.7746e-04, 1.0916e-04, 1.4482e-04], device='cuda:3') 2023-04-27 23:21:07,922 INFO [train.py:904] (3/8) Epoch 3, batch 6000, loss[loss=0.2407, simple_loss=0.3217, pruned_loss=0.07985, over 16780.00 frames. ], tot_loss[loss=0.2855, simple_loss=0.3544, pruned_loss=0.1083, over 3088548.74 frames. ], batch size: 83, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:21:07,923 INFO [train.py:929] (3/8) Computing validation loss 2023-04-27 23:21:18,891 INFO [train.py:938] (3/8) Epoch 3, validation: loss=0.2097, simple_loss=0.3184, pruned_loss=0.05055, over 944034.00 frames. 2023-04-27 23:21:18,892 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-27 23:21:34,656 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:22:08,530 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4425, 3.4284, 3.3400, 2.6998, 3.3089, 2.0166, 3.1791, 3.0451], device='cuda:3'), covar=tensor([0.0090, 0.0062, 0.0097, 0.0330, 0.0066, 0.1373, 0.0081, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0061, 0.0095, 0.0112, 0.0071, 0.0120, 0.0082, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:22:12,976 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6709, 2.6968, 1.6309, 2.6899, 2.0690, 2.7057, 1.8633, 2.3642], device='cuda:3'), covar=tensor([0.0110, 0.0275, 0.1074, 0.0063, 0.0590, 0.0451, 0.0980, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0132, 0.0171, 0.0080, 0.0163, 0.0163, 0.0179, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-27 23:22:36,223 INFO [train.py:904] (3/8) Epoch 3, batch 6050, loss[loss=0.2593, simple_loss=0.346, pruned_loss=0.0863, over 17038.00 frames. ], tot_loss[loss=0.2836, simple_loss=0.3525, pruned_loss=0.1073, over 3091983.06 frames. ], batch size: 50, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:51,471 INFO [train.py:904] (3/8) Epoch 3, batch 6100, loss[loss=0.2439, simple_loss=0.3187, pruned_loss=0.08457, over 16904.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3521, pruned_loss=0.1063, over 3095304.24 frames. ], batch size: 109, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:24:12,635 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 3, batch 6150, loss[loss=0.3306, simple_loss=0.3737, pruned_loss=0.1437, over 11658.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3496, pruned_loss=0.1055, over 3095970.90 frames. ], batch size: 247, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:25:23,313 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:50,230 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:26:28,208 INFO [train.py:904] (3/8) Epoch 3, batch 6200, loss[loss=0.2936, simple_loss=0.3606, pruned_loss=0.1133, over 17001.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3472, pruned_loss=0.1044, over 3114546.93 frames. ], batch size: 41, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:26:48,630 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:27:03,424 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:27:18,519 INFO [optim.py:368] (3/8) 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 23:27:41,897 INFO [train.py:904] (3/8) Epoch 3, batch 6250, loss[loss=0.2586, simple_loss=0.3388, pruned_loss=0.08922, over 16919.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3459, pruned_loss=0.1031, over 3139083.78 frames. ], batch size: 109, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:27:58,188 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:28:00,057 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6786, 5.0630, 4.6724, 4.8666, 3.4427, 4.8863, 4.8609, 4.5005], device='cuda:3'), covar=tensor([0.0826, 0.0308, 0.0442, 0.0306, 0.1895, 0.0419, 0.0272, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0130, 0.0169, 0.0142, 0.0200, 0.0161, 0.0126, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:28:13,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0088, 2.1083, 2.0575, 3.2315, 1.6438, 2.9317, 2.0022, 1.8410], device='cuda:3'), covar=tensor([0.0475, 0.1066, 0.0652, 0.0317, 0.2445, 0.0399, 0.1269, 0.1900], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0234, 0.0198, 0.0257, 0.0311, 0.0211, 0.0223, 0.0298], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:28:54,859 INFO [train.py:904] (3/8) Epoch 3, batch 6300, loss[loss=0.3513, simple_loss=0.3794, pruned_loss=0.1616, over 11164.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3467, pruned_loss=0.1037, over 3110711.78 frames. ], batch size: 247, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:29:02,862 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:29:48,250 INFO [optim.py:368] (3/8) 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 23:29:50,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4118, 2.5070, 2.2950, 3.9025, 1.8179, 3.7618, 2.2940, 2.2443], device='cuda:3'), covar=tensor([0.0372, 0.0832, 0.0546, 0.0214, 0.1919, 0.0230, 0.0959, 0.1395], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0235, 0.0199, 0.0259, 0.0311, 0.0213, 0.0225, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:30:03,827 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 23:30:11,907 INFO [train.py:904] (3/8) Epoch 3, batch 6350, loss[loss=0.3483, simple_loss=0.3811, pruned_loss=0.1578, over 11206.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3486, pruned_loss=0.1065, over 3087342.85 frames. ], batch size: 247, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:31:25,030 INFO [train.py:904] (3/8) Epoch 3, batch 6400, loss[loss=0.3706, simple_loss=0.4152, pruned_loss=0.163, over 11066.00 frames. ], tot_loss[loss=0.281, simple_loss=0.3486, pruned_loss=0.1067, over 3094228.52 frames. ], batch size: 248, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:12,500 INFO [optim.py:368] (3/8) 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 23:32:25,308 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3522, 4.2726, 4.1777, 3.4867, 4.1524, 1.5817, 3.9091, 4.1110], device='cuda:3'), covar=tensor([0.0059, 0.0045, 0.0069, 0.0298, 0.0057, 0.1555, 0.0078, 0.0096], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0061, 0.0094, 0.0110, 0.0071, 0.0120, 0.0083, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:32:35,291 INFO [train.py:904] (3/8) Epoch 3, batch 6450, loss[loss=0.264, simple_loss=0.3324, pruned_loss=0.09777, over 16373.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3478, pruned_loss=0.1055, over 3095001.44 frames. ], batch size: 146, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:42,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7886, 3.5957, 3.8339, 3.7154, 3.7814, 4.1303, 4.0072, 3.7033], device='cuda:3'), covar=tensor([0.1416, 0.1590, 0.1149, 0.1752, 0.1968, 0.1059, 0.0888, 0.1769], device='cuda:3'), in_proj_covar=tensor([0.0231, 0.0313, 0.0291, 0.0277, 0.0359, 0.0313, 0.0256, 0.0370], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:32:47,295 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:05,686 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:08,813 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7656, 3.2999, 2.3227, 4.4255, 4.2386, 3.9801, 1.7919, 3.0273], device='cuda:3'), covar=tensor([0.1267, 0.0388, 0.1163, 0.0059, 0.0174, 0.0288, 0.1106, 0.0649], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0132, 0.0163, 0.0071, 0.0136, 0.0143, 0.0153, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-04-27 23:33:44,605 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0625, 2.5067, 2.4310, 3.2844, 3.0754, 3.2873, 1.8647, 2.7619], device='cuda:3'), covar=tensor([0.1138, 0.0376, 0.0932, 0.0103, 0.0244, 0.0294, 0.1095, 0.0574], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0133, 0.0166, 0.0072, 0.0137, 0.0145, 0.0155, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-27 23:33:53,753 INFO [train.py:904] (3/8) Epoch 3, batch 6500, loss[loss=0.3087, simple_loss=0.3502, pruned_loss=0.1336, over 11962.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.345, pruned_loss=0.1042, over 3108739.47 frames. ], batch size: 250, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:34:01,036 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:29,489 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:44,551 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.621e+02 5.604e+02 6.997e+02 1.424e+03, threshold=1.121e+03, percent-clipped=2.0 2023-04-27 23:35:10,790 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9333, 3.2621, 3.4399, 2.3029, 3.1612, 3.2991, 3.3585, 1.7834], device='cuda:3'), covar=tensor([0.0319, 0.0024, 0.0028, 0.0205, 0.0031, 0.0066, 0.0021, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0049, 0.0055, 0.0107, 0.0050, 0.0061, 0.0057, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:35:12,110 INFO [train.py:904] (3/8) Epoch 3, batch 6550, loss[loss=0.2647, simple_loss=0.3535, pruned_loss=0.08793, over 16399.00 frames. ], tot_loss[loss=0.2789, simple_loss=0.3481, pruned_loss=0.1048, over 3116974.54 frames. ], batch size: 146, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:35:45,474 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:35:48,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0044, 5.2630, 4.9722, 5.0689, 4.5296, 4.4611, 4.7744, 5.3474], device='cuda:3'), covar=tensor([0.0500, 0.0546, 0.0722, 0.0368, 0.0603, 0.0580, 0.0476, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0374, 0.0332, 0.0241, 0.0242, 0.0240, 0.0298, 0.0264], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:36:28,902 INFO [train.py:904] (3/8) Epoch 3, batch 6600, loss[loss=0.2952, simple_loss=0.3587, pruned_loss=0.1159, over 15173.00 frames. ], tot_loss[loss=0.2817, simple_loss=0.3509, pruned_loss=0.1062, over 3113909.68 frames. ], batch size: 190, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:36:33,592 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:36:37,585 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:44,029 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8962, 3.8516, 3.1026, 2.6819, 3.0279, 2.2884, 4.1193, 4.4402], device='cuda:3'), covar=tensor([0.1935, 0.0716, 0.1183, 0.1104, 0.2043, 0.1267, 0.0360, 0.0325], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0240, 0.0258, 0.0215, 0.0297, 0.0193, 0.0220, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:37:20,959 INFO [optim.py:368] (3/8) 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 23:37:28,409 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4872, 3.4456, 3.3839, 2.9591, 3.3748, 2.0729, 3.1721, 3.1610], device='cuda:3'), covar=tensor([0.0065, 0.0053, 0.0073, 0.0202, 0.0053, 0.1075, 0.0067, 0.0096], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0061, 0.0095, 0.0110, 0.0070, 0.0120, 0.0082, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:37:45,805 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:46,399 INFO [train.py:904] (3/8) Epoch 3, batch 6650, loss[loss=0.2641, simple_loss=0.3306, pruned_loss=0.09875, over 16865.00 frames. ], tot_loss[loss=0.2827, simple_loss=0.3511, pruned_loss=0.1071, over 3106546.84 frames. ], batch size: 116, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:37:51,224 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:38:09,002 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:38:15,898 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-04-27 23:39:03,588 INFO [train.py:904] (3/8) Epoch 3, batch 6700, loss[loss=0.3343, simple_loss=0.3715, pruned_loss=0.1485, over 11428.00 frames. ], tot_loss[loss=0.282, simple_loss=0.3497, pruned_loss=0.1072, over 3106372.42 frames. ], batch size: 247, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:39:11,535 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5984, 3.7420, 3.5380, 3.6693, 2.8349, 3.6789, 3.4737, 3.3962], device='cuda:3'), covar=tensor([0.0541, 0.0239, 0.0366, 0.0229, 0.1136, 0.0343, 0.0791, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0128, 0.0168, 0.0140, 0.0197, 0.0160, 0.0126, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:39:19,089 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:22,856 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6145, 4.5413, 4.3747, 3.7191, 4.3996, 1.8776, 4.1581, 4.4231], device='cuda:3'), covar=tensor([0.0047, 0.0040, 0.0058, 0.0247, 0.0048, 0.1258, 0.0060, 0.0075], device='cuda:3'), in_proj_covar=tensor([0.0074, 0.0061, 0.0097, 0.0111, 0.0071, 0.0120, 0.0082, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:39:57,110 INFO [optim.py:368] (3/8) 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 23:40:06,994 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 23:40:21,183 INFO [train.py:904] (3/8) Epoch 3, batch 6750, loss[loss=0.3444, simple_loss=0.3881, pruned_loss=0.1503, over 11795.00 frames. ], tot_loss[loss=0.2827, simple_loss=0.3496, pruned_loss=0.1079, over 3078012.12 frames. ], batch size: 247, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:40:23,765 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 23:40:28,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 23:40:34,491 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7344, 2.8393, 2.2414, 3.9684, 3.7914, 3.6868, 1.5054, 2.8582], device='cuda:3'), covar=tensor([0.1268, 0.0428, 0.1182, 0.0068, 0.0218, 0.0297, 0.1237, 0.0639], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0131, 0.0164, 0.0070, 0.0136, 0.0142, 0.0149, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-27 23:40:46,958 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0321, 4.2087, 3.3083, 2.7775, 3.2512, 2.3925, 4.4692, 4.6864], device='cuda:3'), covar=tensor([0.1869, 0.0606, 0.1092, 0.1005, 0.1927, 0.1305, 0.0314, 0.0332], device='cuda:3'), in_proj_covar=tensor([0.0265, 0.0237, 0.0251, 0.0210, 0.0295, 0.0188, 0.0217, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:40:51,259 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:41:37,189 INFO [train.py:904] (3/8) Epoch 3, batch 6800, loss[loss=0.273, simple_loss=0.35, pruned_loss=0.09798, over 16697.00 frames. ], tot_loss[loss=0.283, simple_loss=0.3499, pruned_loss=0.108, over 3055125.15 frames. ], batch size: 134, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:42:03,929 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:31,274 INFO [optim.py:368] (3/8) 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 23:42:51,757 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:55,392 INFO [train.py:904] (3/8) Epoch 3, batch 6850, loss[loss=0.2656, simple_loss=0.3614, pruned_loss=0.08492, over 16915.00 frames. ], tot_loss[loss=0.284, simple_loss=0.3515, pruned_loss=0.1082, over 3067844.06 frames. ], batch size: 90, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:10,166 INFO [train.py:904] (3/8) Epoch 3, batch 6900, loss[loss=0.3754, simple_loss=0.4065, pruned_loss=0.1721, over 11501.00 frames. ], tot_loss[loss=0.2846, simple_loss=0.354, pruned_loss=0.1076, over 3085378.80 frames. ], batch size: 250, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:22,720 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:29,505 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3944, 3.4466, 3.9473, 3.9498, 3.9308, 3.5203, 3.6211, 3.6443], device='cuda:3'), covar=tensor([0.0320, 0.0384, 0.0349, 0.0386, 0.0433, 0.0318, 0.0795, 0.0353], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0177, 0.0190, 0.0186, 0.0230, 0.0194, 0.0296, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-27 23:45:02,577 INFO [optim.py:368] (3/8) 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 23:45:08,093 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4229, 4.4191, 4.9542, 4.9138, 4.9367, 4.4587, 4.4991, 4.3445], device='cuda:3'), covar=tensor([0.0218, 0.0263, 0.0319, 0.0398, 0.0438, 0.0264, 0.0703, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0177, 0.0192, 0.0187, 0.0232, 0.0195, 0.0299, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-27 23:45:28,455 INFO [train.py:904] (3/8) Epoch 3, batch 6950, loss[loss=0.2981, simple_loss=0.3753, pruned_loss=0.1105, over 16485.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3574, pruned_loss=0.1114, over 3061392.14 frames. ], batch size: 146, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:45:41,655 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:46:43,369 INFO [train.py:904] (3/8) Epoch 3, batch 7000, loss[loss=0.2706, simple_loss=0.3445, pruned_loss=0.09836, over 16664.00 frames. ], tot_loss[loss=0.2896, simple_loss=0.3578, pruned_loss=0.1107, over 3074722.99 frames. ], batch size: 134, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:46:51,004 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:01,975 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 23:47:14,481 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6365, 4.7007, 5.0876, 5.0694, 5.1631, 4.8039, 4.4775, 4.5073], device='cuda:3'), covar=tensor([0.0316, 0.0298, 0.0363, 0.0414, 0.0389, 0.0325, 0.1145, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0176, 0.0190, 0.0185, 0.0231, 0.0197, 0.0297, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-27 23:47:35,868 INFO [optim.py:368] (3/8) 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 23:48:01,107 INFO [train.py:904] (3/8) Epoch 3, batch 7050, loss[loss=0.3376, simple_loss=0.3791, pruned_loss=0.148, over 11641.00 frames. ], tot_loss[loss=0.2896, simple_loss=0.358, pruned_loss=0.1106, over 3071916.51 frames. ], batch size: 247, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:49:19,631 INFO [train.py:904] (3/8) Epoch 3, batch 7100, loss[loss=0.3003, simple_loss=0.3663, pruned_loss=0.1171, over 16386.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3567, pruned_loss=0.1103, over 3057745.74 frames. ], batch size: 146, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:50:12,100 INFO [optim.py:368] (3/8) 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 23:50:36,342 INFO [train.py:904] (3/8) Epoch 3, batch 7150, loss[loss=0.2596, simple_loss=0.3356, pruned_loss=0.09179, over 17019.00 frames. ], tot_loss[loss=0.286, simple_loss=0.3534, pruned_loss=0.1093, over 3049869.70 frames. ], batch size: 55, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:51,150 INFO [train.py:904] (3/8) Epoch 3, batch 7200, loss[loss=0.2231, simple_loss=0.3074, pruned_loss=0.06945, over 17041.00 frames. ], tot_loss[loss=0.2816, simple_loss=0.3498, pruned_loss=0.1068, over 3043959.94 frames. ], batch size: 53, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:55,800 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:52:45,515 INFO [optim.py:368] (3/8) 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 23:52:55,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8784, 3.5139, 3.7093, 2.6510, 3.4377, 3.6111, 3.7217, 1.8217], device='cuda:3'), covar=tensor([0.0326, 0.0024, 0.0028, 0.0190, 0.0033, 0.0052, 0.0018, 0.0325], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0051, 0.0056, 0.0109, 0.0052, 0.0060, 0.0057, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-27 23:53:09,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5908, 3.5501, 3.9764, 3.9640, 3.9693, 3.6135, 3.7539, 3.6381], device='cuda:3'), covar=tensor([0.0222, 0.0307, 0.0303, 0.0333, 0.0338, 0.0261, 0.0613, 0.0324], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0171, 0.0191, 0.0186, 0.0227, 0.0194, 0.0293, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-27 23:53:12,423 INFO [train.py:904] (3/8) Epoch 3, batch 7250, loss[loss=0.2498, simple_loss=0.3225, pruned_loss=0.08856, over 16814.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.3463, pruned_loss=0.1045, over 3058528.12 frames. ], batch size: 102, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:53:26,456 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:53:43,085 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 23:53:58,220 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8500, 2.7760, 2.5749, 4.3772, 1.9339, 4.1855, 2.5682, 2.3742], device='cuda:3'), covar=tensor([0.0347, 0.0824, 0.0509, 0.0188, 0.2069, 0.0235, 0.0969, 0.1669], device='cuda:3'), in_proj_covar=tensor([0.0262, 0.0237, 0.0199, 0.0264, 0.0314, 0.0211, 0.0227, 0.0302], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-27 23:54:26,523 INFO [train.py:904] (3/8) Epoch 3, batch 7300, loss[loss=0.2535, simple_loss=0.3365, pruned_loss=0.08525, over 16478.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3448, pruned_loss=0.1037, over 3051691.40 frames. ], batch size: 75, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:54:35,522 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:55:17,368 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 7350, loss[loss=0.2636, simple_loss=0.3409, pruned_loss=0.09317, over 16854.00 frames. ], tot_loss[loss=0.2755, simple_loss=0.3444, pruned_loss=0.1032, over 3042225.74 frames. ], batch size: 102, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:55:46,793 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:59,424 INFO [train.py:904] (3/8) Epoch 3, batch 7400, loss[loss=0.2796, simple_loss=0.3535, pruned_loss=0.1029, over 16562.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3466, pruned_loss=0.105, over 3031553.36 frames. ], batch size: 68, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:57:20,383 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:58:16,704 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 23:58:18,284 INFO [train.py:904] (3/8) Epoch 3, batch 7450, loss[loss=0.2987, simple_loss=0.3614, pruned_loss=0.118, over 16731.00 frames. ], tot_loss[loss=0.28, simple_loss=0.348, pruned_loss=0.1059, over 3045528.28 frames. ], batch size: 89, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:58:29,784 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-27 23:58:58,373 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:59:39,384 INFO [train.py:904] (3/8) Epoch 3, batch 7500, loss[loss=0.3511, simple_loss=0.3895, pruned_loss=0.1564, over 11595.00 frames. ], tot_loss[loss=0.281, simple_loss=0.3493, pruned_loss=0.1063, over 3022167.30 frames. ], batch size: 248, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:59:44,138 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:33,136 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.883e+02 4.957e+02 5.924e+02 7.629e+02 1.550e+03, threshold=1.185e+03, percent-clipped=6.0 2023-04-28 00:00:55,939 INFO [train.py:904] (3/8) Epoch 3, batch 7550, loss[loss=0.2635, simple_loss=0.3372, pruned_loss=0.09489, over 16698.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3491, pruned_loss=0.1072, over 3016908.48 frames. ], batch size: 134, lr: 1.96e-02, grad_scale: 4.0 2023-04-28 00:00:58,763 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:02:13,359 INFO [train.py:904] (3/8) Epoch 3, batch 7600, loss[loss=0.3237, simple_loss=0.378, pruned_loss=0.1347, over 15307.00 frames. ], tot_loss[loss=0.2805, simple_loss=0.3479, pruned_loss=0.1066, over 3042333.58 frames. ], batch size: 190, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:03:02,836 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 00:03:04,625 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4956, 4.4735, 4.1329, 1.9250, 2.9550, 2.6483, 3.8041, 4.3360], device='cuda:3'), covar=tensor([0.0221, 0.0313, 0.0426, 0.1614, 0.0732, 0.0851, 0.0600, 0.0496], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0121, 0.0159, 0.0148, 0.0142, 0.0135, 0.0149, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 00:03:08,896 INFO [optim.py:368] (3/8) 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 00:03:31,750 INFO [train.py:904] (3/8) Epoch 3, batch 7650, loss[loss=0.3267, simple_loss=0.369, pruned_loss=0.1422, over 11656.00 frames. ], tot_loss[loss=0.283, simple_loss=0.3491, pruned_loss=0.1084, over 3016891.54 frames. ], batch size: 248, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:04:45,971 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 00:04:52,685 INFO [train.py:904] (3/8) Epoch 3, batch 7700, loss[loss=0.2591, simple_loss=0.3331, pruned_loss=0.09257, over 16397.00 frames. ], tot_loss[loss=0.2825, simple_loss=0.3486, pruned_loss=0.1082, over 3038882.04 frames. ], batch size: 146, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:04:58,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6077, 2.4638, 2.1792, 4.0986, 1.8378, 3.6438, 2.3391, 2.3073], device='cuda:3'), covar=tensor([0.0364, 0.0875, 0.0603, 0.0184, 0.1952, 0.0295, 0.1019, 0.1441], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0241, 0.0204, 0.0270, 0.0322, 0.0214, 0.0232, 0.0306], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:05:24,112 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7739, 1.1428, 1.5123, 1.6462, 1.6756, 1.8025, 1.5127, 1.8147], device='cuda:3'), covar=tensor([0.0048, 0.0131, 0.0066, 0.0071, 0.0055, 0.0042, 0.0107, 0.0035], device='cuda:3'), in_proj_covar=tensor([0.0084, 0.0120, 0.0102, 0.0096, 0.0093, 0.0069, 0.0117, 0.0062], device='cuda:3'), out_proj_covar=tensor([1.2737e-04, 1.8712e-04, 1.6411e-04, 1.5321e-04, 1.4275e-04, 1.0497e-04, 1.7853e-04, 9.4072e-05], device='cuda:3') 2023-04-28 00:05:46,975 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 7750, loss[loss=0.3144, simple_loss=0.3621, pruned_loss=0.1333, over 11578.00 frames. ], tot_loss[loss=0.2824, simple_loss=0.3488, pruned_loss=0.108, over 3042155.71 frames. ], batch size: 248, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:06:17,038 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 00:06:40,881 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:07:28,337 INFO [train.py:904] (3/8) Epoch 3, batch 7800, loss[loss=0.2419, simple_loss=0.3249, pruned_loss=0.07942, over 16795.00 frames. ], tot_loss[loss=0.2835, simple_loss=0.3499, pruned_loss=0.1085, over 3054206.35 frames. ], batch size: 102, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:08:22,842 INFO [optim.py:368] (3/8) 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 00:08:45,062 INFO [train.py:904] (3/8) Epoch 3, batch 7850, loss[loss=0.2781, simple_loss=0.3569, pruned_loss=0.09969, over 16868.00 frames. ], tot_loss[loss=0.2836, simple_loss=0.3506, pruned_loss=0.1083, over 3044671.11 frames. ], batch size: 83, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:00,834 INFO [train.py:904] (3/8) Epoch 3, batch 7900, loss[loss=0.2764, simple_loss=0.3492, pruned_loss=0.1018, over 16706.00 frames. ], tot_loss[loss=0.2801, simple_loss=0.3484, pruned_loss=0.1059, over 3063267.38 frames. ], batch size: 134, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:01,923 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3946, 3.3739, 3.1921, 3.2656, 2.9950, 3.3637, 3.2113, 3.1585], device='cuda:3'), covar=tensor([0.0366, 0.0203, 0.0163, 0.0139, 0.0445, 0.0197, 0.0674, 0.0264], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0133, 0.0170, 0.0141, 0.0201, 0.0163, 0.0127, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 00:10:45,267 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:55,796 INFO [optim.py:368] (3/8) 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 00:11:04,775 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8643, 5.0640, 4.7116, 4.8905, 4.4968, 4.3810, 4.5951, 5.1104], device='cuda:3'), covar=tensor([0.0448, 0.0544, 0.0822, 0.0346, 0.0468, 0.0578, 0.0432, 0.0527], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0380, 0.0339, 0.0243, 0.0243, 0.0253, 0.0305, 0.0262], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:11:18,498 INFO [train.py:904] (3/8) Epoch 3, batch 7950, loss[loss=0.2532, simple_loss=0.3354, pruned_loss=0.08547, over 16822.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3486, pruned_loss=0.1065, over 3054089.83 frames. ], batch size: 102, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:11:26,067 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:11:53,368 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7581, 1.1290, 1.3263, 1.6494, 1.6107, 1.7722, 1.3786, 1.7753], device='cuda:3'), covar=tensor([0.0053, 0.0138, 0.0062, 0.0093, 0.0059, 0.0046, 0.0116, 0.0029], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0121, 0.0105, 0.0097, 0.0094, 0.0070, 0.0115, 0.0062], device='cuda:3'), out_proj_covar=tensor([1.2972e-04, 1.8854e-04, 1.6733e-04, 1.5459e-04, 1.4462e-04, 1.0649e-04, 1.7604e-04, 9.3727e-05], device='cuda:3') 2023-04-28 00:12:08,207 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-28 00:12:13,419 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:33,512 INFO [train.py:904] (3/8) Epoch 3, batch 8000, loss[loss=0.2873, simple_loss=0.3513, pruned_loss=0.1117, over 16826.00 frames. ], tot_loss[loss=0.2817, simple_loss=0.3489, pruned_loss=0.1073, over 3047383.56 frames. ], batch size: 116, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:12:56,321 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 4.065e+02 5.202e+02 6.886e+02 1.574e+03, threshold=1.040e+03, percent-clipped=4.0 2023-04-28 00:13:46,419 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:49,305 INFO [train.py:904] (3/8) Epoch 3, batch 8050, loss[loss=0.2845, simple_loss=0.3547, pruned_loss=0.1072, over 15485.00 frames. ], tot_loss[loss=0.2814, simple_loss=0.3486, pruned_loss=0.1071, over 3039565.75 frames. ], batch size: 190, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:14:18,760 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:15:05,404 INFO [train.py:904] (3/8) Epoch 3, batch 8100, loss[loss=0.2582, simple_loss=0.3357, pruned_loss=0.09031, over 16275.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3482, pruned_loss=0.1065, over 3044532.55 frames. ], batch size: 35, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:15:17,592 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 00:15:30,570 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:15:35,123 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4148, 4.0997, 4.1257, 1.8565, 4.3824, 4.2915, 3.2291, 3.2943], device='cuda:3'), covar=tensor([0.0716, 0.0072, 0.0106, 0.1152, 0.0032, 0.0037, 0.0266, 0.0337], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0081, 0.0078, 0.0142, 0.0070, 0.0071, 0.0111, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 00:15:50,615 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0303, 4.0346, 4.5534, 4.5340, 4.5237, 4.1317, 4.1072, 3.9398], device='cuda:3'), covar=tensor([0.0248, 0.0318, 0.0296, 0.0365, 0.0392, 0.0272, 0.0803, 0.0403], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0181, 0.0194, 0.0196, 0.0235, 0.0201, 0.0298, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 00:15:57,042 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 8150, loss[loss=0.2551, simple_loss=0.327, pruned_loss=0.09162, over 16587.00 frames. ], tot_loss[loss=0.2767, simple_loss=0.3448, pruned_loss=0.1043, over 3078609.55 frames. ], batch size: 75, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:17:15,217 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:23,187 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 8200, loss[loss=0.2919, simple_loss=0.3429, pruned_loss=0.1204, over 11641.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3424, pruned_loss=0.1036, over 3075782.99 frames. ], batch size: 247, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:18:31,947 INFO [optim.py:368] (3/8) 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 00:18:50,548 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:55,861 INFO [train.py:904] (3/8) Epoch 3, batch 8250, loss[loss=0.2522, simple_loss=0.3245, pruned_loss=0.08997, over 12303.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3417, pruned_loss=0.1016, over 3052448.32 frames. ], batch size: 246, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:19:00,262 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:19:01,417 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7529, 2.7458, 2.3281, 3.7644, 3.4508, 3.6859, 1.5476, 2.7681], device='cuda:3'), covar=tensor([0.1728, 0.0551, 0.1155, 0.0087, 0.0237, 0.0277, 0.1626, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0133, 0.0157, 0.0072, 0.0137, 0.0142, 0.0150, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-28 00:19:09,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6503, 1.9445, 1.6337, 1.6836, 2.3494, 2.2408, 2.5683, 2.6101], device='cuda:3'), covar=tensor([0.0017, 0.0127, 0.0167, 0.0177, 0.0079, 0.0108, 0.0046, 0.0067], device='cuda:3'), in_proj_covar=tensor([0.0054, 0.0118, 0.0124, 0.0125, 0.0115, 0.0124, 0.0079, 0.0101], device='cuda:3'), out_proj_covar=tensor([7.2321e-05, 1.6536e-04, 1.6933e-04, 1.7457e-04, 1.6360e-04, 1.7426e-04, 1.1147e-04, 1.4339e-04], device='cuda:3') 2023-04-28 00:19:17,422 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7233, 3.5871, 3.3097, 1.7922, 2.7709, 2.1813, 3.1780, 3.5438], device='cuda:3'), covar=tensor([0.0275, 0.0510, 0.0450, 0.1586, 0.0718, 0.0937, 0.0618, 0.0688], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0118, 0.0156, 0.0146, 0.0139, 0.0131, 0.0145, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 00:19:51,757 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:16,983 INFO [train.py:904] (3/8) Epoch 3, batch 8300, loss[loss=0.2248, simple_loss=0.3196, pruned_loss=0.065, over 16704.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3368, pruned_loss=0.09674, over 3030747.59 frames. ], batch size: 76, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:20:33,878 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:20:51,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9555, 3.6769, 3.9362, 4.1482, 4.2472, 3.8058, 4.2590, 4.2177], device='cuda:3'), covar=tensor([0.0703, 0.0840, 0.1270, 0.0551, 0.0416, 0.0825, 0.0388, 0.0425], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0366, 0.0467, 0.0359, 0.0271, 0.0261, 0.0291, 0.0296], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:21:14,648 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 3, batch 8350, loss[loss=0.2725, simple_loss=0.3319, pruned_loss=0.1066, over 12035.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3349, pruned_loss=0.09365, over 3022950.47 frames. ], batch size: 246, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:21:52,242 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 00:22:22,655 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0348, 2.5524, 2.6154, 3.4001, 3.1496, 3.3865, 1.7240, 2.9807], device='cuda:3'), covar=tensor([0.1115, 0.0385, 0.0707, 0.0081, 0.0167, 0.0303, 0.1146, 0.0498], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0132, 0.0158, 0.0070, 0.0133, 0.0141, 0.0151, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-28 00:22:45,727 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 00:22:58,584 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1744, 3.6261, 3.5582, 1.5159, 3.7600, 3.7494, 3.1608, 2.9442], device='cuda:3'), covar=tensor([0.0718, 0.0087, 0.0140, 0.1271, 0.0058, 0.0043, 0.0221, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0080, 0.0076, 0.0143, 0.0070, 0.0070, 0.0109, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 00:23:00,531 INFO [train.py:904] (3/8) Epoch 3, batch 8400, loss[loss=0.2168, simple_loss=0.3085, pruned_loss=0.06256, over 16702.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3308, pruned_loss=0.09007, over 3025126.35 frames. ], batch size: 89, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:44,847 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 00:23:58,229 INFO [optim.py:368] (3/8) 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 00:24:20,120 INFO [train.py:904] (3/8) Epoch 3, batch 8450, loss[loss=0.2303, simple_loss=0.3138, pruned_loss=0.07343, over 17269.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3282, pruned_loss=0.0874, over 3046525.52 frames. ], batch size: 52, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:24:39,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8211, 5.2977, 5.2691, 5.1841, 5.1561, 5.7190, 5.3400, 4.9861], device='cuda:3'), covar=tensor([0.0540, 0.0996, 0.0925, 0.1308, 0.1898, 0.0723, 0.0877, 0.1845], device='cuda:3'), in_proj_covar=tensor([0.0219, 0.0306, 0.0282, 0.0264, 0.0338, 0.0316, 0.0249, 0.0354], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:25:27,691 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 00:25:42,011 INFO [train.py:904] (3/8) Epoch 3, batch 8500, loss[loss=0.204, simple_loss=0.297, pruned_loss=0.05545, over 16802.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3232, pruned_loss=0.08345, over 3057930.55 frames. ], batch size: 102, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:25:54,589 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0116, 3.1921, 2.4955, 4.1495, 3.9744, 3.9619, 1.4578, 2.9533], device='cuda:3'), covar=tensor([0.1204, 0.0398, 0.1049, 0.0072, 0.0163, 0.0288, 0.1343, 0.0651], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0133, 0.0161, 0.0070, 0.0133, 0.0143, 0.0153, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-04-28 00:25:54,611 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8037, 3.6734, 3.4025, 1.6540, 2.8409, 2.2267, 3.0999, 3.5914], device='cuda:3'), covar=tensor([0.0260, 0.0434, 0.0416, 0.1726, 0.0664, 0.0945, 0.0784, 0.0585], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0113, 0.0150, 0.0144, 0.0132, 0.0129, 0.0140, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 00:26:13,899 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4583, 3.3674, 3.3776, 3.0058, 3.3249, 2.1473, 3.1713, 3.1207], device='cuda:3'), covar=tensor([0.0076, 0.0067, 0.0085, 0.0173, 0.0058, 0.1164, 0.0077, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0073, 0.0063, 0.0097, 0.0103, 0.0071, 0.0124, 0.0084, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:26:18,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3097, 3.4406, 3.1214, 3.1840, 2.7249, 3.3200, 3.1335, 3.0949], device='cuda:3'), covar=tensor([0.0521, 0.0286, 0.0313, 0.0232, 0.0910, 0.0298, 0.1058, 0.0359], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0127, 0.0167, 0.0137, 0.0191, 0.0158, 0.0122, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:26:40,692 INFO [optim.py:368] (3/8) 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 00:26:42,918 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7191, 5.0469, 5.1634, 5.1699, 5.2000, 5.6401, 5.4073, 5.1828], device='cuda:3'), covar=tensor([0.0629, 0.1324, 0.1068, 0.1271, 0.1771, 0.0775, 0.0782, 0.1553], device='cuda:3'), in_proj_covar=tensor([0.0219, 0.0305, 0.0284, 0.0264, 0.0337, 0.0314, 0.0249, 0.0351], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:26:49,998 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:00,695 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:05,698 INFO [train.py:904] (3/8) Epoch 3, batch 8550, loss[loss=0.243, simple_loss=0.3265, pruned_loss=0.07981, over 16197.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3199, pruned_loss=0.08213, over 3014789.65 frames. ], batch size: 165, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:27:19,476 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-28 00:27:26,054 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:12,808 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:46,470 INFO [train.py:904] (3/8) Epoch 3, batch 8600, loss[loss=0.1983, simple_loss=0.2892, pruned_loss=0.0537, over 16590.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.321, pruned_loss=0.08103, over 3030114.56 frames. ], batch size: 62, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:29:07,408 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:29:30,868 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:50,786 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:57,553 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.629e+02 3.755e+02 4.530e+02 5.746e+02 8.888e+02, threshold=9.061e+02, percent-clipped=0.0 2023-04-28 00:30:11,173 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:30:25,441 INFO [train.py:904] (3/8) Epoch 3, batch 8650, loss[loss=0.2003, simple_loss=0.2978, pruned_loss=0.05142, over 16905.00 frames. ], tot_loss[loss=0.237, simple_loss=0.318, pruned_loss=0.07799, over 3033942.03 frames. ], batch size: 102, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:30:45,513 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:31:22,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7104, 1.1510, 1.4666, 1.7064, 1.6720, 1.7779, 1.3095, 1.8429], device='cuda:3'), covar=tensor([0.0064, 0.0181, 0.0086, 0.0110, 0.0078, 0.0059, 0.0166, 0.0049], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0123, 0.0110, 0.0100, 0.0096, 0.0071, 0.0117, 0.0064], device='cuda:3'), out_proj_covar=tensor([1.3493e-04, 1.8940e-04, 1.7367e-04, 1.5730e-04, 1.4771e-04, 1.0555e-04, 1.7911e-04, 9.5418e-05], device='cuda:3') 2023-04-28 00:31:30,900 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-28 00:31:44,995 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2723, 4.2381, 4.0620, 4.1238, 3.7405, 4.1486, 4.0295, 3.8780], device='cuda:3'), covar=tensor([0.0339, 0.0257, 0.0193, 0.0130, 0.0590, 0.0286, 0.0311, 0.0350], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0124, 0.0164, 0.0134, 0.0187, 0.0155, 0.0117, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:31:52,872 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:12,024 INFO [train.py:904] (3/8) Epoch 3, batch 8700, loss[loss=0.2552, simple_loss=0.3209, pruned_loss=0.09472, over 12480.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3148, pruned_loss=0.07587, over 3034115.65 frames. ], batch size: 248, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:32:15,231 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:31,867 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7071, 3.6919, 1.7177, 3.7791, 2.3134, 3.6800, 1.9612, 2.8266], device='cuda:3'), covar=tensor([0.0063, 0.0154, 0.1603, 0.0038, 0.0840, 0.0329, 0.1483, 0.0590], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0128, 0.0171, 0.0077, 0.0157, 0.0152, 0.0179, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 00:33:00,326 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 00:33:18,499 INFO [optim.py:368] (3/8) 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 00:33:46,756 INFO [train.py:904] (3/8) Epoch 3, batch 8750, loss[loss=0.2514, simple_loss=0.3368, pruned_loss=0.08299, over 16777.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3136, pruned_loss=0.07477, over 3035426.99 frames. ], batch size: 124, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:34:18,768 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:35:38,460 INFO [train.py:904] (3/8) Epoch 3, batch 8800, loss[loss=0.2619, simple_loss=0.3402, pruned_loss=0.09177, over 15381.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3128, pruned_loss=0.07402, over 3050761.28 frames. ], batch size: 191, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:36:18,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6029, 1.2744, 1.7346, 2.3611, 2.3040, 2.5375, 1.2992, 2.5252], device='cuda:3'), covar=tensor([0.0056, 0.0232, 0.0146, 0.0096, 0.0080, 0.0065, 0.0209, 0.0038], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0124, 0.0111, 0.0100, 0.0097, 0.0071, 0.0119, 0.0063], device='cuda:3'), out_proj_covar=tensor([1.3318e-04, 1.9240e-04, 1.7506e-04, 1.5584e-04, 1.4843e-04, 1.0540e-04, 1.8128e-04, 9.3884e-05], device='cuda:3') 2023-04-28 00:36:52,048 INFO [optim.py:368] (3/8) 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 00:37:01,710 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2902, 3.3472, 2.6432, 2.0404, 2.3946, 1.9534, 3.2740, 3.4153], device='cuda:3'), covar=tensor([0.2182, 0.0678, 0.1223, 0.1274, 0.1745, 0.1451, 0.0465, 0.0412], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0236, 0.0248, 0.0210, 0.0239, 0.0189, 0.0211, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:37:03,504 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:16,994 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 8850, loss[loss=0.2216, simple_loss=0.3149, pruned_loss=0.06417, over 16799.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.314, pruned_loss=0.07314, over 3033801.83 frames. ], batch size: 124, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:38:45,754 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:38:57,572 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:39:06,881 INFO [train.py:904] (3/8) Epoch 3, batch 8900, loss[loss=0.2371, simple_loss=0.3243, pruned_loss=0.07494, over 15269.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3135, pruned_loss=0.07198, over 3034340.39 frames. ], batch size: 191, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:39:40,209 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:35,294 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 3.766e+02 4.506e+02 5.745e+02 1.186e+03, threshold=9.012e+02, percent-clipped=1.0 2023-04-28 00:40:49,336 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8280, 3.1651, 3.0584, 2.0765, 2.9456, 2.9537, 3.0280, 1.7299], device='cuda:3'), covar=tensor([0.0308, 0.0016, 0.0034, 0.0214, 0.0036, 0.0041, 0.0026, 0.0312], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0051, 0.0054, 0.0106, 0.0052, 0.0058, 0.0055, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 00:41:11,853 INFO [train.py:904] (3/8) Epoch 3, batch 8950, loss[loss=0.2225, simple_loss=0.3033, pruned_loss=0.07083, over 15322.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3136, pruned_loss=0.07275, over 3049810.26 frames. ], batch size: 191, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:41:23,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7366, 3.1310, 3.0123, 1.9799, 2.9623, 2.9605, 2.9687, 1.7676], device='cuda:3'), covar=tensor([0.0390, 0.0020, 0.0035, 0.0272, 0.0033, 0.0034, 0.0026, 0.0337], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0052, 0.0055, 0.0107, 0.0052, 0.0058, 0.0056, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 00:42:53,686 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:43:00,670 INFO [train.py:904] (3/8) Epoch 3, batch 9000, loss[loss=0.1856, simple_loss=0.274, pruned_loss=0.04858, over 16514.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3095, pruned_loss=0.07058, over 3062293.92 frames. ], batch size: 68, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:43:00,671 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 00:43:11,838 INFO [train.py:938] (3/8) Epoch 3, validation: loss=0.1886, simple_loss=0.2904, pruned_loss=0.04341, over 944034.00 frames. 2023-04-28 00:43:11,838 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 00:43:57,466 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6224, 3.0459, 2.5664, 4.4023, 4.2995, 4.2101, 1.4253, 3.0512], device='cuda:3'), covar=tensor([0.1475, 0.0481, 0.1036, 0.0064, 0.0146, 0.0232, 0.1400, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0133, 0.0160, 0.0070, 0.0130, 0.0142, 0.0154, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-04-28 00:44:26,933 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 9050, loss[loss=0.2352, simple_loss=0.3018, pruned_loss=0.08436, over 16879.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3108, pruned_loss=0.07195, over 3050148.73 frames. ], batch size: 116, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:45:12,121 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 9100, loss[loss=0.2253, simple_loss=0.3038, pruned_loss=0.07338, over 12139.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3101, pruned_loss=0.07198, over 3071779.26 frames. ], batch size: 248, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:48:08,505 INFO [optim.py:368] (3/8) 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 00:48:09,837 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1340, 3.2582, 1.6021, 3.3149, 2.2486, 3.2351, 1.9345, 2.6525], device='cuda:3'), covar=tensor([0.0093, 0.0196, 0.1603, 0.0043, 0.0765, 0.0359, 0.1315, 0.0581], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0128, 0.0173, 0.0076, 0.0155, 0.0154, 0.0179, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 00:48:40,641 INFO [train.py:904] (3/8) Epoch 3, batch 9150, loss[loss=0.2154, simple_loss=0.3018, pruned_loss=0.06453, over 16567.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3103, pruned_loss=0.07131, over 3061701.79 frames. ], batch size: 148, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:50:09,822 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5952, 4.4678, 4.3488, 4.3055, 4.0065, 4.4874, 4.3786, 4.0379], device='cuda:3'), covar=tensor([0.0246, 0.0188, 0.0146, 0.0117, 0.0544, 0.0150, 0.0211, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0123, 0.0162, 0.0135, 0.0187, 0.0150, 0.0117, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:50:24,877 INFO [train.py:904] (3/8) Epoch 3, batch 9200, loss[loss=0.1854, simple_loss=0.2615, pruned_loss=0.05466, over 11762.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3056, pruned_loss=0.06989, over 3064706.35 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:50:54,907 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:32,421 INFO [optim.py:368] (3/8) 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 00:52:01,382 INFO [train.py:904] (3/8) Epoch 3, batch 9250, loss[loss=0.2161, simple_loss=0.2875, pruned_loss=0.07236, over 12165.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.306, pruned_loss=0.07018, over 3066429.28 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:52:02,454 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:29,837 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:48,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0471, 2.4963, 2.4163, 3.3448, 3.2275, 3.3239, 1.7646, 2.7817], device='cuda:3'), covar=tensor([0.1063, 0.0417, 0.0858, 0.0085, 0.0165, 0.0315, 0.1089, 0.0608], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0130, 0.0158, 0.0068, 0.0129, 0.0140, 0.0152, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-04-28 00:53:31,859 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:50,314 INFO [train.py:904] (3/8) Epoch 3, batch 9300, loss[loss=0.1802, simple_loss=0.2711, pruned_loss=0.04461, over 16831.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3042, pruned_loss=0.06905, over 3068905.78 frames. ], batch size: 90, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:54:15,991 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:54:27,001 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4034, 4.1903, 3.9153, 1.7898, 3.1758, 2.1743, 3.8593, 4.1759], device='cuda:3'), covar=tensor([0.0241, 0.0410, 0.0391, 0.1842, 0.0677, 0.1144, 0.0505, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0110, 0.0152, 0.0144, 0.0134, 0.0130, 0.0140, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 00:54:29,635 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-28 00:55:11,308 INFO [optim.py:368] (3/8) 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 00:55:29,274 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7687, 1.2092, 1.5710, 1.7201, 1.7905, 1.8702, 1.3977, 1.7973], device='cuda:3'), covar=tensor([0.0066, 0.0157, 0.0104, 0.0100, 0.0079, 0.0057, 0.0150, 0.0061], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0124, 0.0111, 0.0101, 0.0098, 0.0071, 0.0117, 0.0064], device='cuda:3'), out_proj_covar=tensor([1.3604e-04, 1.9045e-04, 1.7399e-04, 1.5683e-04, 1.4877e-04, 1.0493e-04, 1.7796e-04, 9.6961e-05], device='cuda:3') 2023-04-28 00:55:29,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3831, 2.9444, 2.6067, 2.1706, 2.1359, 1.9256, 2.9326, 3.0916], device='cuda:3'), covar=tensor([0.1610, 0.0674, 0.1047, 0.1148, 0.1783, 0.1451, 0.0472, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0266, 0.0244, 0.0256, 0.0216, 0.0234, 0.0195, 0.0219, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 00:55:35,075 INFO [train.py:904] (3/8) Epoch 3, batch 9350, loss[loss=0.2533, simple_loss=0.3301, pruned_loss=0.08825, over 16349.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3039, pruned_loss=0.06893, over 3081666.35 frames. ], batch size: 146, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:55:51,602 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:00,148 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:16,960 INFO [train.py:904] (3/8) Epoch 3, batch 9400, loss[loss=0.1891, simple_loss=0.2647, pruned_loss=0.05678, over 12550.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3033, pruned_loss=0.06839, over 3074507.04 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:57:27,902 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:58:01,550 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:58:32,711 INFO [optim.py:368] (3/8) 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 00:58:58,181 INFO [train.py:904] (3/8) Epoch 3, batch 9450, loss[loss=0.2137, simple_loss=0.3032, pruned_loss=0.06209, over 16681.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3057, pruned_loss=0.06874, over 3078743.82 frames. ], batch size: 89, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:37,285 INFO [train.py:904] (3/8) Epoch 3, batch 9500, loss[loss=0.2279, simple_loss=0.3124, pruned_loss=0.07168, over 15411.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3041, pruned_loss=0.06779, over 3070094.92 frames. ], batch size: 192, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:41,478 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:01:17,640 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 01:01:50,889 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 9550, loss[loss=0.2306, simple_loss=0.3178, pruned_loss=0.07174, over 16904.00 frames. ], tot_loss[loss=0.22, simple_loss=0.304, pruned_loss=0.06803, over 3075860.18 frames. ], batch size: 116, lr: 1.89e-02, grad_scale: 4.0 2023-04-28 01:02:49,212 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:26,676 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1275, 4.0398, 3.9151, 3.5329, 3.9162, 1.6464, 3.7245, 3.8215], device='cuda:3'), covar=tensor([0.0060, 0.0050, 0.0084, 0.0182, 0.0057, 0.1489, 0.0074, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0061, 0.0092, 0.0093, 0.0067, 0.0119, 0.0081, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:03:33,122 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5851, 4.5049, 4.3764, 4.3788, 3.9738, 4.4298, 4.4480, 4.1269], device='cuda:3'), covar=tensor([0.0269, 0.0221, 0.0158, 0.0118, 0.0625, 0.0191, 0.0214, 0.0300], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0124, 0.0164, 0.0134, 0.0189, 0.0152, 0.0117, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:03:49,032 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 3, batch 9600, loss[loss=0.2281, simple_loss=0.3142, pruned_loss=0.07104, over 16658.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3061, pruned_loss=0.06917, over 3089498.71 frames. ], batch size: 83, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:04:16,863 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:55,913 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 01:05:18,577 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.589e+02 4.321e+02 5.133e+02 1.002e+03, threshold=8.641e+02, percent-clipped=3.0 2023-04-28 01:05:23,702 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:51,214 INFO [train.py:904] (3/8) Epoch 3, batch 9650, loss[loss=0.2276, simple_loss=0.3154, pruned_loss=0.06995, over 16904.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.308, pruned_loss=0.06953, over 3079934.46 frames. ], batch size: 116, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:06:34,985 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3164, 4.2030, 4.1763, 4.1726, 3.7734, 4.2249, 4.1686, 3.9624], device='cuda:3'), covar=tensor([0.0307, 0.0267, 0.0147, 0.0117, 0.0632, 0.0177, 0.0263, 0.0302], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0124, 0.0164, 0.0133, 0.0189, 0.0152, 0.0116, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:07:41,652 INFO [train.py:904] (3/8) Epoch 3, batch 9700, loss[loss=0.2167, simple_loss=0.305, pruned_loss=0.06422, over 16983.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.307, pruned_loss=0.06924, over 3074720.47 frames. ], batch size: 109, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:08:15,999 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:08:59,979 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 9750, loss[loss=0.2238, simple_loss=0.2938, pruned_loss=0.07695, over 12050.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3055, pruned_loss=0.06909, over 3075994.02 frames. ], batch size: 247, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:09:48,623 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4539, 4.3106, 4.1827, 3.2544, 4.0898, 4.3230, 4.2483, 2.5169], device='cuda:3'), covar=tensor([0.0329, 0.0012, 0.0035, 0.0187, 0.0020, 0.0027, 0.0018, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0053, 0.0056, 0.0108, 0.0053, 0.0060, 0.0055, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:11:02,950 INFO [train.py:904] (3/8) Epoch 3, batch 9800, loss[loss=0.2214, simple_loss=0.3183, pruned_loss=0.0622, over 15342.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3048, pruned_loss=0.06799, over 3094092.29 frames. ], batch size: 190, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:11:26,748 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9321, 4.1711, 3.7124, 2.5307, 3.0512, 2.7036, 4.6836, 4.6949], device='cuda:3'), covar=tensor([0.1806, 0.0486, 0.0812, 0.1034, 0.1612, 0.1002, 0.0243, 0.0251], device='cuda:3'), in_proj_covar=tensor([0.0256, 0.0234, 0.0250, 0.0209, 0.0222, 0.0189, 0.0208, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:11:47,396 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 01:11:48,324 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:12:14,775 INFO [optim.py:368] (3/8) 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 01:12:47,481 INFO [train.py:904] (3/8) Epoch 3, batch 9850, loss[loss=0.2055, simple_loss=0.304, pruned_loss=0.05349, over 17181.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3062, pruned_loss=0.06775, over 3083834.76 frames. ], batch size: 46, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:13:02,425 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:06,698 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:39,080 INFO [train.py:904] (3/8) Epoch 3, batch 9900, loss[loss=0.226, simple_loss=0.2986, pruned_loss=0.07672, over 12259.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3069, pruned_loss=0.06768, over 3089291.24 frames. ], batch size: 249, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:14:54,525 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:59,640 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:16:05,826 INFO [optim.py:368] (3/8) 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 01:16:35,561 INFO [train.py:904] (3/8) Epoch 3, batch 9950, loss[loss=0.2253, simple_loss=0.317, pruned_loss=0.06681, over 15379.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3095, pruned_loss=0.06865, over 3082521.49 frames. ], batch size: 191, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:16:47,150 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:17:23,115 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:37,069 INFO [train.py:904] (3/8) Epoch 3, batch 10000, loss[loss=0.203, simple_loss=0.295, pruned_loss=0.05549, over 15369.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3069, pruned_loss=0.06737, over 3099746.07 frames. ], batch size: 191, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:19:12,455 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:19:55,679 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 3, batch 10050, loss[loss=0.2649, simple_loss=0.3438, pruned_loss=0.09294, over 16691.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3063, pruned_loss=0.067, over 3100301.48 frames. ], batch size: 134, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:20:38,386 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:20:50,750 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:27,280 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3087, 4.6409, 4.3741, 4.3991, 4.0155, 3.9261, 4.1493, 4.6081], device='cuda:3'), covar=tensor([0.0538, 0.0567, 0.0739, 0.0377, 0.0564, 0.0896, 0.0571, 0.0621], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0356, 0.0306, 0.0236, 0.0235, 0.0237, 0.0288, 0.0256], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:21:54,399 INFO [train.py:904] (3/8) Epoch 3, batch 10100, loss[loss=0.1988, simple_loss=0.2895, pruned_loss=0.05405, over 15388.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3068, pruned_loss=0.06771, over 3079357.83 frames. ], batch size: 192, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:22:24,580 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9240, 4.0011, 3.3328, 2.5462, 3.0387, 2.4267, 4.2491, 4.4478], device='cuda:3'), covar=tensor([0.1973, 0.0662, 0.1035, 0.1148, 0.1671, 0.1239, 0.0337, 0.0382], device='cuda:3'), in_proj_covar=tensor([0.0261, 0.0239, 0.0250, 0.0211, 0.0223, 0.0195, 0.0212, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:22:37,216 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:23:00,421 INFO [optim.py:368] (3/8) 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 01:23:02,831 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3894, 4.3095, 4.2368, 3.6139, 4.1418, 1.6469, 3.9712, 4.2098], device='cuda:3'), covar=tensor([0.0055, 0.0046, 0.0063, 0.0189, 0.0053, 0.1396, 0.0065, 0.0100], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0060, 0.0091, 0.0089, 0.0066, 0.0118, 0.0080, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:23:38,627 INFO [train.py:904] (3/8) Epoch 4, batch 0, loss[loss=0.3894, simple_loss=0.3959, pruned_loss=0.1914, over 16673.00 frames. ], tot_loss[loss=0.3894, simple_loss=0.3959, pruned_loss=0.1914, over 16673.00 frames. ], batch size: 134, lr: 1.75e-02, grad_scale: 8.0 2023-04-28 01:23:38,627 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 01:23:46,516 INFO [train.py:938] (3/8) Epoch 4, validation: loss=0.188, simple_loss=0.2904, pruned_loss=0.04284, over 944034.00 frames. 2023-04-28 01:23:46,517 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 01:23:56,982 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4013, 4.0596, 3.5691, 1.8466, 2.8896, 2.2840, 3.5418, 3.9980], device='cuda:3'), covar=tensor([0.0222, 0.0502, 0.0498, 0.1574, 0.0701, 0.1082, 0.0641, 0.0572], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0109, 0.0152, 0.0143, 0.0133, 0.0129, 0.0136, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 01:23:57,992 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:29,165 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:56,025 INFO [train.py:904] (3/8) Epoch 4, batch 50, loss[loss=0.225, simple_loss=0.3064, pruned_loss=0.07181, over 17267.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3287, pruned_loss=0.1013, over 750226.38 frames. ], batch size: 52, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:25:00,488 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-28 01:25:02,247 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:04,729 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0091, 3.9258, 2.9012, 5.3178, 5.1478, 4.5737, 1.8157, 3.2857], device='cuda:3'), covar=tensor([0.1192, 0.0392, 0.1039, 0.0058, 0.0185, 0.0264, 0.1196, 0.0638], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0135, 0.0164, 0.0069, 0.0137, 0.0145, 0.0153, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-04-28 01:25:09,212 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 01:25:11,274 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8094, 4.5438, 4.3742, 1.9459, 4.7207, 4.7455, 3.4131, 3.8966], device='cuda:3'), covar=tensor([0.0701, 0.0067, 0.0152, 0.1166, 0.0033, 0.0034, 0.0279, 0.0259], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0082, 0.0080, 0.0146, 0.0072, 0.0073, 0.0112, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 01:25:41,526 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3058, 4.3604, 4.9294, 4.8988, 4.8428, 4.4014, 4.4798, 4.3048], device='cuda:3'), covar=tensor([0.0235, 0.0329, 0.0275, 0.0329, 0.0380, 0.0279, 0.0658, 0.0315], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0182, 0.0187, 0.0190, 0.0220, 0.0199, 0.0283, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 01:25:49,876 INFO [optim.py:368] (3/8) 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 01:26:02,163 INFO [train.py:904] (3/8) Epoch 4, batch 100, loss[loss=0.2271, simple_loss=0.2905, pruned_loss=0.08186, over 15983.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3236, pruned_loss=0.0971, over 1320222.44 frames. ], batch size: 35, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:26:23,740 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 150, loss[loss=0.2444, simple_loss=0.3056, pruned_loss=0.09155, over 16706.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3228, pruned_loss=0.09559, over 1762068.68 frames. ], batch size: 89, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:27:41,039 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9017, 4.2533, 3.0318, 2.6273, 3.1931, 2.1193, 4.4943, 4.4143], device='cuda:3'), covar=tensor([0.1807, 0.0593, 0.1225, 0.1175, 0.1944, 0.1446, 0.0324, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0245, 0.0258, 0.0219, 0.0259, 0.0198, 0.0219, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:28:04,890 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 3.830e+02 4.808e+02 5.813e+02 1.621e+03, threshold=9.616e+02, percent-clipped=3.0 2023-04-28 01:28:19,172 INFO [train.py:904] (3/8) Epoch 4, batch 200, loss[loss=0.2381, simple_loss=0.3204, pruned_loss=0.07788, over 17092.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3197, pruned_loss=0.09249, over 2110419.61 frames. ], batch size: 53, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:28:24,042 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:26,980 INFO [train.py:904] (3/8) Epoch 4, batch 250, loss[loss=0.2505, simple_loss=0.3422, pruned_loss=0.07935, over 17046.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3177, pruned_loss=0.09143, over 2382392.63 frames. ], batch size: 53, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:29:48,574 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:48,723 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:29:49,932 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3330, 5.1869, 5.0761, 4.4309, 5.0629, 1.9176, 4.7530, 5.2752], device='cuda:3'), covar=tensor([0.0052, 0.0049, 0.0064, 0.0253, 0.0047, 0.1419, 0.0082, 0.0071], device='cuda:3'), in_proj_covar=tensor([0.0076, 0.0067, 0.0104, 0.0106, 0.0074, 0.0125, 0.0089, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:30:15,257 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7901, 4.0492, 3.3600, 2.5205, 3.0087, 2.5559, 4.4536, 4.3548], device='cuda:3'), covar=tensor([0.2127, 0.0669, 0.1057, 0.1254, 0.2141, 0.1166, 0.0311, 0.0516], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0244, 0.0258, 0.0220, 0.0269, 0.0199, 0.0221, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:30:16,701 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 01:30:18,678 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5504, 3.2726, 2.8300, 1.8462, 2.5487, 2.1141, 3.1538, 3.2297], device='cuda:3'), covar=tensor([0.0252, 0.0568, 0.0536, 0.1490, 0.0709, 0.0935, 0.0560, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0118, 0.0153, 0.0144, 0.0134, 0.0129, 0.0138, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 01:30:21,642 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 300, loss[loss=0.2635, simple_loss=0.3157, pruned_loss=0.1057, over 16904.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3145, pruned_loss=0.08954, over 2582522.64 frames. ], batch size: 116, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:30:47,952 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5121, 4.4446, 4.3357, 4.3159, 3.9323, 4.3384, 4.1923, 4.1422], device='cuda:3'), covar=tensor([0.0331, 0.0225, 0.0146, 0.0133, 0.0620, 0.0223, 0.0290, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0148, 0.0190, 0.0157, 0.0223, 0.0178, 0.0134, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:31:16,614 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:31:43,555 INFO [train.py:904] (3/8) Epoch 4, batch 350, loss[loss=0.2568, simple_loss=0.3215, pruned_loss=0.09608, over 15456.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3124, pruned_loss=0.08848, over 2739304.96 frames. ], batch size: 190, lr: 1.74e-02, grad_scale: 1.0 2023-04-28 01:31:48,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-04-28 01:32:20,667 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 4, batch 400, loss[loss=0.2175, simple_loss=0.3085, pruned_loss=0.06324, over 17025.00 frames. ], tot_loss[loss=0.2414, simple_loss=0.3097, pruned_loss=0.08653, over 2872481.75 frames. ], batch size: 50, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:33:11,824 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:33:32,376 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9063, 5.5545, 5.5378, 5.4031, 5.3590, 5.8755, 5.6492, 5.3071], device='cuda:3'), covar=tensor([0.0668, 0.1273, 0.1062, 0.1491, 0.2306, 0.1028, 0.0896, 0.2053], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0354, 0.0324, 0.0300, 0.0396, 0.0366, 0.0280, 0.0403], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:33:40,194 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-28 01:34:01,538 INFO [train.py:904] (3/8) Epoch 4, batch 450, loss[loss=0.228, simple_loss=0.2907, pruned_loss=0.08264, over 16922.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3072, pruned_loss=0.08466, over 2964353.32 frames. ], batch size: 96, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:34:05,025 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:35,347 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:45,841 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 4, batch 500, loss[loss=0.2338, simple_loss=0.2957, pruned_loss=0.08592, over 16897.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3058, pruned_loss=0.08271, over 3046091.06 frames. ], batch size: 96, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:35:28,274 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:35,689 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0056, 4.6950, 4.9308, 5.2621, 5.3358, 4.6539, 5.3471, 5.3066], device='cuda:3'), covar=tensor([0.0574, 0.0578, 0.1113, 0.0356, 0.0365, 0.0463, 0.0345, 0.0300], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0408, 0.0543, 0.0416, 0.0314, 0.0290, 0.0325, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:35:59,037 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:09,062 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 550, loss[loss=0.2858, simple_loss=0.3273, pruned_loss=0.1221, over 16424.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3043, pruned_loss=0.08163, over 3116028.85 frames. ], batch size: 146, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:36:33,162 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:36:33,296 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:36:39,303 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:47,863 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8857, 5.5455, 5.5803, 5.4119, 5.5569, 6.0340, 5.7624, 5.3632], device='cuda:3'), covar=tensor([0.0663, 0.1510, 0.1119, 0.1512, 0.2157, 0.0811, 0.0852, 0.1984], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0360, 0.0333, 0.0302, 0.0406, 0.0361, 0.0285, 0.0409], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:37:13,481 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 600, loss[loss=0.1824, simple_loss=0.2596, pruned_loss=0.05256, over 16297.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3044, pruned_loss=0.08215, over 3150939.33 frames. ], batch size: 36, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:37:38,200 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 01:37:46,200 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:58,202 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:38:36,813 INFO [train.py:904] (3/8) Epoch 4, batch 650, loss[loss=0.2159, simple_loss=0.2727, pruned_loss=0.07954, over 16811.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3025, pruned_loss=0.08126, over 3181019.51 frames. ], batch size: 102, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:38:37,117 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9711, 5.6346, 5.6414, 5.5446, 5.5730, 6.0516, 5.7516, 5.5022], device='cuda:3'), covar=tensor([0.0630, 0.1329, 0.1116, 0.1659, 0.2266, 0.0785, 0.1040, 0.2358], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0352, 0.0326, 0.0296, 0.0395, 0.0353, 0.0281, 0.0394], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:39:30,534 INFO [optim.py:368] (3/8) 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 01:39:36,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2685, 1.5477, 2.3336, 2.9678, 2.9207, 3.3775, 1.9297, 3.1865], device='cuda:3'), covar=tensor([0.0052, 0.0193, 0.0125, 0.0080, 0.0081, 0.0073, 0.0165, 0.0053], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0130, 0.0116, 0.0110, 0.0105, 0.0081, 0.0122, 0.0068], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 01:39:45,330 INFO [train.py:904] (3/8) Epoch 4, batch 700, loss[loss=0.2444, simple_loss=0.3035, pruned_loss=0.09264, over 16708.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3007, pruned_loss=0.07979, over 3213135.76 frames. ], batch size: 134, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:40:46,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3929, 4.7253, 4.3852, 4.5463, 4.2383, 4.1064, 4.2710, 4.7023], device='cuda:3'), covar=tensor([0.0576, 0.0722, 0.0952, 0.0404, 0.0573, 0.0997, 0.0616, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0321, 0.0448, 0.0385, 0.0281, 0.0288, 0.0289, 0.0353, 0.0311], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:40:48,058 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-28 01:40:53,738 INFO [train.py:904] (3/8) Epoch 4, batch 750, loss[loss=0.2014, simple_loss=0.2867, pruned_loss=0.05808, over 17127.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3001, pruned_loss=0.0788, over 3245168.45 frames. ], batch size: 47, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:41:31,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7735, 4.1609, 4.0960, 2.0871, 4.2110, 4.1814, 3.2223, 3.3567], device='cuda:3'), covar=tensor([0.0587, 0.0062, 0.0151, 0.1058, 0.0053, 0.0053, 0.0314, 0.0279], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0141, 0.0073, 0.0074, 0.0112, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 01:41:48,308 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 800, loss[loss=0.2267, simple_loss=0.2932, pruned_loss=0.08008, over 16674.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2989, pruned_loss=0.07815, over 3265538.56 frames. ], batch size: 76, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:42:14,287 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:21,458 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 01:42:44,497 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:55,443 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:10,104 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-28 01:43:11,602 INFO [train.py:904] (3/8) Epoch 4, batch 850, loss[loss=0.2245, simple_loss=0.2886, pruned_loss=0.08019, over 16666.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.2979, pruned_loss=0.07766, over 3279409.24 frames. ], batch size: 134, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:43:24,681 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:30,634 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-28 01:44:07,363 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.752e+02 3.699e+02 4.562e+02 5.706e+02 1.340e+03, threshold=9.123e+02, percent-clipped=5.0 2023-04-28 01:44:19,710 INFO [train.py:904] (3/8) Epoch 4, batch 900, loss[loss=0.2115, simple_loss=0.2971, pruned_loss=0.06297, over 16716.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.2976, pruned_loss=0.07696, over 3294928.48 frames. ], batch size: 57, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:44:32,829 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:42,741 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:45:12,488 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8603, 1.6173, 2.1472, 2.7019, 2.6293, 2.6837, 1.6985, 2.8138], device='cuda:3'), covar=tensor([0.0043, 0.0175, 0.0122, 0.0076, 0.0063, 0.0082, 0.0156, 0.0031], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0129, 0.0115, 0.0110, 0.0104, 0.0081, 0.0122, 0.0068], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 01:45:31,310 INFO [train.py:904] (3/8) Epoch 4, batch 950, loss[loss=0.2378, simple_loss=0.2968, pruned_loss=0.0894, over 16405.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.2979, pruned_loss=0.07679, over 3298276.41 frames. ], batch size: 146, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:45:41,852 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:26,066 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.766e+02 3.401e+02 4.040e+02 4.698e+02 9.697e+02, threshold=8.080e+02, percent-clipped=2.0 2023-04-28 01:46:38,672 INFO [train.py:904] (3/8) Epoch 4, batch 1000, loss[loss=0.1867, simple_loss=0.2653, pruned_loss=0.05408, over 17017.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.296, pruned_loss=0.07661, over 3312366.16 frames. ], batch size: 41, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:46:49,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7408, 5.0309, 4.6144, 4.8259, 4.4605, 4.3947, 4.5924, 5.0609], device='cuda:3'), covar=tensor([0.0545, 0.0740, 0.1073, 0.0412, 0.0606, 0.0676, 0.0628, 0.0722], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0438, 0.0373, 0.0274, 0.0279, 0.0279, 0.0346, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:47:06,315 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:18,287 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8814, 4.8047, 4.6627, 4.6330, 4.1731, 4.6565, 4.6391, 4.3400], device='cuda:3'), covar=tensor([0.0404, 0.0233, 0.0203, 0.0166, 0.0842, 0.0265, 0.0255, 0.0393], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0157, 0.0206, 0.0171, 0.0244, 0.0189, 0.0148, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:47:36,195 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:47:48,996 INFO [train.py:904] (3/8) Epoch 4, batch 1050, loss[loss=0.1879, simple_loss=0.2718, pruned_loss=0.05202, over 17169.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2949, pruned_loss=0.07523, over 3322166.69 frames. ], batch size: 46, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:48:30,566 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0508, 5.0132, 4.8279, 4.7961, 4.3322, 4.8666, 4.8182, 4.4960], device='cuda:3'), covar=tensor([0.0335, 0.0174, 0.0177, 0.0138, 0.0797, 0.0233, 0.0231, 0.0381], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0156, 0.0206, 0.0171, 0.0242, 0.0189, 0.0147, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:48:45,340 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 3.437e+02 4.188e+02 5.032e+02 1.579e+03, threshold=8.377e+02, percent-clipped=2.0 2023-04-28 01:49:00,844 INFO [train.py:904] (3/8) Epoch 4, batch 1100, loss[loss=0.2351, simple_loss=0.307, pruned_loss=0.08158, over 17228.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2946, pruned_loss=0.07528, over 3309981.69 frames. ], batch size: 52, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:49:02,376 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:49:11,008 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:41,918 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:55,715 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 1150, loss[loss=0.2173, simple_loss=0.2823, pruned_loss=0.07614, over 16830.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2949, pruned_loss=0.07472, over 3306458.74 frames. ], batch size: 102, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:50:18,982 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:49,048 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:01,009 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 01:51:04,258 INFO [optim.py:368] (3/8) 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 01:51:19,068 INFO [train.py:904] (3/8) Epoch 4, batch 1200, loss[loss=0.2751, simple_loss=0.319, pruned_loss=0.1156, over 16512.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.294, pruned_loss=0.07456, over 3315453.87 frames. ], batch size: 75, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:51:25,266 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3263, 4.0323, 3.8623, 1.7765, 3.9941, 3.9917, 3.1246, 3.1179], device='cuda:3'), covar=tensor([0.0707, 0.0057, 0.0117, 0.1183, 0.0059, 0.0056, 0.0265, 0.0304], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0081, 0.0081, 0.0145, 0.0074, 0.0076, 0.0113, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 01:51:41,053 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:52:27,377 INFO [train.py:904] (3/8) Epoch 4, batch 1250, loss[loss=0.2051, simple_loss=0.276, pruned_loss=0.06713, over 16757.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2942, pruned_loss=0.07543, over 3318177.88 frames. ], batch size: 39, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:52:31,896 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:47,962 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:53:22,524 INFO [optim.py:368] (3/8) 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 01:53:32,910 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5337, 4.2304, 3.5107, 1.8111, 2.8921, 2.4705, 3.6950, 4.0227], device='cuda:3'), covar=tensor([0.0235, 0.0446, 0.0535, 0.1532, 0.0685, 0.0899, 0.0629, 0.0643], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0127, 0.0152, 0.0142, 0.0133, 0.0127, 0.0140, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 01:53:35,322 INFO [train.py:904] (3/8) Epoch 4, batch 1300, loss[loss=0.223, simple_loss=0.3083, pruned_loss=0.06886, over 17182.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2931, pruned_loss=0.07432, over 3325559.37 frames. ], batch size: 46, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:53:54,699 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:53:54,821 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:26,665 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 1350, loss[loss=0.2309, simple_loss=0.3115, pruned_loss=0.07517, over 16667.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2932, pruned_loss=0.07433, over 3320237.02 frames. ], batch size: 57, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:14,625 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:19,669 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0597, 3.3007, 3.1810, 1.6364, 3.4403, 3.3389, 2.7747, 2.6883], device='cuda:3'), covar=tensor([0.0865, 0.0106, 0.0216, 0.1231, 0.0068, 0.0088, 0.0383, 0.0390], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0082, 0.0081, 0.0146, 0.0074, 0.0078, 0.0111, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 01:55:39,665 INFO [optim.py:368] (3/8) 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 01:55:48,226 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 01:55:50,732 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:52,765 INFO [train.py:904] (3/8) Epoch 4, batch 1400, loss[loss=0.1953, simple_loss=0.2692, pruned_loss=0.06066, over 16502.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.293, pruned_loss=0.07422, over 3327242.08 frames. ], batch size: 68, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:57,018 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2287, 4.0702, 4.1457, 4.1612, 4.1167, 4.6848, 4.3162, 3.9655], device='cuda:3'), covar=tensor([0.1329, 0.1574, 0.1232, 0.1796, 0.2371, 0.0954, 0.1152, 0.2325], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0365, 0.0336, 0.0310, 0.0412, 0.0366, 0.0288, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:56:09,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5109, 4.5472, 4.3474, 4.4283, 3.7200, 4.4599, 4.4065, 4.0075], device='cuda:3'), covar=tensor([0.0512, 0.0304, 0.0264, 0.0185, 0.1045, 0.0291, 0.0397, 0.0417], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0161, 0.0210, 0.0175, 0.0248, 0.0193, 0.0151, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 01:56:32,185 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:40,462 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:02,739 INFO [train.py:904] (3/8) Epoch 4, batch 1450, loss[loss=0.2359, simple_loss=0.2936, pruned_loss=0.08906, over 16696.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.2926, pruned_loss=0.07497, over 3318035.37 frames. ], batch size: 76, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:57:22,438 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:56,639 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:00,366 INFO [optim.py:368] (3/8) 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 01:58:12,722 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 1500, loss[loss=0.2287, simple_loss=0.3124, pruned_loss=0.07251, over 17019.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2932, pruned_loss=0.07507, over 3319314.71 frames. ], batch size: 55, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:58:13,906 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9965, 4.6867, 4.8945, 5.2592, 5.3646, 4.5288, 5.3882, 5.2406], device='cuda:3'), covar=tensor([0.0732, 0.0738, 0.1372, 0.0440, 0.0429, 0.0515, 0.0383, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0424, 0.0574, 0.0437, 0.0331, 0.0316, 0.0346, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 01:58:47,470 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 1550, loss[loss=0.261, simple_loss=0.3032, pruned_loss=0.1094, over 16885.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2958, pruned_loss=0.07685, over 3318118.55 frames. ], batch size: 90, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:59:39,778 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-28 01:59:40,644 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.322e+02 3.491e+02 4.515e+02 5.293e+02 8.780e+02, threshold=9.030e+02, percent-clipped=2.0 2023-04-28 02:00:34,386 INFO [train.py:904] (3/8) Epoch 4, batch 1600, loss[loss=0.2336, simple_loss=0.3253, pruned_loss=0.07099, over 17098.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.2971, pruned_loss=0.07723, over 3315015.32 frames. ], batch size: 48, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:00:46,073 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:52,720 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:33,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7889, 2.5546, 2.3510, 4.3237, 1.8511, 3.5881, 2.1719, 2.3794], device='cuda:3'), covar=tensor([0.0422, 0.1196, 0.0662, 0.0237, 0.2505, 0.0532, 0.1462, 0.1953], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0266, 0.0220, 0.0283, 0.0327, 0.0244, 0.0244, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:01:41,676 INFO [train.py:904] (3/8) Epoch 4, batch 1650, loss[loss=0.224, simple_loss=0.2884, pruned_loss=0.0798, over 16459.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.2977, pruned_loss=0.07707, over 3321007.68 frames. ], batch size: 146, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:01:58,771 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:37,463 INFO [optim.py:368] (3/8) 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 02:02:40,714 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:44,963 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:02:50,054 INFO [train.py:904] (3/8) Epoch 4, batch 1700, loss[loss=0.206, simple_loss=0.2899, pruned_loss=0.06106, over 17173.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3006, pruned_loss=0.07834, over 3321014.72 frames. ], batch size: 46, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:03:29,667 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:53,202 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:04:01,130 INFO [train.py:904] (3/8) Epoch 4, batch 1750, loss[loss=0.2459, simple_loss=0.3232, pruned_loss=0.08433, over 17073.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3023, pruned_loss=0.07868, over 3314325.15 frames. ], batch size: 53, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:04:05,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9376, 3.2109, 2.6788, 4.6028, 4.4308, 4.3441, 1.8431, 3.0445], device='cuda:3'), covar=tensor([0.1111, 0.0375, 0.0924, 0.0046, 0.0164, 0.0210, 0.1093, 0.0578], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0135, 0.0162, 0.0075, 0.0167, 0.0155, 0.0153, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 02:04:48,292 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:58,762 INFO [optim.py:368] (3/8) 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 02:05:09,617 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8591, 1.7172, 2.3761, 2.7610, 2.8180, 2.7811, 1.7245, 3.0410], device='cuda:3'), covar=tensor([0.0049, 0.0159, 0.0114, 0.0081, 0.0059, 0.0074, 0.0158, 0.0031], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0127, 0.0117, 0.0112, 0.0107, 0.0080, 0.0124, 0.0070], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 02:05:11,464 INFO [train.py:904] (3/8) Epoch 4, batch 1800, loss[loss=0.251, simple_loss=0.3126, pruned_loss=0.09471, over 16730.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.3042, pruned_loss=0.07961, over 3303119.57 frames. ], batch size: 83, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:05:39,580 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 1850, loss[loss=0.239, simple_loss=0.3113, pruned_loss=0.08329, over 16281.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3046, pruned_loss=0.07932, over 3292860.22 frames. ], batch size: 165, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:06:26,553 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:28,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2585, 4.1580, 4.0861, 4.1135, 3.7704, 4.1952, 3.8633, 3.8970], device='cuda:3'), covar=tensor([0.0373, 0.0288, 0.0193, 0.0166, 0.0776, 0.0222, 0.0506, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0162, 0.0207, 0.0174, 0.0247, 0.0194, 0.0150, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 02:07:13,036 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 02:07:18,152 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 1900, loss[loss=0.2148, simple_loss=0.284, pruned_loss=0.07285, over 16763.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3024, pruned_loss=0.07812, over 3299972.45 frames. ], batch size: 124, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:07:43,066 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:01,508 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0629, 3.7383, 3.1247, 1.8077, 2.6160, 2.0975, 3.5053, 3.6740], device='cuda:3'), covar=tensor([0.0263, 0.0445, 0.0575, 0.1453, 0.0747, 0.0930, 0.0560, 0.0568], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0130, 0.0154, 0.0142, 0.0135, 0.0127, 0.0141, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:08:04,473 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 02:08:41,339 INFO [train.py:904] (3/8) Epoch 4, batch 1950, loss[loss=0.2283, simple_loss=0.3051, pruned_loss=0.07575, over 16671.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3024, pruned_loss=0.07763, over 3309620.45 frames. ], batch size: 57, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:08:50,794 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:37,979 INFO [optim.py:368] (3/8) 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 02:09:39,328 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:47,992 INFO [train.py:904] (3/8) Epoch 4, batch 2000, loss[loss=0.2093, simple_loss=0.2913, pruned_loss=0.06369, over 17204.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.302, pruned_loss=0.0772, over 3310129.91 frames. ], batch size: 46, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:10:05,342 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 02:10:28,246 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:45,040 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 2050, loss[loss=0.2305, simple_loss=0.3136, pruned_loss=0.07368, over 17108.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3025, pruned_loss=0.07821, over 3316036.37 frames. ], batch size: 49, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:11:33,295 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-28 02:11:34,835 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:44,256 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 4, batch 2100, loss[loss=0.2245, simple_loss=0.3093, pruned_loss=0.06984, over 17121.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3029, pruned_loss=0.07826, over 3320894.24 frames. ], batch size: 49, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:12:36,400 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:47,249 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4425, 4.3553, 4.3084, 1.8876, 4.5726, 4.6844, 3.2862, 3.4318], device='cuda:3'), covar=tensor([0.1025, 0.0096, 0.0191, 0.1290, 0.0100, 0.0040, 0.0288, 0.0398], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0082, 0.0081, 0.0143, 0.0074, 0.0077, 0.0112, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 02:12:50,998 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:51,124 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0121, 3.9066, 3.8082, 3.8375, 3.5557, 3.9181, 3.6109, 3.6635], device='cuda:3'), covar=tensor([0.0410, 0.0335, 0.0203, 0.0160, 0.0635, 0.0257, 0.0733, 0.0350], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0164, 0.0209, 0.0175, 0.0247, 0.0196, 0.0148, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 02:13:17,653 INFO [train.py:904] (3/8) Epoch 4, batch 2150, loss[loss=0.2461, simple_loss=0.3089, pruned_loss=0.0916, over 16902.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3046, pruned_loss=0.07926, over 3316282.48 frames. ], batch size: 116, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:13:24,605 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:42,050 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:47,098 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9061, 2.6140, 2.3858, 4.4846, 1.9254, 3.9919, 2.3819, 2.3988], device='cuda:3'), covar=tensor([0.0370, 0.1161, 0.0667, 0.0203, 0.2297, 0.0438, 0.1279, 0.1891], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0271, 0.0223, 0.0285, 0.0335, 0.0251, 0.0249, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:14:15,066 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 2200, loss[loss=0.2036, simple_loss=0.2886, pruned_loss=0.05933, over 16813.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3041, pruned_loss=0.07875, over 3317791.19 frames. ], batch size: 42, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:14:30,432 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:15:02,686 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7036, 3.5980, 2.7698, 2.2907, 2.7451, 2.1396, 3.6264, 3.8177], device='cuda:3'), covar=tensor([0.1775, 0.0570, 0.1072, 0.1295, 0.1893, 0.1323, 0.0381, 0.0560], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0246, 0.0255, 0.0226, 0.0294, 0.0194, 0.0224, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:15:34,029 INFO [train.py:904] (3/8) Epoch 4, batch 2250, loss[loss=0.223, simple_loss=0.2973, pruned_loss=0.07433, over 16510.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3054, pruned_loss=0.07957, over 3310752.90 frames. ], batch size: 68, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:16:32,021 INFO [optim.py:368] (3/8) 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 02:16:42,204 INFO [train.py:904] (3/8) Epoch 4, batch 2300, loss[loss=0.2417, simple_loss=0.3082, pruned_loss=0.08761, over 16773.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3065, pruned_loss=0.08023, over 3309268.49 frames. ], batch size: 83, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:17:51,369 INFO [train.py:904] (3/8) Epoch 4, batch 2350, loss[loss=0.2119, simple_loss=0.292, pruned_loss=0.06594, over 17236.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3064, pruned_loss=0.08006, over 3322289.07 frames. ], batch size: 43, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:18:17,891 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 02:18:48,982 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 2400, loss[loss=0.2302, simple_loss=0.3143, pruned_loss=0.07305, over 17274.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3086, pruned_loss=0.08185, over 3321066.89 frames. ], batch size: 52, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:20:06,099 INFO [train.py:904] (3/8) Epoch 4, batch 2450, loss[loss=0.2176, simple_loss=0.295, pruned_loss=0.0701, over 16835.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3074, pruned_loss=0.07969, over 3329681.74 frames. ], batch size: 42, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:20:21,833 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0592, 4.3289, 4.3290, 1.9470, 4.5678, 4.6726, 3.3475, 3.7216], device='cuda:3'), covar=tensor([0.0533, 0.0087, 0.0173, 0.1090, 0.0084, 0.0042, 0.0252, 0.0262], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0083, 0.0081, 0.0144, 0.0074, 0.0077, 0.0114, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 02:21:03,694 INFO [optim.py:368] (3/8) 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 02:21:12,638 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0165, 4.6869, 4.6588, 2.2927, 4.9640, 4.9739, 3.4866, 3.8901], device='cuda:3'), covar=tensor([0.0639, 0.0061, 0.0108, 0.1044, 0.0035, 0.0034, 0.0259, 0.0272], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0084, 0.0081, 0.0146, 0.0074, 0.0079, 0.0115, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 02:21:13,283 INFO [train.py:904] (3/8) Epoch 4, batch 2500, loss[loss=0.1893, simple_loss=0.2652, pruned_loss=0.05669, over 16810.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3076, pruned_loss=0.07954, over 3322290.28 frames. ], batch size: 39, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:29,438 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 02:21:34,387 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:21,140 INFO [train.py:904] (3/8) Epoch 4, batch 2550, loss[loss=0.2421, simple_loss=0.3029, pruned_loss=0.09068, over 16873.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.308, pruned_loss=0.0801, over 3317664.38 frames. ], batch size: 116, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:22:57,663 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 4, batch 2600, loss[loss=0.264, simple_loss=0.3274, pruned_loss=0.1003, over 15327.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3086, pruned_loss=0.08034, over 3311476.88 frames. ], batch size: 190, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:23:41,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9021, 4.1401, 1.8782, 4.5075, 2.5075, 4.4579, 1.7199, 2.9425], device='cuda:3'), covar=tensor([0.0124, 0.0264, 0.1610, 0.0036, 0.0758, 0.0250, 0.1882, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0149, 0.0170, 0.0084, 0.0158, 0.0179, 0.0182, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:23:55,328 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:24:02,326 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1349, 2.4540, 2.4109, 4.9290, 1.7717, 4.1319, 2.4156, 2.4073], device='cuda:3'), covar=tensor([0.0395, 0.1371, 0.0712, 0.0163, 0.2568, 0.0430, 0.1329, 0.2085], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0273, 0.0227, 0.0288, 0.0338, 0.0255, 0.0251, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:24:39,064 INFO [train.py:904] (3/8) Epoch 4, batch 2650, loss[loss=0.2729, simple_loss=0.3366, pruned_loss=0.1046, over 16917.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3082, pruned_loss=0.0795, over 3313620.30 frames. ], batch size: 109, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:19,565 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:25:46,692 INFO [train.py:904] (3/8) Epoch 4, batch 2700, loss[loss=0.2046, simple_loss=0.2925, pruned_loss=0.05835, over 17166.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.308, pruned_loss=0.07861, over 3321500.99 frames. ], batch size: 46, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:54,273 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:26:56,821 INFO [train.py:904] (3/8) Epoch 4, batch 2750, loss[loss=0.2236, simple_loss=0.2949, pruned_loss=0.07614, over 16828.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3074, pruned_loss=0.07744, over 3332994.52 frames. ], batch size: 102, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:27:10,591 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2758, 5.0468, 5.1857, 3.8270, 5.0945, 1.8954, 4.8203, 5.2079], device='cuda:3'), covar=tensor([0.0081, 0.0087, 0.0085, 0.0479, 0.0074, 0.1827, 0.0096, 0.0149], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0079, 0.0118, 0.0127, 0.0087, 0.0128, 0.0104, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 02:27:16,289 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.476e+02 3.299e+02 3.913e+02 4.982e+02 9.564e+02, threshold=7.826e+02, percent-clipped=3.0 2023-04-28 02:28:04,360 INFO [train.py:904] (3/8) Epoch 4, batch 2800, loss[loss=0.2123, simple_loss=0.2989, pruned_loss=0.06283, over 16728.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3064, pruned_loss=0.07603, over 3327647.14 frames. ], batch size: 62, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:28:14,483 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-04-28 02:29:10,594 INFO [train.py:904] (3/8) Epoch 4, batch 2850, loss[loss=0.2377, simple_loss=0.324, pruned_loss=0.07569, over 17030.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3052, pruned_loss=0.07604, over 3325741.06 frames. ], batch size: 55, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:29:39,596 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:30:09,813 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 2900, loss[loss=0.2618, simple_loss=0.3106, pruned_loss=0.1065, over 16888.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3046, pruned_loss=0.07658, over 3333519.67 frames. ], batch size: 116, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:31:16,567 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4592, 4.3472, 4.3358, 3.8631, 4.3289, 2.0440, 4.1578, 4.3476], device='cuda:3'), covar=tensor([0.0055, 0.0065, 0.0087, 0.0293, 0.0074, 0.1281, 0.0080, 0.0103], device='cuda:3'), in_proj_covar=tensor([0.0087, 0.0079, 0.0117, 0.0127, 0.0087, 0.0126, 0.0103, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 02:31:29,005 INFO [train.py:904] (3/8) Epoch 4, batch 2950, loss[loss=0.1853, simple_loss=0.2656, pruned_loss=0.0525, over 16950.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3044, pruned_loss=0.07843, over 3327156.09 frames. ], batch size: 41, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:31:44,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0371, 1.6971, 1.3413, 1.4638, 1.7985, 1.7529, 1.8044, 1.8976], device='cuda:3'), covar=tensor([0.0033, 0.0107, 0.0151, 0.0143, 0.0080, 0.0117, 0.0058, 0.0085], device='cuda:3'), in_proj_covar=tensor([0.0077, 0.0137, 0.0137, 0.0136, 0.0133, 0.0141, 0.0107, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-28 02:32:01,322 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:27,159 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.112e+02 3.646e+02 4.580e+02 6.032e+02 1.054e+03, threshold=9.160e+02, percent-clipped=6.0 2023-04-28 02:32:35,894 INFO [train.py:904] (3/8) Epoch 4, batch 3000, loss[loss=0.1909, simple_loss=0.2667, pruned_loss=0.05756, over 15882.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3053, pruned_loss=0.07948, over 3324540.84 frames. ], batch size: 35, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:35,894 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 02:32:42,219 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7198, 2.3378, 2.2282, 4.1998, 1.7649, 3.6090, 2.1639, 1.9020], device='cuda:3'), covar=tensor([0.0415, 0.1478, 0.0811, 0.0184, 0.2645, 0.0449, 0.1609, 0.2527], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0271, 0.0227, 0.0289, 0.0336, 0.0257, 0.0253, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:32:45,554 INFO [train.py:938] (3/8) Epoch 4, validation: loss=0.1627, simple_loss=0.2694, pruned_loss=0.02796, over 944034.00 frames. 2023-04-28 02:32:45,555 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 02:32:52,098 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9338, 4.3625, 4.3989, 1.7206, 4.5917, 4.7023, 3.2003, 3.7640], device='cuda:3'), covar=tensor([0.0647, 0.0090, 0.0131, 0.1249, 0.0087, 0.0054, 0.0337, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0084, 0.0084, 0.0144, 0.0075, 0.0077, 0.0116, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 02:33:04,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1112, 3.6811, 3.2862, 5.2823, 5.0517, 4.5592, 1.9178, 3.5687], device='cuda:3'), covar=tensor([0.1115, 0.0407, 0.0796, 0.0052, 0.0186, 0.0268, 0.1153, 0.0539], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0137, 0.0164, 0.0079, 0.0171, 0.0158, 0.0155, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 02:33:12,889 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:31,072 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7565, 6.0244, 5.7300, 5.8992, 5.3317, 5.0361, 5.5801, 6.1656], device='cuda:3'), covar=tensor([0.0512, 0.0543, 0.0880, 0.0375, 0.0513, 0.0476, 0.0598, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0463, 0.0395, 0.0289, 0.0292, 0.0287, 0.0365, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:33:54,593 INFO [train.py:904] (3/8) Epoch 4, batch 3050, loss[loss=0.2229, simple_loss=0.2854, pruned_loss=0.08016, over 17016.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3051, pruned_loss=0.07963, over 3321126.44 frames. ], batch size: 41, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:34:07,858 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:15,426 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2032, 3.5360, 3.1287, 5.2967, 5.0213, 4.4600, 1.8338, 3.6037], device='cuda:3'), covar=tensor([0.1073, 0.0439, 0.0848, 0.0038, 0.0228, 0.0281, 0.1173, 0.0515], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0137, 0.0165, 0.0079, 0.0171, 0.0156, 0.0155, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 02:34:38,074 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:51,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8612, 5.1823, 4.8875, 4.9611, 4.6445, 4.5429, 4.6899, 5.2521], device='cuda:3'), covar=tensor([0.0660, 0.0577, 0.0761, 0.0370, 0.0513, 0.0621, 0.0577, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0458, 0.0393, 0.0288, 0.0291, 0.0286, 0.0362, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:34:54,165 INFO [optim.py:368] (3/8) 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 02:35:02,732 INFO [train.py:904] (3/8) Epoch 4, batch 3100, loss[loss=0.2277, simple_loss=0.2906, pruned_loss=0.0824, over 16915.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3039, pruned_loss=0.07957, over 3331453.74 frames. ], batch size: 109, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:35:10,119 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 02:36:10,677 INFO [train.py:904] (3/8) Epoch 4, batch 3150, loss[loss=0.2758, simple_loss=0.3301, pruned_loss=0.1108, over 16291.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3037, pruned_loss=0.07857, over 3335635.66 frames. ], batch size: 165, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:36:25,960 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4736, 4.3294, 3.3564, 1.8211, 2.5865, 2.2320, 3.8030, 4.2013], device='cuda:3'), covar=tensor([0.0219, 0.0351, 0.0554, 0.1568, 0.0842, 0.0982, 0.0473, 0.0486], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0129, 0.0151, 0.0140, 0.0132, 0.0126, 0.0140, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:36:39,551 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:07,361 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-28 02:37:11,328 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 3200, loss[loss=0.2224, simple_loss=0.3018, pruned_loss=0.0715, over 16631.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3014, pruned_loss=0.07677, over 3331794.37 frames. ], batch size: 62, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:37:44,851 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:48,785 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-28 02:37:54,591 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:38:04,039 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 02:38:25,339 INFO [train.py:904] (3/8) Epoch 4, batch 3250, loss[loss=0.2253, simple_loss=0.3023, pruned_loss=0.07421, over 16587.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3022, pruned_loss=0.07677, over 3326819.40 frames. ], batch size: 75, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:38:58,372 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,424 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:39:26,581 INFO [optim.py:368] (3/8) 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 02:39:36,484 INFO [train.py:904] (3/8) Epoch 4, batch 3300, loss[loss=0.2342, simple_loss=0.3024, pruned_loss=0.08301, over 16864.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3032, pruned_loss=0.0776, over 3326659.26 frames. ], batch size: 109, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:07,410 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:40:16,427 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8356, 5.1632, 4.8206, 4.9449, 4.6029, 4.3570, 4.6494, 5.1714], device='cuda:3'), covar=tensor([0.0593, 0.0581, 0.0845, 0.0403, 0.0543, 0.0779, 0.0530, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0463, 0.0394, 0.0291, 0.0295, 0.0292, 0.0365, 0.0324], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:40:45,510 INFO [train.py:904] (3/8) Epoch 4, batch 3350, loss[loss=0.2657, simple_loss=0.3231, pruned_loss=0.1042, over 16928.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3035, pruned_loss=0.07727, over 3328514.21 frames. ], batch size: 116, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:58,982 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:21,435 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:41:52,914 INFO [train.py:904] (3/8) Epoch 4, batch 3400, loss[loss=0.2122, simple_loss=0.3006, pruned_loss=0.06188, over 16752.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3034, pruned_loss=0.07695, over 3332710.35 frames. ], batch size: 57, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:42:04,787 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:43:03,522 INFO [train.py:904] (3/8) Epoch 4, batch 3450, loss[loss=0.2111, simple_loss=0.2771, pruned_loss=0.07258, over 16769.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3009, pruned_loss=0.07558, over 3333713.98 frames. ], batch size: 124, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:43:43,304 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-28 02:43:58,384 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:05,161 INFO [optim.py:368] (3/8) 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 02:44:13,346 INFO [train.py:904] (3/8) Epoch 4, batch 3500, loss[loss=0.1987, simple_loss=0.269, pruned_loss=0.06425, over 17011.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2991, pruned_loss=0.07526, over 3330443.48 frames. ], batch size: 41, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:44:48,807 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 02:44:51,929 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 02:45:26,873 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:45:27,576 INFO [train.py:904] (3/8) Epoch 4, batch 3550, loss[loss=0.2163, simple_loss=0.3017, pruned_loss=0.06547, over 17265.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.2986, pruned_loss=0.07508, over 3322104.68 frames. ], batch size: 52, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:45:46,104 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5548, 4.4150, 3.9963, 1.9864, 3.1468, 2.5469, 3.9728, 4.0117], device='cuda:3'), covar=tensor([0.0298, 0.0444, 0.0431, 0.1471, 0.0651, 0.0863, 0.0555, 0.0734], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0134, 0.0155, 0.0141, 0.0134, 0.0127, 0.0142, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:46:11,010 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:46:27,192 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:46:35,600 INFO [train.py:904] (3/8) Epoch 4, batch 3600, loss[loss=0.1972, simple_loss=0.26, pruned_loss=0.06717, over 16868.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2972, pruned_loss=0.07486, over 3307842.32 frames. ], batch size: 83, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:48,226 INFO [train.py:904] (3/8) Epoch 4, batch 3650, loss[loss=0.2239, simple_loss=0.3002, pruned_loss=0.07377, over 16821.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2967, pruned_loss=0.07546, over 3292672.31 frames. ], batch size: 42, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:50,631 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5656, 2.0361, 2.6900, 3.2690, 3.0254, 3.7223, 2.3889, 3.4072], device='cuda:3'), covar=tensor([0.0048, 0.0169, 0.0115, 0.0079, 0.0088, 0.0056, 0.0139, 0.0050], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0132, 0.0122, 0.0115, 0.0115, 0.0084, 0.0128, 0.0073], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 02:47:54,738 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:24,662 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-28 02:48:29,107 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:53,740 INFO [optim.py:368] (3/8) 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 02:48:54,314 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:02,643 INFO [train.py:904] (3/8) Epoch 4, batch 3700, loss[loss=0.2397, simple_loss=0.3017, pruned_loss=0.08884, over 16732.00 frames. ], tot_loss[loss=0.224, simple_loss=0.2949, pruned_loss=0.07656, over 3274227.71 frames. ], batch size: 124, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:49:17,889 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-04-28 02:49:41,422 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:54,187 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7167, 3.7857, 1.7748, 3.7837, 2.6424, 3.7620, 1.9126, 2.8075], device='cuda:3'), covar=tensor([0.0085, 0.0175, 0.1514, 0.0072, 0.0683, 0.0375, 0.1295, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0149, 0.0174, 0.0086, 0.0156, 0.0181, 0.0181, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:49:57,136 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3077, 1.7709, 2.2752, 3.2191, 2.8778, 3.4316, 1.5777, 3.2100], device='cuda:3'), covar=tensor([0.0046, 0.0212, 0.0149, 0.0071, 0.0077, 0.0044, 0.0226, 0.0036], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0133, 0.0121, 0.0115, 0.0115, 0.0084, 0.0129, 0.0073], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 02:50:17,552 INFO [train.py:904] (3/8) Epoch 4, batch 3750, loss[loss=0.2051, simple_loss=0.2741, pruned_loss=0.06802, over 16659.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2958, pruned_loss=0.07818, over 3265835.11 frames. ], batch size: 89, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:50:19,754 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 02:50:25,313 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:51:14,125 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 02:51:17,180 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4418, 4.1966, 4.4367, 4.6814, 4.7244, 4.2887, 4.5651, 4.7124], device='cuda:3'), covar=tensor([0.0720, 0.0680, 0.1133, 0.0438, 0.0468, 0.0636, 0.0787, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0425, 0.0564, 0.0445, 0.0333, 0.0316, 0.0357, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:51:21,418 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 3800, loss[loss=0.2357, simple_loss=0.3023, pruned_loss=0.08454, over 16506.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.2972, pruned_loss=0.07985, over 3274677.58 frames. ], batch size: 75, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:52:36,319 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:52:44,482 INFO [train.py:904] (3/8) Epoch 4, batch 3850, loss[loss=0.2326, simple_loss=0.3013, pruned_loss=0.08197, over 12388.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.2978, pruned_loss=0.08092, over 3263900.35 frames. ], batch size: 246, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:52:50,398 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5584, 4.3251, 3.6890, 1.7923, 2.8674, 2.5254, 3.7715, 4.0228], device='cuda:3'), covar=tensor([0.0167, 0.0307, 0.0507, 0.1638, 0.0733, 0.0841, 0.0504, 0.0541], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0131, 0.0157, 0.0141, 0.0136, 0.0127, 0.0141, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:53:15,004 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1303, 5.4564, 5.1093, 5.2209, 4.7162, 4.5397, 4.9507, 5.4860], device='cuda:3'), covar=tensor([0.0516, 0.0563, 0.0754, 0.0406, 0.0646, 0.0688, 0.0492, 0.0569], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0447, 0.0381, 0.0287, 0.0291, 0.0287, 0.0361, 0.0316], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 02:53:25,080 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 02:53:31,025 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 4, batch 3900, loss[loss=0.19, simple_loss=0.2759, pruned_loss=0.05207, over 17055.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.2968, pruned_loss=0.0808, over 3272489.37 frames. ], batch size: 50, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:54:14,097 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 02:54:40,914 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:55:09,051 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 3950, loss[loss=0.2545, simple_loss=0.3071, pruned_loss=0.101, over 16344.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.2967, pruned_loss=0.082, over 3273204.27 frames. ], batch size: 165, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:56:16,758 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 4000, loss[loss=0.2154, simple_loss=0.293, pruned_loss=0.06888, over 16992.00 frames. ], tot_loss[loss=0.229, simple_loss=0.2957, pruned_loss=0.08117, over 3274965.15 frames. ], batch size: 41, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:31,870 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 2023-04-28 02:57:36,428 INFO [train.py:904] (3/8) Epoch 4, batch 4050, loss[loss=0.217, simple_loss=0.2871, pruned_loss=0.07348, over 16747.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2943, pruned_loss=0.07891, over 3267380.69 frames. ], batch size: 57, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,748 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:58:09,239 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 02:58:27,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0906, 3.2315, 1.4100, 3.1877, 2.1838, 3.2123, 1.7916, 2.4573], device='cuda:3'), covar=tensor([0.0123, 0.0235, 0.1778, 0.0065, 0.0876, 0.0377, 0.1527, 0.0727], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0148, 0.0174, 0.0084, 0.0159, 0.0176, 0.0181, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 02:58:41,514 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.624e+02 2.984e+02 3.642e+02 7.677e+02, threshold=5.968e+02, percent-clipped=0.0 2023-04-28 02:58:48,881 INFO [train.py:904] (3/8) Epoch 4, batch 4100, loss[loss=0.2982, simple_loss=0.3653, pruned_loss=0.1156, over 15284.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.2952, pruned_loss=0.07768, over 3267870.34 frames. ], batch size: 190, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:59:22,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3836, 4.3879, 4.8789, 4.8464, 4.8474, 4.4001, 4.4843, 4.2953], device='cuda:3'), covar=tensor([0.0198, 0.0248, 0.0253, 0.0327, 0.0332, 0.0250, 0.0590, 0.0318], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0212, 0.0220, 0.0223, 0.0264, 0.0231, 0.0328, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 02:59:43,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0100, 1.4967, 2.2919, 2.9078, 2.7280, 3.2369, 1.8005, 2.9664], device='cuda:3'), covar=tensor([0.0047, 0.0216, 0.0113, 0.0084, 0.0091, 0.0052, 0.0179, 0.0044], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0133, 0.0120, 0.0115, 0.0114, 0.0085, 0.0130, 0.0074], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 02:59:54,021 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:56,615 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 4150, loss[loss=0.2651, simple_loss=0.339, pruned_loss=0.09557, over 16700.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3035, pruned_loss=0.0813, over 3239216.48 frames. ], batch size: 134, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:00:38,903 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:45,027 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:15,081 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 4200, loss[loss=0.2836, simple_loss=0.3627, pruned_loss=0.1022, over 16568.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3113, pruned_loss=0.08374, over 3207963.12 frames. ], batch size: 57, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:01:28,560 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:45,995 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8392, 2.6754, 2.5666, 1.9124, 2.5939, 2.5463, 2.6440, 1.7819], device='cuda:3'), covar=tensor([0.0232, 0.0028, 0.0025, 0.0171, 0.0033, 0.0042, 0.0027, 0.0210], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0049, 0.0053, 0.0105, 0.0056, 0.0060, 0.0055, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:02:13,187 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:02:19,279 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:39,120 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 4250, loss[loss=0.2326, simple_loss=0.3134, pruned_loss=0.07596, over 16570.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3144, pruned_loss=0.08367, over 3200408.40 frames. ], batch size: 57, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:02:53,686 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:03:47,976 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:03:56,213 INFO [train.py:904] (3/8) Epoch 4, batch 4300, loss[loss=0.267, simple_loss=0.347, pruned_loss=0.0935, over 16439.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3152, pruned_loss=0.08196, over 3209841.63 frames. ], batch size: 146, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:04:27,346 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:05:03,262 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0219, 1.6273, 1.4992, 1.4682, 1.6519, 1.6309, 1.7300, 1.8753], device='cuda:3'), covar=tensor([0.0026, 0.0109, 0.0144, 0.0131, 0.0084, 0.0115, 0.0048, 0.0079], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0135, 0.0136, 0.0133, 0.0129, 0.0137, 0.0098, 0.0115], device='cuda:3'), out_proj_covar=tensor([9.1829e-05, 1.7925e-04, 1.7486e-04, 1.7253e-04, 1.7272e-04, 1.8383e-04, 1.2958e-04, 1.5470e-04], device='cuda:3') 2023-04-28 03:05:10,251 INFO [train.py:904] (3/8) Epoch 4, batch 4350, loss[loss=0.2412, simple_loss=0.3209, pruned_loss=0.08071, over 16453.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3197, pruned_loss=0.08375, over 3188682.25 frames. ], batch size: 68, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:05:10,584 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:05:12,495 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 03:05:35,856 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2035, 2.1182, 1.7732, 1.9614, 2.5963, 2.2791, 3.1304, 2.9113], device='cuda:3'), covar=tensor([0.0020, 0.0150, 0.0202, 0.0176, 0.0094, 0.0162, 0.0038, 0.0082], device='cuda:3'), in_proj_covar=tensor([0.0071, 0.0137, 0.0138, 0.0135, 0.0130, 0.0140, 0.0100, 0.0116], device='cuda:3'), out_proj_covar=tensor([9.3365e-05, 1.8167e-04, 1.7768e-04, 1.7505e-04, 1.7460e-04, 1.8672e-04, 1.3208e-04, 1.5648e-04], device='cuda:3') 2023-04-28 03:06:03,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9187, 3.1569, 3.0104, 1.4765, 3.3323, 3.3333, 2.6684, 2.5109], device='cuda:3'), covar=tensor([0.0823, 0.0129, 0.0195, 0.1202, 0.0043, 0.0058, 0.0340, 0.0404], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0139, 0.0069, 0.0071, 0.0111, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 03:06:15,520 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.334e+02 4.019e+02 4.887e+02 8.589e+02, threshold=8.038e+02, percent-clipped=2.0 2023-04-28 03:06:20,218 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:22,170 INFO [train.py:904] (3/8) Epoch 4, batch 4400, loss[loss=0.2647, simple_loss=0.3284, pruned_loss=0.1005, over 11423.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.321, pruned_loss=0.08439, over 3192055.69 frames. ], batch size: 246, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:06:41,100 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 03:07:32,093 INFO [train.py:904] (3/8) Epoch 4, batch 4450, loss[loss=0.2612, simple_loss=0.3384, pruned_loss=0.09199, over 16844.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3248, pruned_loss=0.08588, over 3184562.45 frames. ], batch size: 42, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:36,347 INFO [optim.py:368] (3/8) 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 03:08:40,977 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:08:43,009 INFO [train.py:904] (3/8) Epoch 4, batch 4500, loss[loss=0.2441, simple_loss=0.3202, pruned_loss=0.08398, over 15461.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3238, pruned_loss=0.08518, over 3180523.11 frames. ], batch size: 191, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:54,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8660, 2.1582, 2.2116, 3.2196, 2.0220, 2.8121, 2.3294, 2.0255], device='cuda:3'), covar=tensor([0.0419, 0.1230, 0.0620, 0.0290, 0.2044, 0.0481, 0.1183, 0.1752], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0279, 0.0231, 0.0290, 0.0347, 0.0253, 0.0253, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 03:09:23,016 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:09:28,650 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 4550, loss[loss=0.2435, simple_loss=0.3214, pruned_loss=0.0828, over 16867.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3244, pruned_loss=0.08629, over 3171531.49 frames. ], batch size: 42, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:10:48,250 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 03:10:55,229 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-28 03:10:57,622 INFO [optim.py:368] (3/8) 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 03:11:04,059 INFO [train.py:904] (3/8) Epoch 4, batch 4600, loss[loss=0.2392, simple_loss=0.3166, pruned_loss=0.08085, over 17137.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.324, pruned_loss=0.08519, over 3181874.36 frames. ], batch size: 47, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:11:25,504 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:12:01,912 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:12:15,705 INFO [train.py:904] (3/8) Epoch 4, batch 4650, loss[loss=0.2125, simple_loss=0.2968, pruned_loss=0.06415, over 16539.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3224, pruned_loss=0.08435, over 3188378.67 frames. ], batch size: 75, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:13:20,501 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.712e+02 3.466e+02 4.003e+02 6.879e+02, threshold=6.932e+02, percent-clipped=1.0 2023-04-28 03:13:28,256 INFO [train.py:904] (3/8) Epoch 4, batch 4700, loss[loss=0.2324, simple_loss=0.3057, pruned_loss=0.07952, over 16630.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3193, pruned_loss=0.08267, over 3182344.92 frames. ], batch size: 57, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:13:32,477 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:41,638 INFO [train.py:904] (3/8) Epoch 4, batch 4750, loss[loss=0.1895, simple_loss=0.2725, pruned_loss=0.05322, over 17205.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3157, pruned_loss=0.08098, over 3177292.46 frames. ], batch size: 45, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:15:36,155 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-28 03:15:45,876 INFO [optim.py:368] (3/8) 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 03:15:52,085 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:53,763 INFO [train.py:904] (3/8) Epoch 4, batch 4800, loss[loss=0.3105, simple_loss=0.3839, pruned_loss=0.1186, over 15403.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3123, pruned_loss=0.07883, over 3186078.43 frames. ], batch size: 190, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:16:13,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9311, 3.9855, 3.7356, 3.7845, 3.4869, 3.9253, 3.5838, 3.6737], device='cuda:3'), covar=tensor([0.0358, 0.0166, 0.0192, 0.0130, 0.0561, 0.0193, 0.0513, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0151, 0.0197, 0.0160, 0.0224, 0.0178, 0.0142, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:16:30,553 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7191, 3.5021, 3.3945, 2.2272, 3.1698, 3.3270, 3.1861, 1.6912], device='cuda:3'), covar=tensor([0.0356, 0.0020, 0.0026, 0.0242, 0.0049, 0.0058, 0.0043, 0.0368], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0050, 0.0055, 0.0109, 0.0057, 0.0063, 0.0058, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:16:33,401 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:16:38,632 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:05,715 INFO [train.py:904] (3/8) Epoch 4, batch 4850, loss[loss=0.2211, simple_loss=0.312, pruned_loss=0.06508, over 16745.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3137, pruned_loss=0.07864, over 3179336.21 frames. ], batch size: 89, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:17:30,343 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 03:17:41,753 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:48,531 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:18:11,920 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 03:18:19,085 INFO [train.py:904] (3/8) Epoch 4, batch 4900, loss[loss=0.2332, simple_loss=0.3195, pruned_loss=0.07346, over 16245.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3136, pruned_loss=0.07763, over 3175512.26 frames. ], batch size: 165, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:18:26,061 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-04-28 03:18:42,145 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:18:49,182 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9351, 5.6694, 5.7974, 5.5082, 5.6379, 6.0787, 5.8369, 5.5356], device='cuda:3'), covar=tensor([0.0605, 0.1093, 0.1009, 0.1316, 0.2053, 0.0895, 0.0759, 0.1860], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0329, 0.0311, 0.0291, 0.0382, 0.0341, 0.0272, 0.0390], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:18:53,774 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:19:32,910 INFO [train.py:904] (3/8) Epoch 4, batch 4950, loss[loss=0.257, simple_loss=0.333, pruned_loss=0.09054, over 16785.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3138, pruned_loss=0.07753, over 3180176.05 frames. ], batch size: 124, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:19:41,840 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:19:53,434 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:20:22,734 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:37,430 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.040e+02 3.403e+02 3.998e+02 5.018e+02 8.259e+02, threshold=7.996e+02, percent-clipped=7.0 2023-04-28 03:20:41,761 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:45,624 INFO [train.py:904] (3/8) Epoch 4, batch 5000, loss[loss=0.2002, simple_loss=0.2932, pruned_loss=0.05359, over 16687.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3156, pruned_loss=0.07768, over 3196731.43 frames. ], batch size: 83, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:21:57,693 INFO [train.py:904] (3/8) Epoch 4, batch 5050, loss[loss=0.2253, simple_loss=0.3114, pruned_loss=0.06962, over 15527.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3162, pruned_loss=0.07751, over 3194311.89 frames. ], batch size: 191, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:22:39,990 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 4, batch 5100, loss[loss=0.2273, simple_loss=0.3094, pruned_loss=0.07262, over 16358.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3145, pruned_loss=0.07697, over 3183599.86 frames. ], batch size: 146, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:24:08,099 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:24:22,823 INFO [train.py:904] (3/8) Epoch 4, batch 5150, loss[loss=0.2508, simple_loss=0.329, pruned_loss=0.08633, over 12252.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3141, pruned_loss=0.07554, over 3189226.68 frames. ], batch size: 247, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:29,063 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 5200, loss[loss=0.2091, simple_loss=0.2936, pruned_loss=0.06227, over 16726.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3123, pruned_loss=0.07512, over 3178862.91 frames. ], batch size: 89, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:51,286 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4262, 4.7694, 4.3811, 4.4735, 4.1545, 4.1744, 4.2640, 4.7682], device='cuda:3'), covar=tensor([0.0554, 0.0704, 0.0906, 0.0473, 0.0566, 0.0753, 0.0598, 0.0690], device='cuda:3'), in_proj_covar=tensor([0.0321, 0.0431, 0.0372, 0.0279, 0.0275, 0.0279, 0.0343, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 03:26:46,332 INFO [train.py:904] (3/8) Epoch 4, batch 5250, loss[loss=0.208, simple_loss=0.2818, pruned_loss=0.06709, over 17245.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3095, pruned_loss=0.07486, over 3181367.25 frames. ], batch size: 52, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:47,352 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:52,066 INFO [optim.py:368] (3/8) 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 03:27:55,778 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 5300, loss[loss=0.2047, simple_loss=0.2827, pruned_loss=0.06331, over 16893.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3057, pruned_loss=0.07319, over 3194387.82 frames. ], batch size: 96, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:28:28,794 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:30,178 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-28 03:29:02,750 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:11,341 INFO [train.py:904] (3/8) Epoch 4, batch 5350, loss[loss=0.2592, simple_loss=0.3401, pruned_loss=0.08914, over 16342.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3038, pruned_loss=0.07242, over 3194505.78 frames. ], batch size: 146, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:29:24,937 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:56,961 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:17,960 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 5400, loss[loss=0.2566, simple_loss=0.3383, pruned_loss=0.08743, over 16733.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3073, pruned_loss=0.07393, over 3185685.05 frames. ], batch size: 89, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:30:53,684 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:13,819 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:38,572 INFO [train.py:904] (3/8) Epoch 4, batch 5450, loss[loss=0.2605, simple_loss=0.3372, pruned_loss=0.09185, over 16865.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3111, pruned_loss=0.0764, over 3188450.06 frames. ], batch size: 116, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:31:56,009 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 03:32:50,254 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 5500, loss[loss=0.284, simple_loss=0.3596, pruned_loss=0.1042, over 16713.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.321, pruned_loss=0.08372, over 3175268.03 frames. ], batch size: 89, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:16,678 INFO [train.py:904] (3/8) Epoch 4, batch 5550, loss[loss=0.2771, simple_loss=0.3492, pruned_loss=0.1025, over 17034.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3312, pruned_loss=0.09272, over 3119133.11 frames. ], batch size: 55, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:17,786 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:35:02,208 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:35:29,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 03:35:29,646 INFO [optim.py:368] (3/8) 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 03:35:33,005 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 5600, loss[loss=0.3504, simple_loss=0.3848, pruned_loss=0.158, over 10965.00 frames. ], tot_loss[loss=0.268, simple_loss=0.3383, pruned_loss=0.09886, over 3085398.04 frames. ], batch size: 247, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:36:20,419 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:36:25,683 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-04-28 03:36:57,478 INFO [train.py:904] (3/8) Epoch 4, batch 5650, loss[loss=0.2813, simple_loss=0.3435, pruned_loss=0.1096, over 16877.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3454, pruned_loss=0.1053, over 3056469.68 frames. ], batch size: 116, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:37:41,227 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:03,587 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4653, 4.6611, 4.7189, 4.7281, 4.6856, 5.2015, 4.8274, 4.5947], device='cuda:3'), covar=tensor([0.0945, 0.1550, 0.1316, 0.1445, 0.2052, 0.0835, 0.1114, 0.2225], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0351, 0.0336, 0.0312, 0.0404, 0.0363, 0.0285, 0.0420], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:38:09,921 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.856e+02 5.092e+02 6.431e+02 7.893e+02 1.649e+03, threshold=1.286e+03, percent-clipped=5.0 2023-04-28 03:38:17,756 INFO [train.py:904] (3/8) Epoch 4, batch 5700, loss[loss=0.2764, simple_loss=0.3538, pruned_loss=0.09945, over 16400.00 frames. ], tot_loss[loss=0.2796, simple_loss=0.3464, pruned_loss=0.1064, over 3053448.96 frames. ], batch size: 146, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:38:37,455 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5256, 2.8286, 2.5232, 4.2348, 3.7529, 3.8500, 1.4700, 2.9104], device='cuda:3'), covar=tensor([0.1749, 0.0586, 0.1235, 0.0131, 0.0278, 0.0346, 0.1639, 0.0793], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0134, 0.0160, 0.0072, 0.0149, 0.0154, 0.0152, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 03:38:42,659 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:05,522 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:13,391 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:36,185 INFO [train.py:904] (3/8) Epoch 4, batch 5750, loss[loss=0.2845, simple_loss=0.3556, pruned_loss=0.1067, over 16872.00 frames. ], tot_loss[loss=0.2827, simple_loss=0.3494, pruned_loss=0.108, over 3043785.03 frames. ], batch size: 109, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:39:49,547 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3395, 1.4979, 2.2296, 2.9222, 2.8268, 3.4715, 1.6328, 3.1682], device='cuda:3'), covar=tensor([0.0042, 0.0244, 0.0149, 0.0106, 0.0102, 0.0040, 0.0223, 0.0045], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0129, 0.0116, 0.0111, 0.0113, 0.0080, 0.0129, 0.0076], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 03:40:29,638 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:37,017 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9826, 3.8720, 3.8955, 2.7561, 3.8894, 1.5949, 3.5940, 3.7663], device='cuda:3'), covar=tensor([0.0127, 0.0113, 0.0109, 0.0581, 0.0092, 0.2185, 0.0137, 0.0230], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0069, 0.0105, 0.0116, 0.0078, 0.0125, 0.0091, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 03:40:42,459 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 03:40:55,818 INFO [train.py:904] (3/8) Epoch 4, batch 5800, loss[loss=0.274, simple_loss=0.3479, pruned_loss=0.1001, over 16375.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3477, pruned_loss=0.1057, over 3044187.95 frames. ], batch size: 146, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:40:59,820 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 03:41:17,639 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 03:42:12,358 INFO [train.py:904] (3/8) Epoch 4, batch 5850, loss[loss=0.3003, simple_loss=0.3515, pruned_loss=0.1245, over 11754.00 frames. ], tot_loss[loss=0.2769, simple_loss=0.3461, pruned_loss=0.1039, over 3037768.90 frames. ], batch size: 246, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:43:29,068 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 5900, loss[loss=0.2609, simple_loss=0.3259, pruned_loss=0.09791, over 16453.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3448, pruned_loss=0.1029, over 3054737.28 frames. ], batch size: 35, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:44:56,221 INFO [train.py:904] (3/8) Epoch 4, batch 5950, loss[loss=0.2781, simple_loss=0.3561, pruned_loss=0.1001, over 16927.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3457, pruned_loss=0.1017, over 3066509.64 frames. ], batch size: 109, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:45:37,412 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:45:47,074 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6031, 2.4533, 2.3076, 4.1636, 1.7973, 3.4914, 2.3196, 2.2383], device='cuda:3'), covar=tensor([0.0473, 0.1299, 0.0794, 0.0253, 0.2549, 0.0484, 0.1401, 0.1837], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0280, 0.0231, 0.0293, 0.0347, 0.0252, 0.0255, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 03:46:10,128 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 6000, loss[loss=0.2291, simple_loss=0.3073, pruned_loss=0.07543, over 16377.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3448, pruned_loss=0.1015, over 3055673.90 frames. ], batch size: 35, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:46:14,570 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 03:46:25,218 INFO [train.py:938] (3/8) Epoch 4, validation: loss=0.1966, simple_loss=0.3069, pruned_loss=0.04309, over 944034.00 frames. 2023-04-28 03:46:25,218 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 03:46:34,073 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7264, 3.6430, 3.7544, 3.6453, 3.6532, 4.1015, 3.8894, 3.6395], device='cuda:3'), covar=tensor([0.1684, 0.1817, 0.1447, 0.2298, 0.2671, 0.1396, 0.1312, 0.2425], device='cuda:3'), in_proj_covar=tensor([0.0247, 0.0341, 0.0330, 0.0304, 0.0401, 0.0363, 0.0281, 0.0407], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:46:49,524 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:01,320 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:44,053 INFO [train.py:904] (3/8) Epoch 4, batch 6050, loss[loss=0.255, simple_loss=0.3404, pruned_loss=0.08482, over 16725.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.3432, pruned_loss=0.1, over 3070193.17 frames. ], batch size: 124, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:48:06,439 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:33,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7849, 3.1812, 3.1885, 1.4191, 3.2483, 3.3312, 2.5631, 2.4349], device='cuda:3'), covar=tensor([0.0885, 0.0103, 0.0145, 0.1282, 0.0076, 0.0072, 0.0427, 0.0482], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0084, 0.0079, 0.0143, 0.0069, 0.0073, 0.0115, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 03:48:42,914 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:58,311 INFO [optim.py:368] (3/8) 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 03:49:03,047 INFO [train.py:904] (3/8) Epoch 4, batch 6100, loss[loss=0.2666, simple_loss=0.3431, pruned_loss=0.09504, over 16782.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3411, pruned_loss=0.09749, over 3100988.82 frames. ], batch size: 102, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:49:19,171 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8981, 1.6612, 1.4652, 1.5187, 1.8846, 1.7414, 1.8129, 1.9297], device='cuda:3'), covar=tensor([0.0027, 0.0107, 0.0157, 0.0154, 0.0081, 0.0120, 0.0065, 0.0081], device='cuda:3'), in_proj_covar=tensor([0.0069, 0.0141, 0.0144, 0.0141, 0.0132, 0.0144, 0.0101, 0.0121], device='cuda:3'), out_proj_covar=tensor([8.9031e-05, 1.8607e-04, 1.8272e-04, 1.8144e-04, 1.7507e-04, 1.8927e-04, 1.3031e-04, 1.6032e-04], device='cuda:3') 2023-04-28 03:49:31,447 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 03:50:22,075 INFO [train.py:904] (3/8) Epoch 4, batch 6150, loss[loss=0.2726, simple_loss=0.3418, pruned_loss=0.1016, over 15251.00 frames. ], tot_loss[loss=0.266, simple_loss=0.339, pruned_loss=0.09657, over 3113505.85 frames. ], batch size: 190, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:23,097 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-28 03:50:29,640 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:34,983 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:41,714 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:11,844 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2628, 3.1473, 3.2261, 3.3984, 3.4195, 3.1906, 3.3777, 3.4328], device='cuda:3'), covar=tensor([0.0587, 0.0665, 0.1056, 0.0426, 0.0473, 0.1569, 0.0676, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0338, 0.0416, 0.0534, 0.0410, 0.0312, 0.0301, 0.0335, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 03:51:38,877 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 6200, loss[loss=0.2306, simple_loss=0.3007, pruned_loss=0.08021, over 17267.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3361, pruned_loss=0.09527, over 3111221.16 frames. ], batch size: 52, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:52:07,249 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:11,530 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:17,387 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7593, 4.7140, 4.5126, 4.5384, 4.0738, 4.6521, 4.5820, 4.3607], device='cuda:3'), covar=tensor([0.0403, 0.0235, 0.0221, 0.0164, 0.0816, 0.0248, 0.0247, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0156, 0.0195, 0.0160, 0.0223, 0.0185, 0.0144, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:52:17,451 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:18,674 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4128, 4.3855, 4.4124, 3.0015, 3.9996, 4.3226, 4.1444, 2.4239], device='cuda:3'), covar=tensor([0.0293, 0.0016, 0.0028, 0.0184, 0.0038, 0.0046, 0.0027, 0.0246], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0049, 0.0055, 0.0109, 0.0056, 0.0062, 0.0057, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 03:52:57,844 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:59,110 INFO [train.py:904] (3/8) Epoch 4, batch 6250, loss[loss=0.2859, simple_loss=0.3561, pruned_loss=0.1078, over 15454.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.336, pruned_loss=0.09523, over 3134176.60 frames. ], batch size: 191, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:53:12,762 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-04-28 03:53:18,979 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3135, 3.1801, 3.2743, 3.4271, 3.4499, 3.1987, 3.4008, 3.4893], device='cuda:3'), covar=tensor([0.0584, 0.0584, 0.0956, 0.0461, 0.0548, 0.1443, 0.0696, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0411, 0.0527, 0.0410, 0.0314, 0.0303, 0.0338, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 03:53:46,308 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 03:54:08,512 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2137, 4.3724, 1.9403, 4.6869, 2.7843, 4.7638, 2.3459, 3.2316], device='cuda:3'), covar=tensor([0.0088, 0.0222, 0.1647, 0.0023, 0.0760, 0.0199, 0.1246, 0.0555], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0147, 0.0176, 0.0077, 0.0160, 0.0179, 0.0185, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 03:54:09,136 INFO [optim.py:368] (3/8) 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 03:54:14,287 INFO [train.py:904] (3/8) Epoch 4, batch 6300, loss[loss=0.2318, simple_loss=0.3209, pruned_loss=0.07138, over 16748.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3361, pruned_loss=0.09462, over 3145105.85 frames. ], batch size: 83, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:54:25,674 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3786, 4.0526, 3.7026, 1.9345, 2.8967, 2.6295, 3.6167, 3.7981], device='cuda:3'), covar=tensor([0.0299, 0.0479, 0.0475, 0.1667, 0.0753, 0.0867, 0.0702, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0126, 0.0157, 0.0146, 0.0141, 0.0132, 0.0147, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 03:54:31,258 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:55:23,690 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 03:55:32,004 INFO [train.py:904] (3/8) Epoch 4, batch 6350, loss[loss=0.2764, simple_loss=0.3504, pruned_loss=0.1012, over 16835.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3377, pruned_loss=0.09636, over 3127845.26 frames. ], batch size: 116, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:55:38,845 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:56:11,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1103, 2.4875, 2.5801, 3.2419, 2.9035, 3.2998, 1.6612, 2.9456], device='cuda:3'), covar=tensor([0.1034, 0.0364, 0.0694, 0.0089, 0.0197, 0.0274, 0.1120, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0136, 0.0163, 0.0076, 0.0157, 0.0160, 0.0156, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 03:56:28,010 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:43,683 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 6400, loss[loss=0.339, simple_loss=0.3765, pruned_loss=0.1507, over 11413.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3387, pruned_loss=0.09836, over 3091767.44 frames. ], batch size: 248, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:56:49,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6482, 1.4196, 1.9556, 2.4029, 2.5026, 2.7972, 1.4498, 2.7629], device='cuda:3'), covar=tensor([0.0062, 0.0229, 0.0156, 0.0110, 0.0094, 0.0061, 0.0212, 0.0046], device='cuda:3'), in_proj_covar=tensor([0.0100, 0.0131, 0.0119, 0.0113, 0.0117, 0.0082, 0.0131, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 03:57:10,839 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:57:39,965 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:04,368 INFO [train.py:904] (3/8) Epoch 4, batch 6450, loss[loss=0.2406, simple_loss=0.3265, pruned_loss=0.07738, over 16543.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.337, pruned_loss=0.09618, over 3096427.46 frames. ], batch size: 68, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:58:20,093 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:50,908 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6788, 3.4444, 3.3615, 1.3037, 3.4603, 3.5544, 2.7702, 2.4667], device='cuda:3'), covar=tensor([0.1276, 0.0126, 0.0176, 0.1625, 0.0072, 0.0072, 0.0398, 0.0624], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0083, 0.0079, 0.0140, 0.0066, 0.0071, 0.0111, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 03:58:54,868 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 03:59:18,064 INFO [optim.py:368] (3/8) 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 03:59:23,779 INFO [train.py:904] (3/8) Epoch 4, batch 6500, loss[loss=0.2624, simple_loss=0.3221, pruned_loss=0.1013, over 16791.00 frames. ], tot_loss[loss=0.2614, simple_loss=0.3335, pruned_loss=0.09464, over 3108443.75 frames. ], batch size: 39, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:59:35,237 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:38,114 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 03:59:38,826 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:44,061 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:49,578 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:56,662 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:18,962 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 04:00:25,656 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2242, 5.0923, 5.0037, 4.8836, 4.3815, 4.9892, 5.0078, 4.7025], device='cuda:3'), covar=tensor([0.0425, 0.0416, 0.0181, 0.0159, 0.0924, 0.0318, 0.0183, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0155, 0.0192, 0.0158, 0.0221, 0.0184, 0.0143, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 04:00:42,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5560, 3.5612, 4.0316, 3.9873, 3.9866, 3.6726, 3.7162, 3.6985], device='cuda:3'), covar=tensor([0.0267, 0.0423, 0.0348, 0.0439, 0.0418, 0.0311, 0.0731, 0.0370], device='cuda:3'), in_proj_covar=tensor([0.0217, 0.0205, 0.0212, 0.0218, 0.0256, 0.0223, 0.0325, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 04:00:44,410 INFO [train.py:904] (3/8) Epoch 4, batch 6550, loss[loss=0.3248, simple_loss=0.3693, pruned_loss=0.1402, over 11437.00 frames. ], tot_loss[loss=0.2645, simple_loss=0.337, pruned_loss=0.09601, over 3107539.24 frames. ], batch size: 246, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:00:50,589 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:02,968 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-28 04:01:13,110 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:56,563 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 6600, loss[loss=0.2918, simple_loss=0.3507, pruned_loss=0.1165, over 11626.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3403, pruned_loss=0.09763, over 3083094.39 frames. ], batch size: 250, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:02:06,619 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1684, 5.1596, 4.9463, 4.2022, 4.9735, 1.9608, 4.7793, 5.0345], device='cuda:3'), covar=tensor([0.0050, 0.0034, 0.0060, 0.0276, 0.0039, 0.1356, 0.0060, 0.0077], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0069, 0.0105, 0.0114, 0.0078, 0.0127, 0.0094, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:02:07,961 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3321, 2.2243, 1.6404, 2.0450, 2.8878, 2.5895, 3.3725, 3.2548], device='cuda:3'), covar=tensor([0.0017, 0.0167, 0.0241, 0.0185, 0.0086, 0.0146, 0.0061, 0.0067], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0139, 0.0143, 0.0138, 0.0133, 0.0144, 0.0102, 0.0120], device='cuda:3'), out_proj_covar=tensor([8.7061e-05, 1.8126e-04, 1.8095e-04, 1.7576e-04, 1.7453e-04, 1.8800e-04, 1.3057e-04, 1.5816e-04], device='cuda:3') 2023-04-28 04:02:09,068 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:02:17,008 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4507, 4.3974, 4.2731, 4.1856, 3.8016, 4.3145, 4.2265, 4.0221], device='cuda:3'), covar=tensor([0.0397, 0.0227, 0.0212, 0.0185, 0.0793, 0.0297, 0.0301, 0.0420], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0155, 0.0193, 0.0160, 0.0222, 0.0187, 0.0145, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 04:02:22,658 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:03:19,967 INFO [train.py:904] (3/8) Epoch 4, batch 6650, loss[loss=0.2596, simple_loss=0.3392, pruned_loss=0.09003, over 16400.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3405, pruned_loss=0.0984, over 3081678.24 frames. ], batch size: 146, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:03:45,623 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7993, 2.6541, 2.3907, 3.7758, 3.3620, 3.6732, 1.5812, 2.8194], device='cuda:3'), covar=tensor([0.1311, 0.0479, 0.1035, 0.0065, 0.0283, 0.0315, 0.1298, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0136, 0.0163, 0.0074, 0.0157, 0.0157, 0.0156, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 04:04:33,021 INFO [optim.py:368] (3/8) 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 04:04:37,774 INFO [train.py:904] (3/8) Epoch 4, batch 6700, loss[loss=0.2639, simple_loss=0.3336, pruned_loss=0.09712, over 16942.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.339, pruned_loss=0.09813, over 3074167.98 frames. ], batch size: 109, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:53,075 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:05:47,882 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9641, 3.9735, 3.8210, 3.8076, 3.4881, 3.9224, 3.6331, 3.6389], device='cuda:3'), covar=tensor([0.0358, 0.0196, 0.0177, 0.0151, 0.0697, 0.0207, 0.0603, 0.0395], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0156, 0.0194, 0.0161, 0.0222, 0.0184, 0.0146, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 04:05:53,496 INFO [train.py:904] (3/8) Epoch 4, batch 6750, loss[loss=0.2307, simple_loss=0.3176, pruned_loss=0.07186, over 16651.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.338, pruned_loss=0.0984, over 3068921.79 frames. ], batch size: 134, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:05:58,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8883, 2.1658, 2.2309, 3.2126, 2.0472, 2.7850, 2.2585, 1.9445], device='cuda:3'), covar=tensor([0.0491, 0.1335, 0.0631, 0.0269, 0.2066, 0.0564, 0.1361, 0.1729], device='cuda:3'), in_proj_covar=tensor([0.0295, 0.0282, 0.0233, 0.0295, 0.0348, 0.0259, 0.0259, 0.0347], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:06:34,416 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9036, 1.5774, 2.0760, 2.6792, 2.6964, 3.0445, 1.6386, 3.1162], device='cuda:3'), covar=tensor([0.0056, 0.0219, 0.0147, 0.0097, 0.0099, 0.0057, 0.0210, 0.0046], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0131, 0.0119, 0.0111, 0.0119, 0.0081, 0.0131, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 04:06:43,828 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 04:06:49,915 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4418, 4.2761, 4.1848, 3.4407, 4.1587, 1.6033, 3.9897, 4.1031], device='cuda:3'), covar=tensor([0.0053, 0.0047, 0.0078, 0.0318, 0.0059, 0.1661, 0.0082, 0.0111], device='cuda:3'), in_proj_covar=tensor([0.0082, 0.0071, 0.0109, 0.0119, 0.0080, 0.0132, 0.0096, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:07:01,366 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.46 vs. limit=5.0 2023-04-28 04:07:06,101 INFO [optim.py:368] (3/8) 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 04:07:08,237 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3366, 4.6273, 4.3763, 4.4172, 4.0531, 4.0061, 4.2732, 4.6222], device='cuda:3'), covar=tensor([0.0581, 0.0593, 0.0887, 0.0409, 0.0551, 0.0840, 0.0510, 0.0626], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0430, 0.0378, 0.0284, 0.0277, 0.0285, 0.0351, 0.0306], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:07:10,550 INFO [train.py:904] (3/8) Epoch 4, batch 6800, loss[loss=0.2816, simple_loss=0.3544, pruned_loss=0.1044, over 16202.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3377, pruned_loss=0.09786, over 3078068.37 frames. ], batch size: 165, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:07:27,127 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:27,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3232, 1.8393, 1.3909, 1.6097, 2.2728, 2.0414, 2.3982, 2.3726], device='cuda:3'), covar=tensor([0.0035, 0.0153, 0.0247, 0.0201, 0.0094, 0.0161, 0.0059, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0068, 0.0137, 0.0144, 0.0136, 0.0132, 0.0142, 0.0101, 0.0120], device='cuda:3'), out_proj_covar=tensor([8.7013e-05, 1.7924e-04, 1.8233e-04, 1.7306e-04, 1.7342e-04, 1.8607e-04, 1.2889e-04, 1.5839e-04], device='cuda:3') 2023-04-28 04:07:31,491 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:35,282 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:37,306 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:37,325 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:00,360 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4652, 3.6328, 2.8048, 2.2347, 2.6185, 2.1215, 3.7061, 3.8571], device='cuda:3'), covar=tensor([0.2193, 0.0606, 0.1146, 0.1308, 0.1737, 0.1315, 0.0429, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0281, 0.0247, 0.0262, 0.0234, 0.0310, 0.0197, 0.0231, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:08:27,076 INFO [train.py:904] (3/8) Epoch 4, batch 6850, loss[loss=0.2847, simple_loss=0.359, pruned_loss=0.1052, over 16295.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3399, pruned_loss=0.09914, over 3077167.27 frames. ], batch size: 165, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:08:38,764 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:43,642 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:46,841 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:49,159 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:08,835 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:36,159 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.986e+02 4.056e+02 4.810e+02 6.733e+02 1.178e+03, threshold=9.620e+02, percent-clipped=3.0 2023-04-28 04:09:39,829 INFO [train.py:904] (3/8) Epoch 4, batch 6900, loss[loss=0.3091, simple_loss=0.3753, pruned_loss=0.1215, over 16462.00 frames. ], tot_loss[loss=0.2707, simple_loss=0.3429, pruned_loss=0.09929, over 3074800.65 frames. ], batch size: 146, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:09:48,131 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 04:09:52,854 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:59,887 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:10:55,352 INFO [train.py:904] (3/8) Epoch 4, batch 6950, loss[loss=0.2744, simple_loss=0.3495, pruned_loss=0.09968, over 16410.00 frames. ], tot_loss[loss=0.272, simple_loss=0.3438, pruned_loss=0.1001, over 3097149.34 frames. ], batch size: 146, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:11:00,665 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:11:32,938 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:10,357 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 7000, loss[loss=0.2119, simple_loss=0.3037, pruned_loss=0.06004, over 16813.00 frames. ], tot_loss[loss=0.2706, simple_loss=0.3437, pruned_loss=0.0988, over 3116497.57 frames. ], batch size: 39, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:12:28,099 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:12:34,195 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2933, 4.2393, 4.1369, 4.0451, 3.7037, 4.1559, 4.0869, 3.8899], device='cuda:3'), covar=tensor([0.0370, 0.0213, 0.0179, 0.0154, 0.0693, 0.0269, 0.0342, 0.0382], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0154, 0.0190, 0.0158, 0.0222, 0.0185, 0.0144, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 04:13:29,622 INFO [train.py:904] (3/8) Epoch 4, batch 7050, loss[loss=0.2961, simple_loss=0.3602, pruned_loss=0.116, over 15289.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3447, pruned_loss=0.099, over 3116608.09 frames. ], batch size: 190, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:13:43,150 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:14:17,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6154, 4.8998, 4.5609, 4.5732, 4.2713, 4.1588, 4.4312, 4.8948], device='cuda:3'), covar=tensor([0.0529, 0.0635, 0.0981, 0.0495, 0.0592, 0.0815, 0.0541, 0.0682], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0428, 0.0377, 0.0284, 0.0275, 0.0286, 0.0349, 0.0305], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:14:45,408 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 3.107e+02 4.211e+02 5.076e+02 6.594e+02 1.566e+03, threshold=1.015e+03, percent-clipped=2.0 2023-04-28 04:14:46,710 INFO [train.py:904] (3/8) Epoch 4, batch 7100, loss[loss=0.2677, simple_loss=0.3473, pruned_loss=0.09407, over 16439.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3427, pruned_loss=0.09838, over 3118098.43 frames. ], batch size: 75, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:15:13,009 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:15:51,807 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 04:15:57,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5216, 3.3789, 2.9021, 1.7533, 2.5498, 2.0501, 3.0325, 3.2328], device='cuda:3'), covar=tensor([0.0310, 0.0517, 0.0563, 0.1746, 0.0829, 0.1023, 0.0631, 0.0596], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0124, 0.0155, 0.0142, 0.0137, 0.0128, 0.0143, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 04:16:02,655 INFO [train.py:904] (3/8) Epoch 4, batch 7150, loss[loss=0.2951, simple_loss=0.3594, pruned_loss=0.1154, over 15297.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3407, pruned_loss=0.09818, over 3102080.50 frames. ], batch size: 190, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:16:23,199 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:24,831 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:38,280 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 04:17:18,900 INFO [train.py:904] (3/8) Epoch 4, batch 7200, loss[loss=0.2022, simple_loss=0.2865, pruned_loss=0.05893, over 16700.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3388, pruned_loss=0.09702, over 3074127.45 frames. ], batch size: 89, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:17:31,686 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 04:17:33,296 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:35,457 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:18:40,008 INFO [train.py:904] (3/8) Epoch 4, batch 7250, loss[loss=0.2479, simple_loss=0.3246, pruned_loss=0.08559, over 15386.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3355, pruned_loss=0.09442, over 3087031.37 frames. ], batch size: 190, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:18:52,049 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:10,334 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:55,474 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 3.869e+02 4.810e+02 5.847e+02 1.256e+03, threshold=9.620e+02, percent-clipped=3.0 2023-04-28 04:19:57,416 INFO [train.py:904] (3/8) Epoch 4, batch 7300, loss[loss=0.2659, simple_loss=0.3438, pruned_loss=0.09402, over 16470.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3349, pruned_loss=0.09457, over 3071927.09 frames. ], batch size: 146, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:21:14,972 INFO [train.py:904] (3/8) Epoch 4, batch 7350, loss[loss=0.2568, simple_loss=0.3311, pruned_loss=0.09122, over 16766.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3343, pruned_loss=0.0947, over 3041551.82 frames. ], batch size: 124, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:22:29,891 INFO [optim.py:368] (3/8) 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 04:22:31,804 INFO [train.py:904] (3/8) Epoch 4, batch 7400, loss[loss=0.2832, simple_loss=0.365, pruned_loss=0.1007, over 16669.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3358, pruned_loss=0.09528, over 3058615.96 frames. ], batch size: 76, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:23:19,096 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-28 04:23:22,648 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 04:23:49,015 INFO [train.py:904] (3/8) Epoch 4, batch 7450, loss[loss=0.2451, simple_loss=0.3192, pruned_loss=0.08553, over 16630.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3376, pruned_loss=0.09698, over 3059318.87 frames. ], batch size: 57, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:24:26,709 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:24:39,463 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2534, 4.1767, 4.1065, 3.4749, 4.1030, 1.6226, 3.9686, 4.0142], device='cuda:3'), covar=tensor([0.0053, 0.0055, 0.0073, 0.0265, 0.0054, 0.1622, 0.0064, 0.0108], device='cuda:3'), in_proj_covar=tensor([0.0080, 0.0069, 0.0105, 0.0114, 0.0078, 0.0128, 0.0093, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:25:06,399 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.829e+02 4.545e+02 5.686e+02 6.937e+02 1.455e+03, threshold=1.137e+03, percent-clipped=4.0 2023-04-28 04:25:07,721 INFO [train.py:904] (3/8) Epoch 4, batch 7500, loss[loss=0.2518, simple_loss=0.3326, pruned_loss=0.08551, over 16207.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3383, pruned_loss=0.09675, over 3059515.40 frames. ], batch size: 165, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:25:09,404 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8451, 4.1344, 3.8364, 3.9378, 3.6200, 3.7308, 3.8450, 4.0576], device='cuda:3'), covar=tensor([0.0692, 0.0711, 0.1049, 0.0514, 0.0619, 0.1196, 0.0629, 0.0870], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0433, 0.0378, 0.0282, 0.0277, 0.0290, 0.0355, 0.0307], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:25:40,563 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:50,240 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:27,335 INFO [train.py:904] (3/8) Epoch 4, batch 7550, loss[loss=0.2359, simple_loss=0.315, pruned_loss=0.07843, over 16714.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3374, pruned_loss=0.09721, over 3042862.26 frames. ], batch size: 134, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:26:54,918 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:23,979 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:27,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0051, 3.8610, 3.9282, 4.2201, 4.2950, 3.9945, 4.3292, 4.2777], device='cuda:3'), covar=tensor([0.0777, 0.0765, 0.1509, 0.0569, 0.0568, 0.0903, 0.0596, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0402, 0.0520, 0.0416, 0.0309, 0.0300, 0.0336, 0.0334], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:27:36,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4127, 3.4503, 3.1982, 3.2437, 3.0700, 3.3582, 3.2281, 3.1309], device='cuda:3'), covar=tensor([0.0414, 0.0223, 0.0180, 0.0148, 0.0524, 0.0245, 0.0862, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0154, 0.0191, 0.0158, 0.0219, 0.0186, 0.0144, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:27:38,142 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 7600, loss[loss=0.2542, simple_loss=0.3343, pruned_loss=0.0871, over 16891.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.336, pruned_loss=0.09676, over 3058433.80 frames. ], batch size: 96, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:28:05,542 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:28:55,058 INFO [train.py:904] (3/8) Epoch 4, batch 7650, loss[loss=0.2525, simple_loss=0.3299, pruned_loss=0.08752, over 16383.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3373, pruned_loss=0.09825, over 3049031.96 frames. ], batch size: 35, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:29:09,888 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-28 04:30:08,813 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.664e+02 4.200e+02 5.231e+02 7.341e+02 1.286e+03, threshold=1.046e+03, percent-clipped=4.0 2023-04-28 04:30:09,972 INFO [train.py:904] (3/8) Epoch 4, batch 7700, loss[loss=0.2884, simple_loss=0.3604, pruned_loss=0.1082, over 16859.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.3382, pruned_loss=0.09906, over 3061551.42 frames. ], batch size: 116, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:31:26,738 INFO [train.py:904] (3/8) Epoch 4, batch 7750, loss[loss=0.2287, simple_loss=0.3159, pruned_loss=0.07076, over 16466.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3388, pruned_loss=0.09924, over 3059839.59 frames. ], batch size: 68, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:32:17,197 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0055, 3.9948, 4.5094, 4.4745, 4.4406, 4.0252, 4.1305, 3.9640], device='cuda:3'), covar=tensor([0.0238, 0.0370, 0.0286, 0.0362, 0.0410, 0.0284, 0.0746, 0.0427], device='cuda:3'), in_proj_covar=tensor([0.0216, 0.0210, 0.0216, 0.0218, 0.0261, 0.0225, 0.0329, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 04:32:40,399 INFO [optim.py:368] (3/8) 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 04:32:42,180 INFO [train.py:904] (3/8) Epoch 4, batch 7800, loss[loss=0.2749, simple_loss=0.3387, pruned_loss=0.1056, over 16701.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3394, pruned_loss=0.1002, over 3056041.99 frames. ], batch size: 62, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:33:10,635 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7636, 1.2801, 1.5482, 1.5633, 1.7708, 1.7985, 1.3965, 1.7848], device='cuda:3'), covar=tensor([0.0074, 0.0157, 0.0091, 0.0107, 0.0079, 0.0057, 0.0147, 0.0045], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0131, 0.0118, 0.0111, 0.0116, 0.0083, 0.0129, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 04:33:58,979 INFO [train.py:904] (3/8) Epoch 4, batch 7850, loss[loss=0.2809, simple_loss=0.3526, pruned_loss=0.1046, over 16466.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3402, pruned_loss=0.1008, over 3023770.87 frames. ], batch size: 75, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:34:50,563 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:12,442 INFO [optim.py:368] (3/8) 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 04:35:14,726 INFO [train.py:904] (3/8) Epoch 4, batch 7900, loss[loss=0.2659, simple_loss=0.3444, pruned_loss=0.09373, over 16878.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3394, pruned_loss=0.09983, over 3027201.90 frames. ], batch size: 116, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:35:28,530 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:42,173 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2244, 2.1456, 2.1036, 3.6457, 1.7700, 3.0713, 2.2129, 2.0730], device='cuda:3'), covar=tensor([0.0516, 0.1523, 0.0874, 0.0278, 0.2517, 0.0567, 0.1440, 0.1829], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0280, 0.0234, 0.0291, 0.0349, 0.0258, 0.0255, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:35:43,917 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6766, 2.6734, 1.7573, 2.7238, 2.0723, 2.7687, 1.9006, 2.3654], device='cuda:3'), covar=tensor([0.0123, 0.0290, 0.1056, 0.0059, 0.0620, 0.0548, 0.0932, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0144, 0.0176, 0.0076, 0.0160, 0.0173, 0.0183, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 04:36:34,670 INFO [train.py:904] (3/8) Epoch 4, batch 7950, loss[loss=0.2805, simple_loss=0.3355, pruned_loss=0.1128, over 11680.00 frames. ], tot_loss[loss=0.2692, simple_loss=0.3392, pruned_loss=0.09957, over 3043961.82 frames. ], batch size: 246, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:37:05,097 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:37:38,435 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6432, 3.6542, 2.9011, 2.2814, 2.6434, 2.1937, 3.7786, 3.9640], device='cuda:3'), covar=tensor([0.2020, 0.0608, 0.1207, 0.1401, 0.2341, 0.1309, 0.0412, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0241, 0.0260, 0.0232, 0.0302, 0.0192, 0.0231, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:37:49,209 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 8000, loss[loss=0.2602, simple_loss=0.3343, pruned_loss=0.09307, over 16716.00 frames. ], tot_loss[loss=0.268, simple_loss=0.338, pruned_loss=0.09898, over 3045856.76 frames. ], batch size: 134, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:39:04,876 INFO [train.py:904] (3/8) Epoch 4, batch 8050, loss[loss=0.3409, simple_loss=0.3856, pruned_loss=0.1481, over 11624.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.338, pruned_loss=0.09872, over 3050970.19 frames. ], batch size: 246, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:40:21,960 INFO [optim.py:368] (3/8) 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 04:40:23,266 INFO [train.py:904] (3/8) Epoch 4, batch 8100, loss[loss=0.2377, simple_loss=0.3142, pruned_loss=0.08061, over 16880.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3368, pruned_loss=0.09708, over 3074056.08 frames. ], batch size: 116, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:41:41,522 INFO [train.py:904] (3/8) Epoch 4, batch 8150, loss[loss=0.3225, simple_loss=0.3646, pruned_loss=0.1402, over 11651.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3336, pruned_loss=0.09542, over 3074142.38 frames. ], batch size: 247, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:42:34,019 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:42:41,333 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9940, 2.6955, 2.6627, 1.7343, 2.8337, 2.8138, 2.3843, 2.2713], device='cuda:3'), covar=tensor([0.0606, 0.0135, 0.0161, 0.0888, 0.0091, 0.0094, 0.0321, 0.0425], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0087, 0.0087, 0.0146, 0.0073, 0.0077, 0.0117, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 04:42:57,099 INFO [optim.py:368] (3/8) 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 04:42:59,072 INFO [train.py:904] (3/8) Epoch 4, batch 8200, loss[loss=0.243, simple_loss=0.3147, pruned_loss=0.08567, over 16570.00 frames. ], tot_loss[loss=0.2592, simple_loss=0.3304, pruned_loss=0.094, over 3087709.04 frames. ], batch size: 62, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:43:24,842 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3829, 3.4971, 1.6194, 3.4698, 2.4420, 3.6636, 1.8724, 2.8215], device='cuda:3'), covar=tensor([0.0107, 0.0202, 0.1381, 0.0050, 0.0668, 0.0286, 0.1259, 0.0482], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0144, 0.0175, 0.0078, 0.0158, 0.0175, 0.0183, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 04:43:53,411 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 8250, loss[loss=0.2442, simple_loss=0.313, pruned_loss=0.08775, over 11834.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3299, pruned_loss=0.09226, over 3049813.43 frames. ], batch size: 247, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:44:49,018 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:45:43,965 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 8300, loss[loss=0.2192, simple_loss=0.3028, pruned_loss=0.06779, over 16722.00 frames. ], tot_loss[loss=0.2508, simple_loss=0.326, pruned_loss=0.08783, over 3064439.56 frames. ], batch size: 57, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:07,345 INFO [train.py:904] (3/8) Epoch 4, batch 8350, loss[loss=0.2499, simple_loss=0.3287, pruned_loss=0.08551, over 16753.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3249, pruned_loss=0.08525, over 3059642.51 frames. ], batch size: 124, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:25,851 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4096, 4.2778, 4.7450, 4.7311, 4.7486, 4.4284, 4.4668, 4.2672], device='cuda:3'), covar=tensor([0.0200, 0.0311, 0.0345, 0.0348, 0.0313, 0.0234, 0.0674, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0201, 0.0210, 0.0214, 0.0254, 0.0222, 0.0317, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 04:47:44,532 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:47:59,762 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3505, 3.2609, 3.3312, 3.4777, 3.5066, 3.1746, 3.4830, 3.5409], device='cuda:3'), covar=tensor([0.0691, 0.0617, 0.1035, 0.0483, 0.0498, 0.2058, 0.0657, 0.0463], device='cuda:3'), in_proj_covar=tensor([0.0338, 0.0409, 0.0521, 0.0420, 0.0310, 0.0301, 0.0336, 0.0343], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:48:29,398 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 3.176e+02 3.884e+02 4.976e+02 1.172e+03, threshold=7.768e+02, percent-clipped=5.0 2023-04-28 04:48:30,621 INFO [train.py:904] (3/8) Epoch 4, batch 8400, loss[loss=0.2153, simple_loss=0.2959, pruned_loss=0.06734, over 12405.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3212, pruned_loss=0.08218, over 3054084.30 frames. ], batch size: 246, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:48:42,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 04:49:26,924 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:49:51,920 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1388, 3.8941, 3.7529, 4.2674, 4.4482, 3.9573, 4.3707, 4.4428], device='cuda:3'), covar=tensor([0.0788, 0.0812, 0.2117, 0.0819, 0.0661, 0.1022, 0.0765, 0.0609], device='cuda:3'), in_proj_covar=tensor([0.0331, 0.0405, 0.0513, 0.0415, 0.0305, 0.0299, 0.0326, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 04:49:53,139 INFO [train.py:904] (3/8) Epoch 4, batch 8450, loss[loss=0.1903, simple_loss=0.2835, pruned_loss=0.04857, over 16856.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3188, pruned_loss=0.08026, over 3032227.45 frames. ], batch size: 116, lr: 1.56e-02, grad_scale: 4.0 2023-04-28 04:51:13,821 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 8500, loss[loss=0.1901, simple_loss=0.2801, pruned_loss=0.05003, over 16756.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3136, pruned_loss=0.07618, over 3047408.83 frames. ], batch size: 102, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:52:20,511 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-28 04:52:39,866 INFO [train.py:904] (3/8) Epoch 4, batch 8550, loss[loss=0.2275, simple_loss=0.3162, pruned_loss=0.06937, over 16879.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3106, pruned_loss=0.07444, over 3033712.81 frames. ], batch size: 96, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:53:06,238 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-28 04:53:09,930 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:53:26,717 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 04:54:21,291 INFO [optim.py:368] (3/8) 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 04:54:21,306 INFO [train.py:904] (3/8) Epoch 4, batch 8600, loss[loss=0.216, simple_loss=0.3089, pruned_loss=0.06156, over 16709.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3109, pruned_loss=0.07334, over 3016859.09 frames. ], batch size: 89, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:54:50,964 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:34,765 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-04-28 04:55:58,610 INFO [train.py:904] (3/8) Epoch 4, batch 8650, loss[loss=0.1971, simple_loss=0.289, pruned_loss=0.05256, over 16803.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3081, pruned_loss=0.07131, over 3014740.06 frames. ], batch size: 124, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:44,932 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 4, batch 8700, loss[loss=0.2263, simple_loss=0.3089, pruned_loss=0.07185, over 15346.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3049, pruned_loss=0.0698, over 3014081.45 frames. ], batch size: 191, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:58:36,851 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:50,366 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:59:20,254 INFO [train.py:904] (3/8) Epoch 4, batch 8750, loss[loss=0.239, simple_loss=0.3308, pruned_loss=0.07364, over 16753.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3041, pruned_loss=0.06912, over 3024556.78 frames. ], batch size: 83, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:01:04,987 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:01:10,820 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4166, 3.2620, 2.6317, 2.1172, 2.3243, 1.9973, 3.3321, 3.4097], device='cuda:3'), covar=tensor([0.1988, 0.0697, 0.1241, 0.1419, 0.1779, 0.1343, 0.0439, 0.0535], device='cuda:3'), in_proj_covar=tensor([0.0263, 0.0234, 0.0249, 0.0225, 0.0253, 0.0188, 0.0218, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:01:14,441 INFO [train.py:904] (3/8) Epoch 4, batch 8800, loss[loss=0.2106, simple_loss=0.2965, pruned_loss=0.06238, over 16768.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3019, pruned_loss=0.06739, over 3039190.48 frames. ], batch size: 124, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:01:15,986 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.302e+02 3.583e+02 4.376e+02 5.208e+02 1.205e+03, threshold=8.753e+02, percent-clipped=6.0 2023-04-28 05:02:23,544 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 8850, loss[loss=0.1978, simple_loss=0.2772, pruned_loss=0.05925, over 12286.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.304, pruned_loss=0.06617, over 3046761.41 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:04:32,317 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 8900, loss[loss=0.232, simple_loss=0.3131, pruned_loss=0.07548, over 12679.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3041, pruned_loss=0.06489, over 3054711.87 frames. ], batch size: 250, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:04:49,517 INFO [optim.py:368] (3/8) 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 05:05:08,765 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5965, 1.4041, 1.8096, 2.6762, 2.3693, 2.5951, 1.5866, 2.5174], device='cuda:3'), covar=tensor([0.0056, 0.0247, 0.0163, 0.0083, 0.0109, 0.0095, 0.0232, 0.0069], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0133, 0.0118, 0.0112, 0.0116, 0.0081, 0.0131, 0.0072], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 05:05:33,504 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0648, 4.8755, 4.7259, 2.4450, 3.9704, 3.2921, 4.2175, 4.3630], device='cuda:3'), covar=tensor([0.0161, 0.0279, 0.0286, 0.1312, 0.0427, 0.0679, 0.0463, 0.0613], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0115, 0.0155, 0.0140, 0.0133, 0.0128, 0.0138, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 05:06:47,800 INFO [train.py:904] (3/8) Epoch 4, batch 8950, loss[loss=0.2148, simple_loss=0.2961, pruned_loss=0.06674, over 17001.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3038, pruned_loss=0.0656, over 3061121.81 frames. ], batch size: 109, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:07:01,607 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 05:08:35,742 INFO [train.py:904] (3/8) Epoch 4, batch 9000, loss[loss=0.1936, simple_loss=0.2776, pruned_loss=0.05483, over 16938.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.3, pruned_loss=0.06347, over 3065739.86 frames. ], batch size: 116, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:08:35,742 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 05:08:45,782 INFO [train.py:938] (3/8) Epoch 4, validation: loss=0.1802, simple_loss=0.283, pruned_loss=0.0387, over 944034.00 frames. 2023-04-28 05:08:45,783 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 05:08:49,857 INFO [optim.py:368] (3/8) 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 05:09:11,772 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:09:43,630 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:29,876 INFO [train.py:904] (3/8) Epoch 4, batch 9050, loss[loss=0.2156, simple_loss=0.3012, pruned_loss=0.06498, over 16792.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.3016, pruned_loss=0.06452, over 3076898.89 frames. ], batch size: 124, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:11:17,790 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:21,903 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:12:14,620 INFO [train.py:904] (3/8) Epoch 4, batch 9100, loss[loss=0.195, simple_loss=0.2815, pruned_loss=0.05431, over 16510.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.301, pruned_loss=0.06467, over 3069692.97 frames. ], batch size: 36, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:12:18,749 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 3.376e+02 4.042e+02 5.182e+02 1.418e+03, threshold=8.084e+02, percent-clipped=5.0 2023-04-28 05:12:45,208 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9384, 2.6789, 2.6658, 1.8370, 2.4767, 2.5777, 2.6252, 1.7500], device='cuda:3'), covar=tensor([0.0259, 0.0023, 0.0040, 0.0183, 0.0045, 0.0042, 0.0032, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0049, 0.0056, 0.0108, 0.0055, 0.0062, 0.0058, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 05:14:15,379 INFO [train.py:904] (3/8) Epoch 4, batch 9150, loss[loss=0.2201, simple_loss=0.301, pruned_loss=0.06959, over 12242.00 frames. ], tot_loss[loss=0.215, simple_loss=0.301, pruned_loss=0.06445, over 3056973.24 frames. ], batch size: 250, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:14:53,072 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6567, 3.5688, 3.0381, 1.7964, 2.6567, 2.2375, 2.9914, 3.3079], device='cuda:3'), covar=tensor([0.0342, 0.0410, 0.0495, 0.1598, 0.0746, 0.0885, 0.0797, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0114, 0.0152, 0.0140, 0.0133, 0.0126, 0.0139, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 05:15:01,352 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5043, 1.9944, 1.8179, 1.7790, 2.4611, 2.1530, 2.6401, 2.6776], device='cuda:3'), covar=tensor([0.0022, 0.0192, 0.0210, 0.0240, 0.0109, 0.0183, 0.0059, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0064, 0.0142, 0.0142, 0.0141, 0.0134, 0.0144, 0.0099, 0.0116], device='cuda:3'), out_proj_covar=tensor([8.0100e-05, 1.8250e-04, 1.7791e-04, 1.7611e-04, 1.7236e-04, 1.8517e-04, 1.2081e-04, 1.4859e-04], device='cuda:3') 2023-04-28 05:15:41,142 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:15:43,179 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 9200, loss[loss=0.188, simple_loss=0.2785, pruned_loss=0.0487, over 15341.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2966, pruned_loss=0.0632, over 3080068.36 frames. ], batch size: 191, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:16:04,311 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.381e+02 3.414e+02 4.371e+02 6.281e+02 1.474e+03, threshold=8.741e+02, percent-clipped=12.0 2023-04-28 05:17:35,079 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:17:35,648 INFO [train.py:904] (3/8) Epoch 4, batch 9250, loss[loss=0.1911, simple_loss=0.2917, pruned_loss=0.04529, over 16676.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2964, pruned_loss=0.06317, over 3075707.19 frames. ], batch size: 89, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:18:15,127 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0465, 2.5589, 2.5213, 4.6101, 1.9314, 3.8352, 2.5787, 2.4735], device='cuda:3'), covar=tensor([0.0417, 0.1453, 0.0790, 0.0173, 0.2578, 0.0480, 0.1414, 0.2007], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0281, 0.0231, 0.0287, 0.0345, 0.0256, 0.0257, 0.0330], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:19:26,128 INFO [train.py:904] (3/8) Epoch 4, batch 9300, loss[loss=0.2118, simple_loss=0.289, pruned_loss=0.06731, over 12580.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2947, pruned_loss=0.06237, over 3066213.62 frames. ], batch size: 248, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:30,018 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.034e+02 3.554e+02 4.321e+02 7.893e+02, threshold=7.107e+02, percent-clipped=0.0 2023-04-28 05:19:51,190 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8803, 2.3584, 2.5794, 4.4848, 1.9531, 3.5787, 2.5004, 2.4744], device='cuda:3'), covar=tensor([0.0421, 0.1432, 0.0728, 0.0191, 0.2605, 0.0554, 0.1385, 0.1930], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0281, 0.0230, 0.0287, 0.0345, 0.0256, 0.0256, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:20:06,140 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:11,784 INFO [train.py:904] (3/8) Epoch 4, batch 9350, loss[loss=0.2395, simple_loss=0.3168, pruned_loss=0.08113, over 16352.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2943, pruned_loss=0.06187, over 3086083.94 frames. ], batch size: 146, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:21:22,319 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8687, 2.9648, 3.0061, 1.9122, 2.7731, 2.9460, 3.0219, 1.7736], device='cuda:3'), covar=tensor([0.0355, 0.0026, 0.0037, 0.0262, 0.0050, 0.0058, 0.0037, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0050, 0.0056, 0.0110, 0.0056, 0.0063, 0.0059, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 05:21:41,267 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:48,577 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:05,657 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:30,263 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:51,014 INFO [train.py:904] (3/8) Epoch 4, batch 9400, loss[loss=0.1755, simple_loss=0.2559, pruned_loss=0.04762, over 12250.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2946, pruned_loss=0.06205, over 3069211.74 frames. ], batch size: 247, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:22:57,590 INFO [optim.py:368] (3/8) 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 05:23:41,186 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:32,547 INFO [train.py:904] (3/8) Epoch 4, batch 9450, loss[loss=0.1859, simple_loss=0.2769, pruned_loss=0.0474, over 16681.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.297, pruned_loss=0.06269, over 3072457.15 frames. ], batch size: 83, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:24:47,352 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3330, 3.2055, 3.3502, 3.5024, 3.4936, 3.2277, 3.4807, 3.5242], device='cuda:3'), covar=tensor([0.0764, 0.0568, 0.0949, 0.0499, 0.0538, 0.1442, 0.0724, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0385, 0.0484, 0.0395, 0.0298, 0.0285, 0.0315, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:25:54,250 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 9500, loss[loss=0.2015, simple_loss=0.289, pruned_loss=0.057, over 16711.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2962, pruned_loss=0.06206, over 3095353.89 frames. ], batch size: 134, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:26:21,223 INFO [optim.py:368] (3/8) 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 05:27:31,419 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:28:04,403 INFO [train.py:904] (3/8) Epoch 4, batch 9550, loss[loss=0.2419, simple_loss=0.327, pruned_loss=0.07841, over 16868.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2957, pruned_loss=0.0622, over 3098069.90 frames. ], batch size: 102, lr: 1.53e-02, grad_scale: 2.0 2023-04-28 05:29:46,541 INFO [train.py:904] (3/8) Epoch 4, batch 9600, loss[loss=0.2448, simple_loss=0.3325, pruned_loss=0.07849, over 16265.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2981, pruned_loss=0.06359, over 3086030.00 frames. ], batch size: 165, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:29:50,837 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 05:29:52,055 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.278e+02 3.502e+02 4.489e+02 5.494e+02 1.109e+03, threshold=8.977e+02, percent-clipped=3.0 2023-04-28 05:31:23,817 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8780, 3.8302, 3.4510, 1.8043, 2.8343, 2.3366, 3.1493, 3.5363], device='cuda:3'), covar=tensor([0.0308, 0.0447, 0.0443, 0.1575, 0.0698, 0.0911, 0.0809, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0114, 0.0155, 0.0141, 0.0134, 0.0127, 0.0139, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 05:31:33,068 INFO [train.py:904] (3/8) Epoch 4, batch 9650, loss[loss=0.2278, simple_loss=0.3024, pruned_loss=0.0766, over 12205.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2998, pruned_loss=0.06402, over 3072197.87 frames. ], batch size: 246, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:32:17,596 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:17,714 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1895, 3.2687, 3.2350, 1.6217, 3.5305, 3.5801, 3.0334, 2.6788], device='cuda:3'), covar=tensor([0.0710, 0.0124, 0.0141, 0.1144, 0.0064, 0.0059, 0.0258, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0080, 0.0073, 0.0136, 0.0066, 0.0070, 0.0106, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 05:32:25,199 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:47,563 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6821, 3.6808, 3.4214, 1.6710, 2.8610, 2.1502, 3.2179, 3.2331], device='cuda:3'), covar=tensor([0.0246, 0.0346, 0.0422, 0.1708, 0.0646, 0.0947, 0.0638, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0113, 0.0154, 0.0142, 0.0134, 0.0127, 0.0139, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 05:33:12,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0936, 3.1676, 3.5543, 3.5408, 3.5408, 3.1995, 3.3701, 3.3017], device='cuda:3'), covar=tensor([0.0288, 0.0498, 0.0484, 0.0465, 0.0399, 0.0376, 0.0641, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0190, 0.0199, 0.0202, 0.0232, 0.0208, 0.0292, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-28 05:33:21,183 INFO [train.py:904] (3/8) Epoch 4, batch 9700, loss[loss=0.1852, simple_loss=0.2789, pruned_loss=0.04571, over 16843.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.298, pruned_loss=0.06333, over 3063677.02 frames. ], batch size: 76, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:33:26,546 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.999e+02 3.154e+02 3.870e+02 5.322e+02 1.510e+03, threshold=7.740e+02, percent-clipped=2.0 2023-04-28 05:33:53,098 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:54,823 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:57,761 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7686, 2.8851, 2.5609, 4.2344, 3.8903, 3.9379, 1.4676, 2.9844], device='cuda:3'), covar=tensor([0.1367, 0.0543, 0.1068, 0.0083, 0.0168, 0.0366, 0.1407, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0135, 0.0163, 0.0074, 0.0139, 0.0158, 0.0159, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 05:35:03,565 INFO [train.py:904] (3/8) Epoch 4, batch 9750, loss[loss=0.209, simple_loss=0.2984, pruned_loss=0.05975, over 16397.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2966, pruned_loss=0.06303, over 3070277.43 frames. ], batch size: 146, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:35:38,794 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:36:17,646 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7932, 2.2557, 2.2835, 4.4239, 1.8404, 3.3360, 2.2986, 2.2525], device='cuda:3'), covar=tensor([0.0500, 0.1609, 0.0881, 0.0204, 0.2839, 0.0654, 0.1643, 0.2219], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0283, 0.0231, 0.0284, 0.0343, 0.0257, 0.0255, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:36:45,112 INFO [train.py:904] (3/8) Epoch 4, batch 9800, loss[loss=0.2004, simple_loss=0.2828, pruned_loss=0.05896, over 12192.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2963, pruned_loss=0.0619, over 3071986.46 frames. ], batch size: 247, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:36:51,067 INFO [optim.py:368] (3/8) 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 05:36:56,602 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:24,808 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:37,837 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:17,356 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 9850, loss[loss=0.1934, simple_loss=0.2868, pruned_loss=0.05004, over 16131.00 frames. ], tot_loss[loss=0.211, simple_loss=0.298, pruned_loss=0.06205, over 3076795.72 frames. ], batch size: 165, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:38:31,195 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8654, 1.4546, 2.0997, 2.7079, 2.5259, 3.0477, 1.5738, 2.9271], device='cuda:3'), covar=tensor([0.0049, 0.0201, 0.0135, 0.0100, 0.0093, 0.0047, 0.0213, 0.0052], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0132, 0.0118, 0.0112, 0.0114, 0.0080, 0.0130, 0.0071], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 05:39:08,895 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2559, 2.0827, 2.2003, 3.5813, 1.7292, 2.8273, 2.1479, 1.9137], device='cuda:3'), covar=tensor([0.0460, 0.1654, 0.0792, 0.0294, 0.2773, 0.0677, 0.1530, 0.2184], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0280, 0.0228, 0.0283, 0.0342, 0.0254, 0.0253, 0.0327], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:39:31,341 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:39,128 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:02,716 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 05:40:05,814 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:40:21,696 INFO [train.py:904] (3/8) Epoch 4, batch 9900, loss[loss=0.2023, simple_loss=0.2997, pruned_loss=0.05251, over 16730.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2985, pruned_loss=0.062, over 3072863.00 frames. ], batch size: 83, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:40:27,910 INFO [optim.py:368] (3/8) 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 05:40:29,192 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1003, 3.4414, 3.1142, 4.9891, 4.7003, 4.5700, 1.6457, 3.8161], device='cuda:3'), covar=tensor([0.1157, 0.0446, 0.0815, 0.0058, 0.0125, 0.0219, 0.1281, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0134, 0.0162, 0.0073, 0.0139, 0.0157, 0.0156, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 05:40:45,503 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 05:41:38,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8056, 4.1504, 3.8752, 3.9847, 3.6923, 3.6847, 3.8389, 4.0328], device='cuda:3'), covar=tensor([0.0700, 0.0716, 0.0904, 0.0445, 0.0609, 0.1177, 0.0619, 0.0997], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0417, 0.0348, 0.0270, 0.0266, 0.0278, 0.0334, 0.0301], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:41:39,098 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0277, 2.5043, 2.3053, 3.2525, 2.8838, 3.2391, 1.7544, 2.7546], device='cuda:3'), covar=tensor([0.1072, 0.0480, 0.0893, 0.0078, 0.0162, 0.0386, 0.1149, 0.0625], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0134, 0.0163, 0.0074, 0.0139, 0.0158, 0.0158, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 05:41:56,835 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:42:18,075 INFO [train.py:904] (3/8) Epoch 4, batch 9950, loss[loss=0.1767, simple_loss=0.2751, pruned_loss=0.03915, over 16882.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.3007, pruned_loss=0.06258, over 3074315.62 frames. ], batch size: 102, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:43:13,957 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 4, batch 10000, loss[loss=0.2157, simple_loss=0.3061, pruned_loss=0.06271, over 16680.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.298, pruned_loss=0.06116, over 3082587.67 frames. ], batch size: 134, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:44:26,638 INFO [optim.py:368] (3/8) 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 05:44:33,376 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9592, 3.6887, 3.9178, 4.1048, 4.1770, 3.7031, 4.1731, 4.1508], device='cuda:3'), covar=tensor([0.0744, 0.0740, 0.1221, 0.0520, 0.0440, 0.1007, 0.0430, 0.0393], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0407, 0.0503, 0.0408, 0.0305, 0.0297, 0.0330, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:45:01,303 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:01,343 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:54,672 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9136, 2.6674, 2.6036, 1.6428, 2.7985, 2.8551, 2.5741, 2.3362], device='cuda:3'), covar=tensor([0.0755, 0.0134, 0.0154, 0.1099, 0.0111, 0.0110, 0.0261, 0.0421], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0085, 0.0077, 0.0143, 0.0070, 0.0073, 0.0111, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 05:46:02,920 INFO [train.py:904] (3/8) Epoch 4, batch 10050, loss[loss=0.2182, simple_loss=0.3089, pruned_loss=0.06374, over 15439.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2973, pruned_loss=0.06073, over 3080932.29 frames. ], batch size: 191, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:46:07,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1816, 2.1303, 2.0604, 3.5156, 1.7519, 2.9275, 2.0736, 1.9793], device='cuda:3'), covar=tensor([0.0481, 0.1482, 0.0836, 0.0299, 0.2603, 0.0636, 0.1618, 0.1963], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0286, 0.0232, 0.0290, 0.0345, 0.0259, 0.0261, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:46:38,831 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:38,802 INFO [train.py:904] (3/8) Epoch 4, batch 10100, loss[loss=0.1917, simple_loss=0.2821, pruned_loss=0.05063, over 16904.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2985, pruned_loss=0.06175, over 3081986.36 frames. ], batch size: 96, lr: 1.52e-02, grad_scale: 8.0 2023-04-28 05:47:39,292 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,161 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,814 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.191e+02 3.527e+02 4.401e+02 5.596e+02 1.013e+03, threshold=8.802e+02, percent-clipped=4.0 2023-04-28 05:48:27,676 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 0, loss[loss=0.2172, simple_loss=0.2971, pruned_loss=0.06865, over 17169.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2971, pruned_loss=0.06865, over 17169.00 frames. ], batch size: 46, lr: 1.42e-02, grad_scale: 8.0 2023-04-28 05:49:25,172 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 05:49:32,552 INFO [train.py:938] (3/8) Epoch 5, validation: loss=0.1789, simple_loss=0.2817, pruned_loss=0.03802, over 944034.00 frames. 2023-04-28 05:49:32,552 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 05:49:49,567 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5601, 3.5595, 2.7081, 2.3794, 2.6679, 2.2621, 3.5076, 3.5387], device='cuda:3'), covar=tensor([0.1657, 0.0494, 0.1034, 0.1212, 0.1491, 0.1225, 0.0390, 0.0551], device='cuda:3'), in_proj_covar=tensor([0.0259, 0.0238, 0.0253, 0.0227, 0.0234, 0.0190, 0.0220, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:50:11,884 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 50, loss[loss=0.2145, simple_loss=0.3092, pruned_loss=0.05988, over 17283.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3257, pruned_loss=0.09546, over 747342.82 frames. ], batch size: 52, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:50:43,213 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 05:51:09,261 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 05:51:29,204 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 100, loss[loss=0.2304, simple_loss=0.3143, pruned_loss=0.07328, over 16755.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3178, pruned_loss=0.08944, over 1317258.46 frames. ], batch size: 57, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:51:59,801 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4020, 4.0107, 3.6827, 1.9389, 3.2051, 2.3861, 3.7067, 3.8030], device='cuda:3'), covar=tensor([0.0262, 0.0443, 0.0456, 0.1524, 0.0581, 0.0969, 0.0625, 0.0884], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0115, 0.0153, 0.0142, 0.0132, 0.0125, 0.0135, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 05:52:07,391 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:52:31,101 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3211, 2.4849, 2.5133, 4.8554, 1.9735, 3.5458, 2.4289, 2.4441], device='cuda:3'), covar=tensor([0.0397, 0.1717, 0.0834, 0.0196, 0.2763, 0.0723, 0.1613, 0.2365], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0296, 0.0239, 0.0298, 0.0353, 0.0267, 0.0264, 0.0347], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 05:52:59,300 INFO [train.py:904] (3/8) Epoch 5, batch 150, loss[loss=0.2138, simple_loss=0.298, pruned_loss=0.06485, over 17117.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3138, pruned_loss=0.08656, over 1760367.37 frames. ], batch size: 49, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:53:08,164 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.263e+02 3.680e+02 4.214e+02 5.175e+02 1.144e+03, threshold=8.427e+02, percent-clipped=1.0 2023-04-28 05:53:12,804 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6360, 3.7495, 1.5478, 3.8087, 2.4536, 3.8575, 1.8410, 2.9406], device='cuda:3'), covar=tensor([0.0089, 0.0204, 0.1438, 0.0064, 0.0666, 0.0315, 0.1227, 0.0481], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0144, 0.0174, 0.0080, 0.0157, 0.0172, 0.0184, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 05:53:44,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0504, 5.7190, 5.8145, 5.6256, 5.6919, 6.1839, 5.9147, 5.6168], device='cuda:3'), covar=tensor([0.0773, 0.1301, 0.1105, 0.1636, 0.2286, 0.0858, 0.0951, 0.1889], device='cuda:3'), in_proj_covar=tensor([0.0257, 0.0377, 0.0362, 0.0327, 0.0426, 0.0385, 0.0297, 0.0432], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 05:54:09,222 INFO [train.py:904] (3/8) Epoch 5, batch 200, loss[loss=0.1946, simple_loss=0.2677, pruned_loss=0.06076, over 16836.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3106, pruned_loss=0.08357, over 2107484.85 frames. ], batch size: 39, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:54:40,856 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2382, 2.5914, 1.9770, 2.1041, 2.9399, 2.6827, 3.3522, 3.1412], device='cuda:3'), covar=tensor([0.0024, 0.0138, 0.0206, 0.0201, 0.0089, 0.0144, 0.0075, 0.0077], device='cuda:3'), in_proj_covar=tensor([0.0070, 0.0145, 0.0147, 0.0145, 0.0138, 0.0146, 0.0108, 0.0122], device='cuda:3'), out_proj_covar=tensor([8.5322e-05, 1.8420e-04, 1.8202e-04, 1.7898e-04, 1.7649e-04, 1.8619e-04, 1.3156e-04, 1.5391e-04], device='cuda:3') 2023-04-28 05:55:01,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9750, 4.2824, 1.9047, 4.5351, 2.7854, 4.4842, 2.1248, 3.0888], device='cuda:3'), covar=tensor([0.0110, 0.0184, 0.1358, 0.0030, 0.0695, 0.0272, 0.1316, 0.0555], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0146, 0.0174, 0.0081, 0.0158, 0.0175, 0.0185, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 05:55:13,917 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:17,652 INFO [train.py:904] (3/8) Epoch 5, batch 250, loss[loss=0.2594, simple_loss=0.3264, pruned_loss=0.09623, over 16431.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3075, pruned_loss=0.08238, over 2385609.46 frames. ], batch size: 68, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:55:17,991 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:25,559 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 3.707e+02 4.511e+02 5.174e+02 8.449e+02, threshold=9.022e+02, percent-clipped=1.0 2023-04-28 05:55:38,044 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:50,359 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:24,970 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:26,474 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:27,683 INFO [train.py:904] (3/8) Epoch 5, batch 300, loss[loss=0.2056, simple_loss=0.2839, pruned_loss=0.06363, over 16010.00 frames. ], tot_loss[loss=0.233, simple_loss=0.305, pruned_loss=0.08048, over 2589541.52 frames. ], batch size: 35, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:56:47,600 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-28 05:56:58,126 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:05,040 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:07,287 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:39,677 INFO [train.py:904] (3/8) Epoch 5, batch 350, loss[loss=0.2066, simple_loss=0.2725, pruned_loss=0.07029, over 16816.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3016, pruned_loss=0.07834, over 2753870.90 frames. ], batch size: 102, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:57:48,102 INFO [optim.py:368] (3/8) 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 05:57:54,089 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:58:14,879 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:28,101 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:44,238 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:48,863 INFO [train.py:904] (3/8) Epoch 5, batch 400, loss[loss=0.1925, simple_loss=0.2712, pruned_loss=0.0569, over 16968.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.2989, pruned_loss=0.0771, over 2877639.31 frames. ], batch size: 41, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 05:58:59,023 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:36,591 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:51,074 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 450, loss[loss=0.2121, simple_loss=0.2767, pruned_loss=0.07374, over 16923.00 frames. ], tot_loss[loss=0.225, simple_loss=0.2974, pruned_loss=0.07632, over 2966917.70 frames. ], batch size: 109, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:00:03,024 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-28 06:00:09,409 INFO [optim.py:368] (3/8) 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 06:00:11,843 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 500, loss[loss=0.222, simple_loss=0.2847, pruned_loss=0.07963, over 16671.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2955, pruned_loss=0.07442, over 3049197.52 frames. ], batch size: 89, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:01:12,319 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:16,677 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:02:14,489 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 550, loss[loss=0.2314, simple_loss=0.301, pruned_loss=0.08088, over 15592.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2949, pruned_loss=0.074, over 3110340.49 frames. ], batch size: 191, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:02:27,242 INFO [optim.py:368] (3/8) 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 06:02:35,984 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:22,002 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:27,688 INFO [train.py:904] (3/8) Epoch 5, batch 600, loss[loss=0.1913, simple_loss=0.2703, pruned_loss=0.05615, over 16820.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2939, pruned_loss=0.07407, over 3161161.01 frames. ], batch size: 42, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:03:54,870 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:04:33,941 INFO [train.py:904] (3/8) Epoch 5, batch 650, loss[loss=0.2042, simple_loss=0.2901, pruned_loss=0.05914, over 17074.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2921, pruned_loss=0.07306, over 3200508.68 frames. ], batch size: 55, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:04:40,903 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:04:42,445 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.330e+02 3.128e+02 3.726e+02 4.799e+02 9.270e+02, threshold=7.452e+02, percent-clipped=1.0 2023-04-28 06:05:39,909 INFO [train.py:904] (3/8) Epoch 5, batch 700, loss[loss=0.211, simple_loss=0.2829, pruned_loss=0.06952, over 16843.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2915, pruned_loss=0.07205, over 3221135.00 frames. ], batch size: 42, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:05:48,991 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:12,111 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 06:06:49,380 INFO [train.py:904] (3/8) Epoch 5, batch 750, loss[loss=0.1959, simple_loss=0.2771, pruned_loss=0.05738, over 16980.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2917, pruned_loss=0.07261, over 3225334.78 frames. ], batch size: 41, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:06:52,746 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:55,089 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 06:07:58,338 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 06:07:58,821 INFO [train.py:904] (3/8) Epoch 5, batch 800, loss[loss=0.2575, simple_loss=0.3059, pruned_loss=0.1045, over 16804.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2918, pruned_loss=0.07202, over 3250567.42 frames. ], batch size: 124, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:07:59,096 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:08,623 INFO [train.py:904] (3/8) Epoch 5, batch 850, loss[loss=0.2223, simple_loss=0.3012, pruned_loss=0.07169, over 17113.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2911, pruned_loss=0.07124, over 3257348.47 frames. ], batch size: 48, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:09:16,401 INFO [optim.py:368] (3/8) 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 06:09:17,870 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:40,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5341, 3.8578, 4.4375, 2.8317, 3.9773, 4.4429, 4.0688, 2.3811], device='cuda:3'), covar=tensor([0.0262, 0.0027, 0.0017, 0.0212, 0.0032, 0.0021, 0.0024, 0.0239], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0059, 0.0060, 0.0113, 0.0059, 0.0066, 0.0062, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:10:16,011 INFO [train.py:904] (3/8) Epoch 5, batch 900, loss[loss=0.2286, simple_loss=0.3024, pruned_loss=0.07744, over 16661.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2893, pruned_loss=0.07057, over 3267375.25 frames. ], batch size: 62, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:10:45,074 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:27,562 INFO [train.py:904] (3/8) Epoch 5, batch 950, loss[loss=0.1955, simple_loss=0.2803, pruned_loss=0.05531, over 17082.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2887, pruned_loss=0.06981, over 3276119.25 frames. ], batch size: 53, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:11:34,481 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:11:35,283 INFO [optim.py:368] (3/8) 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 06:11:43,169 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2515, 4.0246, 4.2440, 4.5046, 4.5747, 4.0475, 4.2834, 4.5526], device='cuda:3'), covar=tensor([0.0835, 0.0775, 0.1231, 0.0490, 0.0484, 0.0906, 0.1124, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0385, 0.0472, 0.0607, 0.0479, 0.0362, 0.0350, 0.0378, 0.0401], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:11:53,684 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:01,065 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2773, 3.4221, 4.1687, 2.8073, 3.8198, 4.1076, 3.9897, 2.1195], device='cuda:3'), covar=tensor([0.0303, 0.0095, 0.0021, 0.0194, 0.0031, 0.0034, 0.0027, 0.0275], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0059, 0.0060, 0.0113, 0.0059, 0.0067, 0.0062, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:12:37,219 INFO [train.py:904] (3/8) Epoch 5, batch 1000, loss[loss=0.1854, simple_loss=0.2635, pruned_loss=0.05367, over 15810.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2877, pruned_loss=0.07024, over 3277021.15 frames. ], batch size: 35, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:12:41,513 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:35,704 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9517, 2.6011, 2.4336, 4.4783, 1.8387, 3.6442, 2.2814, 2.4546], device='cuda:3'), covar=tensor([0.0497, 0.1751, 0.0934, 0.0287, 0.2879, 0.0696, 0.1777, 0.2443], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0308, 0.0250, 0.0311, 0.0359, 0.0292, 0.0274, 0.0378], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:13:45,703 INFO [train.py:904] (3/8) Epoch 5, batch 1050, loss[loss=0.2574, simple_loss=0.308, pruned_loss=0.1034, over 16707.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2871, pruned_loss=0.06977, over 3294038.01 frames. ], batch size: 134, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:13:49,016 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:50,324 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:54,768 INFO [optim.py:368] (3/8) 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 06:14:22,059 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 06:14:56,088 INFO [train.py:904] (3/8) Epoch 5, batch 1100, loss[loss=0.2066, simple_loss=0.2755, pruned_loss=0.06889, over 16352.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2861, pruned_loss=0.06878, over 3301465.75 frames. ], batch size: 165, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:14:56,429 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:56,482 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:15:15,775 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:16:02,694 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:05,299 INFO [train.py:904] (3/8) Epoch 5, batch 1150, loss[loss=0.1984, simple_loss=0.2679, pruned_loss=0.0644, over 16856.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2856, pruned_loss=0.06814, over 3291058.49 frames. ], batch size: 96, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:16:12,831 INFO [optim.py:368] (3/8) 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 06:16:15,017 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:19,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9445, 4.2111, 2.0592, 4.4970, 2.6387, 4.4421, 2.1743, 3.0839], device='cuda:3'), covar=tensor([0.0106, 0.0177, 0.1492, 0.0039, 0.0786, 0.0272, 0.1389, 0.0572], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0151, 0.0172, 0.0085, 0.0158, 0.0181, 0.0183, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 06:17:05,704 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 06:17:07,044 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4883, 4.3000, 4.3964, 4.6760, 4.7704, 4.2928, 4.6165, 4.7445], device='cuda:3'), covar=tensor([0.0659, 0.0623, 0.1086, 0.0446, 0.0389, 0.0670, 0.0733, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0494, 0.0626, 0.0502, 0.0378, 0.0363, 0.0397, 0.0414], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:17:14,374 INFO [train.py:904] (3/8) Epoch 5, batch 1200, loss[loss=0.1918, simple_loss=0.2714, pruned_loss=0.05615, over 17238.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2855, pruned_loss=0.06797, over 3302290.07 frames. ], batch size: 44, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:17:21,141 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:17:34,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8936, 3.9428, 3.1813, 2.5875, 3.1887, 2.4575, 4.2855, 4.3326], device='cuda:3'), covar=tensor([0.1992, 0.0819, 0.1260, 0.1393, 0.2079, 0.1362, 0.0392, 0.0545], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0254, 0.0269, 0.0241, 0.0295, 0.0201, 0.0238, 0.0259], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:17:49,398 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 06:18:05,997 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 06:18:23,576 INFO [train.py:904] (3/8) Epoch 5, batch 1250, loss[loss=0.2678, simple_loss=0.3177, pruned_loss=0.109, over 16855.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2859, pruned_loss=0.06791, over 3313847.79 frames. ], batch size: 116, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:18:24,322 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 06:18:31,521 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.380e+02 4.131e+02 4.913e+02 1.055e+03, threshold=8.263e+02, percent-clipped=6.0 2023-04-28 06:19:24,972 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 06:19:30,805 INFO [train.py:904] (3/8) Epoch 5, batch 1300, loss[loss=0.2292, simple_loss=0.287, pruned_loss=0.0857, over 12031.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2866, pruned_loss=0.06863, over 3308907.02 frames. ], batch size: 247, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:19:38,976 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:05,883 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7911, 5.0241, 5.1622, 5.1938, 5.0365, 5.7131, 5.3381, 5.0223], device='cuda:3'), covar=tensor([0.0849, 0.1587, 0.1308, 0.1609, 0.2714, 0.0903, 0.1065, 0.1952], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0394, 0.0378, 0.0335, 0.0451, 0.0405, 0.0312, 0.0453], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:20:13,317 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0005, 5.5142, 5.6430, 5.5537, 5.4776, 6.0999, 5.7071, 5.4004], device='cuda:3'), covar=tensor([0.0611, 0.1369, 0.1264, 0.1451, 0.2260, 0.0781, 0.0951, 0.1804], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0392, 0.0377, 0.0334, 0.0448, 0.0403, 0.0311, 0.0450], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:20:38,016 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:20:42,275 INFO [train.py:904] (3/8) Epoch 5, batch 1350, loss[loss=0.2366, simple_loss=0.3065, pruned_loss=0.08333, over 11925.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2868, pruned_loss=0.06867, over 3312725.13 frames. ], batch size: 250, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:20:51,204 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.407e+02 4.000e+02 4.882e+02 1.065e+03, threshold=8.000e+02, percent-clipped=1.0 2023-04-28 06:20:51,904 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-04-28 06:21:05,570 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:12,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0445, 4.3641, 2.0902, 4.6311, 2.7946, 4.5788, 2.3034, 3.1750], device='cuda:3'), covar=tensor([0.0117, 0.0153, 0.1474, 0.0040, 0.0753, 0.0273, 0.1260, 0.0550], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0153, 0.0172, 0.0086, 0.0158, 0.0184, 0.0183, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 06:21:55,632 INFO [train.py:904] (3/8) Epoch 5, batch 1400, loss[loss=0.23, simple_loss=0.3129, pruned_loss=0.07356, over 17011.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2863, pruned_loss=0.06856, over 3311217.61 frames. ], batch size: 55, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:22:08,287 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:22:09,339 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 06:22:59,102 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7297, 3.1398, 2.6382, 4.3846, 3.9761, 4.0664, 1.6251, 2.9668], device='cuda:3'), covar=tensor([0.1171, 0.0404, 0.0870, 0.0061, 0.0211, 0.0275, 0.1112, 0.0599], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0140, 0.0164, 0.0083, 0.0171, 0.0169, 0.0158, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 06:23:05,347 INFO [train.py:904] (3/8) Epoch 5, batch 1450, loss[loss=0.2156, simple_loss=0.2915, pruned_loss=0.06986, over 16764.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.285, pruned_loss=0.06857, over 3311246.24 frames. ], batch size: 57, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:23:15,558 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 3.059e+02 3.831e+02 4.678e+02 9.639e+02, threshold=7.661e+02, percent-clipped=1.0 2023-04-28 06:24:14,180 INFO [train.py:904] (3/8) Epoch 5, batch 1500, loss[loss=0.2054, simple_loss=0.2703, pruned_loss=0.0703, over 16369.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2852, pruned_loss=0.06898, over 3314189.32 frames. ], batch size: 146, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:10,932 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 06:25:21,155 INFO [train.py:904] (3/8) Epoch 5, batch 1550, loss[loss=0.2367, simple_loss=0.2943, pruned_loss=0.08954, over 16823.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2864, pruned_loss=0.06992, over 3324049.05 frames. ], batch size: 102, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:32,954 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.189e+02 3.587e+02 4.011e+02 4.553e+02 8.694e+02, threshold=8.021e+02, percent-clipped=2.0 2023-04-28 06:25:45,822 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 06:26:32,628 INFO [train.py:904] (3/8) Epoch 5, batch 1600, loss[loss=0.2137, simple_loss=0.2764, pruned_loss=0.07546, over 16763.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2895, pruned_loss=0.07161, over 3314783.03 frames. ], batch size: 124, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:26:35,201 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5689, 4.4960, 4.4670, 3.9207, 4.4217, 1.7138, 4.2365, 4.3178], device='cuda:3'), covar=tensor([0.0067, 0.0056, 0.0079, 0.0280, 0.0057, 0.1595, 0.0082, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0082, 0.0129, 0.0136, 0.0098, 0.0143, 0.0112, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:27:39,756 INFO [train.py:904] (3/8) Epoch 5, batch 1650, loss[loss=0.2392, simple_loss=0.3196, pruned_loss=0.07941, over 16578.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2912, pruned_loss=0.07248, over 3309447.55 frames. ], batch size: 62, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:49,823 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:49,287 INFO [train.py:904] (3/8) Epoch 5, batch 1700, loss[loss=0.2449, simple_loss=0.3005, pruned_loss=0.09471, over 16486.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2932, pruned_loss=0.07327, over 3316224.99 frames. ], batch size: 146, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:28:54,201 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:29:01,207 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 1750, loss[loss=0.1815, simple_loss=0.2582, pruned_loss=0.05245, over 17007.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.2946, pruned_loss=0.07403, over 3313952.69 frames. ], batch size: 41, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:30:05,725 INFO [optim.py:368] (3/8) 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 06:30:05,949 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:30:31,933 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3275, 4.2874, 4.7009, 4.6962, 4.7457, 4.2631, 4.3575, 4.2372], device='cuda:3'), covar=tensor([0.0271, 0.0360, 0.0330, 0.0420, 0.0357, 0.0298, 0.0725, 0.0405], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0237, 0.0241, 0.0241, 0.0289, 0.0258, 0.0365, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 06:30:35,460 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:31:06,096 INFO [train.py:904] (3/8) Epoch 5, batch 1800, loss[loss=0.2546, simple_loss=0.3172, pruned_loss=0.09599, over 16419.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.296, pruned_loss=0.07389, over 3319451.27 frames. ], batch size: 146, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:01,396 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:06,164 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 06:32:16,840 INFO [train.py:904] (3/8) Epoch 5, batch 1850, loss[loss=0.2163, simple_loss=0.2878, pruned_loss=0.07241, over 16415.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.2964, pruned_loss=0.07406, over 3316486.77 frames. ], batch size: 146, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:26,220 INFO [optim.py:368] (3/8) 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 06:32:52,604 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0818, 4.2857, 3.2415, 2.3287, 3.3701, 2.2809, 4.6233, 4.5488], device='cuda:3'), covar=tensor([0.2015, 0.0694, 0.1274, 0.1607, 0.2242, 0.1467, 0.0345, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0254, 0.0269, 0.0243, 0.0300, 0.0202, 0.0241, 0.0261], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:33:25,574 INFO [train.py:904] (3/8) Epoch 5, batch 1900, loss[loss=0.1801, simple_loss=0.2612, pruned_loss=0.04948, over 16728.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2957, pruned_loss=0.07278, over 3307029.56 frames. ], batch size: 39, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:33:32,587 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:52,018 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6214, 4.9626, 4.6589, 4.7430, 4.4539, 4.3664, 4.4249, 4.9297], device='cuda:3'), covar=tensor([0.0686, 0.0712, 0.0917, 0.0464, 0.0621, 0.0881, 0.0693, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0371, 0.0507, 0.0419, 0.0321, 0.0315, 0.0320, 0.0399, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:34:35,504 INFO [train.py:904] (3/8) Epoch 5, batch 1950, loss[loss=0.1621, simple_loss=0.239, pruned_loss=0.04264, over 16763.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.295, pruned_loss=0.0717, over 3322288.25 frames. ], batch size: 39, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:34:47,175 INFO [optim.py:368] (3/8) 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 06:34:52,221 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:52,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0057, 3.7473, 3.0770, 5.2964, 4.8181, 4.5348, 1.8330, 3.5989], device='cuda:3'), covar=tensor([0.1173, 0.0456, 0.0899, 0.0058, 0.0235, 0.0302, 0.1240, 0.0556], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0141, 0.0168, 0.0086, 0.0176, 0.0172, 0.0160, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 06:34:59,430 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 2000, loss[loss=0.1909, simple_loss=0.2823, pruned_loss=0.0498, over 17260.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2943, pruned_loss=0.07056, over 3323612.06 frames. ], batch size: 52, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:35:51,736 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:36:00,887 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:31,234 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 06:36:47,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7080, 4.7239, 5.1828, 5.1881, 5.1770, 4.7324, 4.7868, 4.5738], device='cuda:3'), covar=tensor([0.0217, 0.0285, 0.0275, 0.0313, 0.0352, 0.0243, 0.0583, 0.0285], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0237, 0.0242, 0.0242, 0.0291, 0.0255, 0.0364, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 06:36:51,835 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3657, 4.2088, 4.3428, 4.3491, 4.2746, 4.8577, 4.5694, 4.2949], device='cuda:3'), covar=tensor([0.1267, 0.1847, 0.1492, 0.1675, 0.2613, 0.1087, 0.1132, 0.2191], device='cuda:3'), in_proj_covar=tensor([0.0281, 0.0409, 0.0385, 0.0347, 0.0466, 0.0414, 0.0320, 0.0461], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:36:55,141 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:57,536 INFO [train.py:904] (3/8) Epoch 5, batch 2050, loss[loss=0.2288, simple_loss=0.2926, pruned_loss=0.08252, over 16800.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.294, pruned_loss=0.0706, over 3319803.05 frames. ], batch size: 83, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:36:59,032 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:37:06,886 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 3.076e+02 3.576e+02 4.219e+02 7.856e+02, threshold=7.152e+02, percent-clipped=1.0 2023-04-28 06:38:05,313 INFO [train.py:904] (3/8) Epoch 5, batch 2100, loss[loss=0.2058, simple_loss=0.2862, pruned_loss=0.06265, over 17187.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2956, pruned_loss=0.07181, over 3327785.07 frames. ], batch size: 46, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:38:18,799 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:19,155 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 06:38:52,937 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 2150, loss[loss=0.27, simple_loss=0.3256, pruned_loss=0.1072, over 15448.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2966, pruned_loss=0.07267, over 3311741.65 frames. ], batch size: 191, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:39:24,110 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.183e+02 3.251e+02 3.882e+02 4.566e+02 8.930e+02, threshold=7.764e+02, percent-clipped=4.0 2023-04-28 06:39:41,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6076, 4.6583, 5.1028, 5.0789, 5.0992, 4.6859, 4.6403, 4.4207], device='cuda:3'), covar=tensor([0.0234, 0.0307, 0.0314, 0.0380, 0.0381, 0.0265, 0.0719, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0235, 0.0239, 0.0243, 0.0289, 0.0254, 0.0363, 0.0216], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 06:40:23,567 INFO [train.py:904] (3/8) Epoch 5, batch 2200, loss[loss=0.2081, simple_loss=0.2821, pruned_loss=0.06699, over 16785.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2962, pruned_loss=0.07249, over 3306600.08 frames. ], batch size: 39, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:40:35,226 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6861, 3.4814, 3.5518, 3.4814, 3.5617, 4.0785, 3.8313, 3.4853], device='cuda:3'), covar=tensor([0.2251, 0.2315, 0.1753, 0.2617, 0.3086, 0.1683, 0.1306, 0.2852], device='cuda:3'), in_proj_covar=tensor([0.0280, 0.0406, 0.0385, 0.0344, 0.0464, 0.0413, 0.0320, 0.0457], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:40:51,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1732, 5.7659, 5.7058, 5.7012, 5.7238, 6.1533, 5.8045, 5.5808], device='cuda:3'), covar=tensor([0.0625, 0.1376, 0.1371, 0.1418, 0.2255, 0.0815, 0.1075, 0.1998], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0410, 0.0389, 0.0348, 0.0468, 0.0415, 0.0323, 0.0460], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:41:06,777 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0056, 4.9247, 4.7903, 4.6735, 4.4012, 4.8598, 4.7743, 4.5051], device='cuda:3'), covar=tensor([0.0390, 0.0250, 0.0177, 0.0178, 0.0820, 0.0250, 0.0331, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0193, 0.0231, 0.0202, 0.0266, 0.0221, 0.0169, 0.0254], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:41:34,695 INFO [train.py:904] (3/8) Epoch 5, batch 2250, loss[loss=0.2435, simple_loss=0.3079, pruned_loss=0.08954, over 17008.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.2972, pruned_loss=0.07356, over 3308102.25 frames. ], batch size: 41, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:41:43,705 INFO [optim.py:368] (3/8) 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 06:41:45,984 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:49,539 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:52,974 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 06:42:14,342 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:32,851 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4747, 5.8128, 5.5565, 5.7498, 5.1582, 4.9231, 5.3401, 5.9317], device='cuda:3'), covar=tensor([0.0847, 0.0647, 0.0871, 0.0374, 0.0630, 0.0567, 0.0518, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0361, 0.0492, 0.0414, 0.0317, 0.0311, 0.0314, 0.0386, 0.0346], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:42:44,587 INFO [train.py:904] (3/8) Epoch 5, batch 2300, loss[loss=0.251, simple_loss=0.3135, pruned_loss=0.09427, over 16512.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2974, pruned_loss=0.07388, over 3305315.39 frames. ], batch size: 68, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:43:00,032 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0793, 2.3889, 2.3124, 4.6153, 1.8935, 3.6843, 2.4239, 2.4320], device='cuda:3'), covar=tensor([0.0448, 0.1848, 0.0962, 0.0256, 0.2831, 0.0707, 0.1735, 0.2328], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0313, 0.0252, 0.0309, 0.0359, 0.0300, 0.0279, 0.0384], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:43:10,956 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 06:43:11,603 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:26,207 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9089, 2.3455, 2.3167, 4.4449, 1.9254, 3.4064, 2.3221, 2.3457], device='cuda:3'), covar=tensor([0.0479, 0.1932, 0.1012, 0.0272, 0.2924, 0.0818, 0.1820, 0.2502], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0312, 0.0252, 0.0309, 0.0358, 0.0300, 0.0279, 0.0385], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:43:38,275 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:53,244 INFO [train.py:904] (3/8) Epoch 5, batch 2350, loss[loss=0.2105, simple_loss=0.2989, pruned_loss=0.06107, over 17148.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2976, pruned_loss=0.07407, over 3299801.15 frames. ], batch size: 48, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:44:03,301 INFO [optim.py:368] (3/8) 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 06:44:56,612 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-28 06:45:02,226 INFO [train.py:904] (3/8) Epoch 5, batch 2400, loss[loss=0.2487, simple_loss=0.3255, pruned_loss=0.08598, over 16568.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2969, pruned_loss=0.07306, over 3310820.05 frames. ], batch size: 62, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:45:08,770 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:52,152 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 2450, loss[loss=0.1825, simple_loss=0.2621, pruned_loss=0.05143, over 16763.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.298, pruned_loss=0.07256, over 3312511.72 frames. ], batch size: 39, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:46:26,010 INFO [optim.py:368] (3/8) 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 06:46:57,984 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:47:23,968 INFO [train.py:904] (3/8) Epoch 5, batch 2500, loss[loss=0.2412, simple_loss=0.3158, pruned_loss=0.08329, over 16893.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2977, pruned_loss=0.07205, over 3312962.83 frames. ], batch size: 116, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:47:36,296 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 06:48:33,486 INFO [train.py:904] (3/8) Epoch 5, batch 2550, loss[loss=0.2019, simple_loss=0.2754, pruned_loss=0.06424, over 16826.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2978, pruned_loss=0.07203, over 3317135.35 frames. ], batch size: 102, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:45,558 INFO [optim.py:368] (3/8) 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 06:48:48,813 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:04,612 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 06:49:15,325 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1374, 5.6315, 5.7127, 5.5181, 5.5523, 6.0829, 5.7687, 5.4782], device='cuda:3'), covar=tensor([0.0694, 0.1659, 0.1270, 0.1573, 0.2459, 0.0898, 0.1034, 0.2098], device='cuda:3'), in_proj_covar=tensor([0.0281, 0.0398, 0.0383, 0.0341, 0.0465, 0.0413, 0.0321, 0.0457], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 06:49:43,647 INFO [train.py:904] (3/8) Epoch 5, batch 2600, loss[loss=0.2286, simple_loss=0.2956, pruned_loss=0.0808, over 16841.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.2967, pruned_loss=0.07123, over 3326147.64 frames. ], batch size: 102, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:49:55,553 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:01,747 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:53,870 INFO [train.py:904] (3/8) Epoch 5, batch 2650, loss[loss=0.2205, simple_loss=0.2905, pruned_loss=0.07526, over 16911.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2967, pruned_loss=0.07109, over 3331627.94 frames. ], batch size: 96, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:51:05,580 INFO [optim.py:368] (3/8) 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 06:51:15,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2305, 5.5357, 5.2430, 5.4546, 4.9359, 4.7776, 5.0956, 5.6451], device='cuda:3'), covar=tensor([0.0729, 0.0715, 0.0935, 0.0431, 0.0626, 0.0612, 0.0575, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0377, 0.0511, 0.0431, 0.0326, 0.0321, 0.0328, 0.0409, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:51:25,867 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8734, 4.5807, 4.8500, 5.1860, 5.3161, 4.6112, 5.2389, 5.2759], device='cuda:3'), covar=tensor([0.0789, 0.0764, 0.1355, 0.0452, 0.0399, 0.0598, 0.0432, 0.0377], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0493, 0.0638, 0.0511, 0.0378, 0.0377, 0.0398, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:51:58,874 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3774, 3.5330, 1.8417, 3.5073, 2.4320, 3.6082, 1.8157, 2.5961], device='cuda:3'), covar=tensor([0.0129, 0.0266, 0.1301, 0.0119, 0.0699, 0.0377, 0.1248, 0.0526], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0155, 0.0175, 0.0088, 0.0158, 0.0187, 0.0185, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 06:52:02,500 INFO [train.py:904] (3/8) Epoch 5, batch 2700, loss[loss=0.2017, simple_loss=0.2814, pruned_loss=0.06095, over 16845.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.297, pruned_loss=0.07108, over 3328175.96 frames. ], batch size: 42, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:52:09,065 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:12,572 INFO [train.py:904] (3/8) Epoch 5, batch 2750, loss[loss=0.2205, simple_loss=0.2927, pruned_loss=0.07411, over 16727.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.296, pruned_loss=0.06982, over 3335622.85 frames. ], batch size: 89, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:53:15,875 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:25,427 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.783e+02 3.325e+02 4.453e+02 8.515e+02, threshold=6.649e+02, percent-clipped=2.0 2023-04-28 06:53:35,023 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9560, 4.3197, 3.3724, 2.3761, 3.2487, 2.4430, 4.4893, 4.3266], device='cuda:3'), covar=tensor([0.1959, 0.0496, 0.1106, 0.1552, 0.2119, 0.1391, 0.0329, 0.0530], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0249, 0.0264, 0.0241, 0.0302, 0.0200, 0.0237, 0.0259], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:54:07,299 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4214, 5.8571, 5.5554, 5.6088, 5.0970, 4.8897, 5.3203, 5.9261], device='cuda:3'), covar=tensor([0.0740, 0.0645, 0.0883, 0.0441, 0.0623, 0.0594, 0.0570, 0.0716], device='cuda:3'), in_proj_covar=tensor([0.0377, 0.0510, 0.0429, 0.0327, 0.0318, 0.0326, 0.0406, 0.0358], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 06:54:22,964 INFO [train.py:904] (3/8) Epoch 5, batch 2800, loss[loss=0.1737, simple_loss=0.2579, pruned_loss=0.04474, over 16828.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2965, pruned_loss=0.06998, over 3342589.84 frames. ], batch size: 39, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:33,348 INFO [train.py:904] (3/8) Epoch 5, batch 2850, loss[loss=0.2127, simple_loss=0.2789, pruned_loss=0.07324, over 16818.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2958, pruned_loss=0.07031, over 3328122.16 frames. ], batch size: 116, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:45,525 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 5, batch 2900, loss[loss=0.276, simple_loss=0.3274, pruned_loss=0.1123, over 15482.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2957, pruned_loss=0.0711, over 3327865.93 frames. ], batch size: 190, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:57:00,349 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:01,407 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:28,489 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:49,461 INFO [train.py:904] (3/8) Epoch 5, batch 2950, loss[loss=0.2316, simple_loss=0.2916, pruned_loss=0.08578, over 16423.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.2956, pruned_loss=0.07196, over 3326338.12 frames. ], batch size: 75, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:02,029 INFO [optim.py:368] (3/8) 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 06:58:05,993 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:24,088 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:34,162 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:59,870 INFO [train.py:904] (3/8) Epoch 5, batch 3000, loss[loss=0.1905, simple_loss=0.2848, pruned_loss=0.04815, over 17109.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2965, pruned_loss=0.07289, over 3324711.50 frames. ], batch size: 48, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:59,871 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 06:59:08,834 INFO [train.py:938] (3/8) Epoch 5, validation: loss=0.1564, simple_loss=0.2632, pruned_loss=0.02477, over 944034.00 frames. 2023-04-28 06:59:08,834 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 06:59:56,495 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1940, 4.4480, 2.2016, 4.8213, 2.9104, 4.7495, 2.4893, 3.4761], device='cuda:3'), covar=tensor([0.0096, 0.0209, 0.1372, 0.0039, 0.0661, 0.0266, 0.1162, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0158, 0.0175, 0.0089, 0.0159, 0.0190, 0.0185, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 07:00:18,524 INFO [train.py:904] (3/8) Epoch 5, batch 3050, loss[loss=0.2179, simple_loss=0.3025, pruned_loss=0.06661, over 16635.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.2957, pruned_loss=0.07221, over 3324458.31 frames. ], batch size: 62, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:00:31,425 INFO [optim.py:368] (3/8) 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 07:00:40,293 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 07:01:25,947 INFO [train.py:904] (3/8) Epoch 5, batch 3100, loss[loss=0.2322, simple_loss=0.2963, pruned_loss=0.0841, over 16906.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2941, pruned_loss=0.0707, over 3330176.50 frames. ], batch size: 116, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:01:48,567 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 07:01:50,580 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5046, 3.5574, 2.7286, 2.2185, 2.5463, 2.1631, 3.4039, 3.5620], device='cuda:3'), covar=tensor([0.2026, 0.0521, 0.1069, 0.1474, 0.2039, 0.1300, 0.0384, 0.0619], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0251, 0.0264, 0.0241, 0.0305, 0.0199, 0.0236, 0.0263], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:02:32,727 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0414, 4.9380, 4.8393, 4.6388, 4.3974, 4.8484, 4.9521, 4.5651], device='cuda:3'), covar=tensor([0.0406, 0.0273, 0.0215, 0.0191, 0.0883, 0.0309, 0.0227, 0.0508], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0232, 0.0202, 0.0267, 0.0228, 0.0171, 0.0255], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:02:33,554 INFO [train.py:904] (3/8) Epoch 5, batch 3150, loss[loss=0.21, simple_loss=0.2883, pruned_loss=0.0658, over 16749.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2935, pruned_loss=0.07086, over 3329832.00 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:46,878 INFO [optim.py:368] (3/8) 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 07:03:22,267 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:03:42,786 INFO [train.py:904] (3/8) Epoch 5, batch 3200, loss[loss=0.2185, simple_loss=0.2934, pruned_loss=0.07181, over 16979.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2931, pruned_loss=0.07082, over 3319211.91 frames. ], batch size: 41, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:04:48,183 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:04:52,114 INFO [train.py:904] (3/8) Epoch 5, batch 3250, loss[loss=0.2665, simple_loss=0.3232, pruned_loss=0.1049, over 16697.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2939, pruned_loss=0.0712, over 3309911.51 frames. ], batch size: 134, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:04:52,666 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2341, 3.8895, 3.1904, 1.8867, 2.7420, 2.3425, 3.7105, 3.6197], device='cuda:3'), covar=tensor([0.0217, 0.0436, 0.0641, 0.1517, 0.0705, 0.0933, 0.0463, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0137, 0.0155, 0.0142, 0.0134, 0.0127, 0.0143, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 07:05:06,669 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:05:28,564 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9941, 3.5062, 2.5729, 4.8518, 4.2713, 4.1233, 1.8317, 3.1873], device='cuda:3'), covar=tensor([0.1121, 0.0429, 0.1014, 0.0058, 0.0276, 0.0323, 0.1188, 0.0643], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0140, 0.0163, 0.0085, 0.0182, 0.0171, 0.0156, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 07:06:03,319 INFO [train.py:904] (3/8) Epoch 5, batch 3300, loss[loss=0.2038, simple_loss=0.2986, pruned_loss=0.05451, over 17146.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.2954, pruned_loss=0.0714, over 3311988.91 frames. ], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:06:11,215 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4704, 3.5700, 3.1275, 3.2107, 3.0561, 3.3664, 3.2590, 3.1153], device='cuda:3'), covar=tensor([0.0444, 0.0264, 0.0209, 0.0202, 0.0584, 0.0262, 0.0842, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0199, 0.0232, 0.0204, 0.0271, 0.0230, 0.0171, 0.0258], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:06:14,780 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:12,748 INFO [train.py:904] (3/8) Epoch 5, batch 3350, loss[loss=0.2051, simple_loss=0.282, pruned_loss=0.06405, over 16796.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2958, pruned_loss=0.07156, over 3313571.79 frames. ], batch size: 102, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:07:24,309 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:07:39,202 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:08:22,242 INFO [train.py:904] (3/8) Epoch 5, batch 3400, loss[loss=0.1939, simple_loss=0.2769, pruned_loss=0.05544, over 17106.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2959, pruned_loss=0.07158, over 3313880.78 frames. ], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:08:47,637 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:09:31,565 INFO [train.py:904] (3/8) Epoch 5, batch 3450, loss[loss=0.2075, simple_loss=0.2759, pruned_loss=0.06953, over 16514.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2936, pruned_loss=0.07063, over 3317331.68 frames. ], batch size: 75, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:09:44,808 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 3.034e+02 3.683e+02 4.468e+02 1.074e+03, threshold=7.367e+02, percent-clipped=2.0 2023-04-28 07:10:33,651 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8983, 4.5902, 4.8834, 5.1576, 5.3040, 4.6460, 5.2881, 5.2531], device='cuda:3'), covar=tensor([0.0857, 0.0807, 0.1285, 0.0511, 0.0324, 0.0585, 0.0334, 0.0355], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0509, 0.0655, 0.0525, 0.0388, 0.0393, 0.0395, 0.0430], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:10:39,130 INFO [train.py:904] (3/8) Epoch 5, batch 3500, loss[loss=0.2543, simple_loss=0.3113, pruned_loss=0.09865, over 16826.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2928, pruned_loss=0.07015, over 3312388.04 frames. ], batch size: 96, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:11:13,799 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1855, 4.0874, 4.0116, 3.5875, 4.0698, 1.7342, 3.9151, 3.8850], device='cuda:3'), covar=tensor([0.0075, 0.0066, 0.0096, 0.0260, 0.0065, 0.1605, 0.0093, 0.0140], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0087, 0.0130, 0.0138, 0.0101, 0.0139, 0.0116, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:11:37,812 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:11:49,615 INFO [train.py:904] (3/8) Epoch 5, batch 3550, loss[loss=0.2039, simple_loss=0.2897, pruned_loss=0.05902, over 17087.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2912, pruned_loss=0.06895, over 3312352.13 frames. ], batch size: 53, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:12:03,018 INFO [optim.py:368] (3/8) 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 07:12:16,286 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-28 07:12:17,262 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:12:46,626 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4040, 4.1780, 3.6057, 1.9796, 3.0764, 2.3534, 3.6756, 3.8597], device='cuda:3'), covar=tensor([0.0284, 0.0467, 0.0552, 0.1548, 0.0696, 0.0958, 0.0659, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0133, 0.0153, 0.0138, 0.0131, 0.0124, 0.0140, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 07:12:59,205 INFO [train.py:904] (3/8) Epoch 5, batch 3600, loss[loss=0.1848, simple_loss=0.2737, pruned_loss=0.04789, over 17116.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2901, pruned_loss=0.06869, over 3310554.33 frames. ], batch size: 47, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:13:08,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5389, 4.1070, 4.3372, 2.9279, 3.8996, 4.4483, 4.0613, 2.3915], device='cuda:3'), covar=tensor([0.0305, 0.0026, 0.0023, 0.0218, 0.0035, 0.0029, 0.0030, 0.0285], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0059, 0.0061, 0.0113, 0.0062, 0.0069, 0.0066, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:13:23,275 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:13:37,272 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4435, 4.4562, 5.0157, 4.9988, 5.0041, 4.5358, 4.5672, 4.3670], device='cuda:3'), covar=tensor([0.0314, 0.0613, 0.0361, 0.0401, 0.0412, 0.0356, 0.0868, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0256, 0.0250, 0.0251, 0.0257, 0.0303, 0.0267, 0.0382, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 07:14:09,667 INFO [train.py:904] (3/8) Epoch 5, batch 3650, loss[loss=0.2163, simple_loss=0.2888, pruned_loss=0.07185, over 15659.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2892, pruned_loss=0.06975, over 3310509.52 frames. ], batch size: 191, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:14:22,914 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1060, 1.2835, 1.9089, 2.1254, 2.2553, 2.1648, 1.4142, 2.2002], device='cuda:3'), covar=tensor([0.0087, 0.0209, 0.0114, 0.0116, 0.0082, 0.0086, 0.0192, 0.0051], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0143, 0.0129, 0.0132, 0.0128, 0.0094, 0.0140, 0.0082], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 07:14:25,432 INFO [optim.py:368] (3/8) 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 07:14:30,509 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:15:22,467 INFO [train.py:904] (3/8) Epoch 5, batch 3700, loss[loss=0.2209, simple_loss=0.2901, pruned_loss=0.07588, over 16490.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2887, pruned_loss=0.07192, over 3292990.62 frames. ], batch size: 146, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:15:44,546 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:15:47,214 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1371, 3.3034, 3.2605, 1.5865, 3.4708, 3.5215, 2.9279, 2.6858], device='cuda:3'), covar=tensor([0.0734, 0.0121, 0.0160, 0.1153, 0.0074, 0.0068, 0.0333, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0087, 0.0085, 0.0143, 0.0074, 0.0080, 0.0118, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 07:16:30,105 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-04-28 07:16:36,558 INFO [train.py:904] (3/8) Epoch 5, batch 3750, loss[loss=0.2189, simple_loss=0.2753, pruned_loss=0.08124, over 16451.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.2899, pruned_loss=0.07375, over 3274552.30 frames. ], batch size: 75, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:16:52,716 INFO [optim.py:368] (3/8) 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 07:17:08,049 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 07:17:49,965 INFO [train.py:904] (3/8) Epoch 5, batch 3800, loss[loss=0.2361, simple_loss=0.2942, pruned_loss=0.08898, over 16848.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.2906, pruned_loss=0.07507, over 3282496.51 frames. ], batch size: 116, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:18:34,882 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8055, 5.2779, 4.9418, 4.9490, 4.5950, 4.5389, 4.7055, 5.3010], device='cuda:3'), covar=tensor([0.0751, 0.0552, 0.0979, 0.0443, 0.0612, 0.0775, 0.0616, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0364, 0.0486, 0.0413, 0.0320, 0.0310, 0.0320, 0.0395, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:18:38,449 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-28 07:18:50,506 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 3850, loss[loss=0.2086, simple_loss=0.2762, pruned_loss=0.07046, over 16828.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.2909, pruned_loss=0.07563, over 3277208.81 frames. ], batch size: 102, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:19:16,581 INFO [optim.py:368] (3/8) 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 07:19:19,512 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4400, 4.3419, 4.3408, 3.8401, 4.3593, 1.8197, 4.1851, 4.1986], device='cuda:3'), covar=tensor([0.0064, 0.0059, 0.0076, 0.0281, 0.0057, 0.1531, 0.0086, 0.0121], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0085, 0.0126, 0.0135, 0.0098, 0.0138, 0.0112, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:19:57,670 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:20:11,834 INFO [train.py:904] (3/8) Epoch 5, batch 3900, loss[loss=0.2171, simple_loss=0.287, pruned_loss=0.07363, over 16207.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2897, pruned_loss=0.07573, over 3276442.92 frames. ], batch size: 35, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:20:37,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3201, 3.9617, 3.3198, 1.8690, 2.9026, 2.4458, 3.6778, 3.6564], device='cuda:3'), covar=tensor([0.0254, 0.0463, 0.0657, 0.1765, 0.0739, 0.0983, 0.0471, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0132, 0.0152, 0.0139, 0.0130, 0.0123, 0.0138, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 07:20:49,863 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:22,466 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 3950, loss[loss=0.2083, simple_loss=0.2869, pruned_loss=0.06484, over 17224.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.2892, pruned_loss=0.07572, over 3278664.41 frames. ], batch size: 45, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:21:37,715 INFO [optim.py:368] (3/8) 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 07:21:44,209 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:08,308 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5968, 4.3184, 3.6569, 1.9632, 3.0157, 2.7196, 3.7814, 4.0827], device='cuda:3'), covar=tensor([0.0161, 0.0309, 0.0473, 0.1459, 0.0667, 0.0736, 0.0493, 0.0494], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0133, 0.0153, 0.0140, 0.0132, 0.0124, 0.0140, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 07:22:16,000 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:34,881 INFO [train.py:904] (3/8) Epoch 5, batch 4000, loss[loss=0.212, simple_loss=0.2789, pruned_loss=0.07252, over 16856.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.2885, pruned_loss=0.075, over 3291264.63 frames. ], batch size: 90, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:22:49,558 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:54,862 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:23:02,205 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5298, 3.7969, 2.8547, 2.3234, 2.6874, 2.2155, 3.7177, 3.8317], device='cuda:3'), covar=tensor([0.2361, 0.0563, 0.1264, 0.1630, 0.2027, 0.1430, 0.0434, 0.0535], device='cuda:3'), in_proj_covar=tensor([0.0279, 0.0253, 0.0268, 0.0248, 0.0312, 0.0207, 0.0239, 0.0267], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:23:06,901 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2277, 4.1114, 4.1541, 3.5920, 4.1191, 1.6361, 3.8913, 3.9560], device='cuda:3'), covar=tensor([0.0075, 0.0076, 0.0090, 0.0297, 0.0067, 0.1754, 0.0105, 0.0143], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0086, 0.0126, 0.0135, 0.0097, 0.0139, 0.0113, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:23:45,855 INFO [train.py:904] (3/8) Epoch 5, batch 4050, loss[loss=0.2082, simple_loss=0.2857, pruned_loss=0.06531, over 16514.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2872, pruned_loss=0.07282, over 3286133.79 frames. ], batch size: 75, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:24:02,341 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-28 07:24:02,948 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.760e+02 3.326e+02 4.025e+02 7.816e+02, threshold=6.651e+02, percent-clipped=1.0 2023-04-28 07:24:04,434 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:25:01,112 INFO [train.py:904] (3/8) Epoch 5, batch 4100, loss[loss=0.2178, simple_loss=0.2946, pruned_loss=0.07047, over 16644.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.287, pruned_loss=0.07092, over 3285913.74 frames. ], batch size: 57, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:04,075 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 07:26:13,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9366, 4.8826, 4.6764, 4.0311, 4.7866, 1.8594, 4.5093, 4.7117], device='cuda:3'), covar=tensor([0.0041, 0.0041, 0.0075, 0.0287, 0.0039, 0.1515, 0.0066, 0.0086], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0084, 0.0123, 0.0133, 0.0095, 0.0138, 0.0110, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:26:16,728 INFO [train.py:904] (3/8) Epoch 5, batch 4150, loss[loss=0.2255, simple_loss=0.3131, pruned_loss=0.06895, over 16834.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.2952, pruned_loss=0.07545, over 3219380.47 frames. ], batch size: 102, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:34,909 INFO [optim.py:368] (3/8) 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 07:27:35,372 INFO [train.py:904] (3/8) Epoch 5, batch 4200, loss[loss=0.3043, simple_loss=0.3629, pruned_loss=0.1229, over 11232.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3026, pruned_loss=0.07786, over 3189963.30 frames. ], batch size: 246, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:28:51,233 INFO [train.py:904] (3/8) Epoch 5, batch 4250, loss[loss=0.2175, simple_loss=0.3044, pruned_loss=0.06534, over 16687.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3054, pruned_loss=0.07719, over 3191703.14 frames. ], batch size: 62, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:29:07,185 INFO [optim.py:368] (3/8) 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 07:29:38,863 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 4300, loss[loss=0.2783, simple_loss=0.3375, pruned_loss=0.1095, over 11575.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3072, pruned_loss=0.07635, over 3181041.19 frames. ], batch size: 247, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:30:13,087 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 4350, loss[loss=0.2375, simple_loss=0.3149, pruned_loss=0.08006, over 16924.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3114, pruned_loss=0.0781, over 3180335.48 frames. ], batch size: 109, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:31:36,571 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 5, batch 4400, loss[loss=0.2565, simple_loss=0.3306, pruned_loss=0.09125, over 17128.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3133, pruned_loss=0.07911, over 3182817.49 frames. ], batch size: 49, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:27,577 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4720, 4.4549, 4.2148, 3.4959, 4.3095, 1.5786, 4.1399, 4.0661], device='cuda:3'), covar=tensor([0.0049, 0.0039, 0.0080, 0.0322, 0.0049, 0.1800, 0.0076, 0.0134], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0081, 0.0119, 0.0130, 0.0091, 0.0136, 0.0106, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:33:49,611 INFO [train.py:904] (3/8) Epoch 5, batch 4450, loss[loss=0.2407, simple_loss=0.3231, pruned_loss=0.07918, over 16685.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3161, pruned_loss=0.07964, over 3192262.49 frames. ], batch size: 57, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:56,464 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:34:13,242 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 07:34:33,629 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 4500, loss[loss=0.2162, simple_loss=0.2928, pruned_loss=0.06979, over 16513.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3164, pruned_loss=0.08013, over 3184826.93 frames. ], batch size: 35, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:35:26,559 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:36:02,219 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:36:07,546 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2378, 4.1754, 3.9861, 3.3611, 4.0657, 1.6121, 3.9289, 3.9024], device='cuda:3'), covar=tensor([0.0048, 0.0042, 0.0078, 0.0284, 0.0050, 0.1710, 0.0062, 0.0102], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0079, 0.0117, 0.0127, 0.0090, 0.0135, 0.0104, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:36:13,664 INFO [train.py:904] (3/8) Epoch 5, batch 4550, loss[loss=0.2716, simple_loss=0.3534, pruned_loss=0.09487, over 16544.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3172, pruned_loss=0.08046, over 3195544.43 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:36:30,358 INFO [optim.py:368] (3/8) 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 07:37:00,845 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 4600, loss[loss=0.2117, simple_loss=0.2984, pruned_loss=0.06248, over 16680.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3176, pruned_loss=0.0798, over 3204081.91 frames. ], batch size: 57, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:37:35,976 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:12,281 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:22,068 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2023-04-28 07:38:40,728 INFO [train.py:904] (3/8) Epoch 5, batch 4650, loss[loss=0.2329, simple_loss=0.3159, pruned_loss=0.07493, over 16674.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3164, pruned_loss=0.07924, over 3204340.76 frames. ], batch size: 134, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:38:45,218 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:57,178 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.863e+02 2.661e+02 3.124e+02 3.951e+02 6.864e+02, threshold=6.249e+02, percent-clipped=3.0 2023-04-28 07:39:25,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1363, 2.4058, 2.6059, 4.8411, 2.1141, 3.3158, 2.7265, 2.5422], device='cuda:3'), covar=tensor([0.0451, 0.1886, 0.0932, 0.0219, 0.2754, 0.0847, 0.1518, 0.2207], device='cuda:3'), in_proj_covar=tensor([0.0316, 0.0319, 0.0257, 0.0306, 0.0368, 0.0304, 0.0283, 0.0388], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:39:55,266 INFO [train.py:904] (3/8) Epoch 5, batch 4700, loss[loss=0.212, simple_loss=0.2819, pruned_loss=0.07102, over 11431.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3127, pruned_loss=0.07745, over 3210488.06 frames. ], batch size: 248, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:40:33,602 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8736, 4.6162, 4.8193, 5.1107, 5.2104, 4.5728, 5.2110, 5.1929], device='cuda:3'), covar=tensor([0.0814, 0.0850, 0.1213, 0.0411, 0.0323, 0.0523, 0.0346, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0362, 0.0445, 0.0564, 0.0453, 0.0342, 0.0341, 0.0358, 0.0373], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:41:03,084 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:41:05,144 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 07:41:06,793 INFO [train.py:904] (3/8) Epoch 5, batch 4750, loss[loss=0.205, simple_loss=0.288, pruned_loss=0.06105, over 16697.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3085, pruned_loss=0.07549, over 3210907.49 frames. ], batch size: 62, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:41:22,808 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.710e+02 3.330e+02 4.145e+02 7.902e+02, threshold=6.661e+02, percent-clipped=5.0 2023-04-28 07:42:20,534 INFO [train.py:904] (3/8) Epoch 5, batch 4800, loss[loss=0.2165, simple_loss=0.3072, pruned_loss=0.06289, over 17227.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3049, pruned_loss=0.07321, over 3212740.97 frames. ], batch size: 52, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:42:33,268 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:42:38,680 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:43:14,510 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:43:33,748 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1636, 3.3891, 1.5343, 3.5303, 2.3325, 3.4611, 1.8845, 2.5678], device='cuda:3'), covar=tensor([0.0132, 0.0226, 0.1705, 0.0048, 0.0779, 0.0392, 0.1388, 0.0634], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0147, 0.0172, 0.0079, 0.0157, 0.0178, 0.0183, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 07:43:34,380 INFO [train.py:904] (3/8) Epoch 5, batch 4850, loss[loss=0.214, simple_loss=0.3053, pruned_loss=0.06136, over 16317.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3058, pruned_loss=0.07256, over 3216316.51 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:43:38,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7916, 1.6904, 1.4675, 1.5226, 1.9027, 1.6622, 1.8483, 1.9634], device='cuda:3'), covar=tensor([0.0041, 0.0136, 0.0188, 0.0165, 0.0093, 0.0151, 0.0066, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0078, 0.0149, 0.0151, 0.0147, 0.0145, 0.0154, 0.0121, 0.0133], device='cuda:3'), out_proj_covar=tensor([9.8886e-05, 1.8351e-04, 1.8275e-04, 1.7724e-04, 1.8135e-04, 1.9100e-04, 1.4537e-04, 1.6534e-04], device='cuda:3') 2023-04-28 07:43:50,677 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.510e+02 3.182e+02 3.913e+02 7.711e+02, threshold=6.364e+02, percent-clipped=1.0 2023-04-28 07:44:47,402 INFO [train.py:904] (3/8) Epoch 5, batch 4900, loss[loss=0.2067, simple_loss=0.2987, pruned_loss=0.05735, over 16650.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.305, pruned_loss=0.07112, over 3215021.01 frames. ], batch size: 134, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:01,018 INFO [train.py:904] (3/8) Epoch 5, batch 4950, loss[loss=0.2334, simple_loss=0.3139, pruned_loss=0.07648, over 16932.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3049, pruned_loss=0.07094, over 3214208.27 frames. ], batch size: 109, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:15,481 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.470e+02 3.018e+02 3.573e+02 4.428e+02 9.346e+02, threshold=7.147e+02, percent-clipped=9.0 2023-04-28 07:47:03,531 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9980, 4.0651, 4.4036, 4.3556, 4.3677, 4.0129, 4.0307, 3.8479], device='cuda:3'), covar=tensor([0.0215, 0.0302, 0.0299, 0.0400, 0.0396, 0.0229, 0.0708, 0.0393], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0219, 0.0223, 0.0225, 0.0272, 0.0238, 0.0338, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 07:47:06,380 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-28 07:47:13,814 INFO [train.py:904] (3/8) Epoch 5, batch 5000, loss[loss=0.2777, simple_loss=0.3442, pruned_loss=0.1056, over 12283.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3066, pruned_loss=0.07091, over 3230921.28 frames. ], batch size: 248, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:19,935 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 07:48:24,540 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8189, 3.1309, 2.6011, 4.8554, 4.1671, 4.1507, 1.7461, 3.1810], device='cuda:3'), covar=tensor([0.1249, 0.0543, 0.1027, 0.0057, 0.0169, 0.0255, 0.1304, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0141, 0.0166, 0.0084, 0.0171, 0.0170, 0.0161, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 07:48:27,066 INFO [train.py:904] (3/8) Epoch 5, batch 5050, loss[loss=0.2359, simple_loss=0.3203, pruned_loss=0.07569, over 16745.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3066, pruned_loss=0.07074, over 3231612.23 frames. ], batch size: 124, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:30,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4172, 4.4560, 4.2543, 4.1327, 3.7614, 4.3429, 4.2230, 4.0407], device='cuda:3'), covar=tensor([0.0412, 0.0236, 0.0196, 0.0187, 0.0880, 0.0249, 0.0258, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0182, 0.0214, 0.0186, 0.0244, 0.0212, 0.0153, 0.0229], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 07:48:33,415 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 07:48:43,127 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.781e+02 3.302e+02 3.985e+02 6.386e+02, threshold=6.604e+02, percent-clipped=0.0 2023-04-28 07:49:39,972 INFO [train.py:904] (3/8) Epoch 5, batch 5100, loss[loss=0.1969, simple_loss=0.2881, pruned_loss=0.05287, over 16771.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3038, pruned_loss=0.06929, over 3239290.91 frames. ], batch size: 83, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:49:42,404 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:49:44,452 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:49:52,048 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 07:49:55,998 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:50:33,731 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:50:53,044 INFO [train.py:904] (3/8) Epoch 5, batch 5150, loss[loss=0.2511, simple_loss=0.3264, pruned_loss=0.08794, over 11818.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3042, pruned_loss=0.06878, over 3226720.05 frames. ], batch size: 247, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:51:04,006 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9614, 5.2987, 4.9364, 4.9638, 4.6263, 4.5242, 4.7049, 5.2982], device='cuda:3'), covar=tensor([0.0544, 0.0566, 0.0807, 0.0460, 0.0651, 0.0647, 0.0603, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0455, 0.0399, 0.0304, 0.0294, 0.0304, 0.0374, 0.0330], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:51:06,958 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:51:10,548 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:21,232 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 07:51:26,054 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:42,660 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 5200, loss[loss=0.1952, simple_loss=0.2742, pruned_loss=0.05808, over 16759.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3036, pruned_loss=0.06933, over 3217347.96 frames. ], batch size: 57, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:52:54,186 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 5250, loss[loss=0.2225, simple_loss=0.2935, pruned_loss=0.07579, over 16322.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3015, pruned_loss=0.06923, over 3214377.40 frames. ], batch size: 35, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:53:27,345 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.734e+02 3.218e+02 3.993e+02 9.159e+02, threshold=6.436e+02, percent-clipped=4.0 2023-04-28 07:54:26,348 INFO [train.py:904] (3/8) Epoch 5, batch 5300, loss[loss=0.2508, simple_loss=0.3162, pruned_loss=0.09265, over 12299.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2989, pruned_loss=0.06825, over 3195795.77 frames. ], batch size: 249, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:54:53,143 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:55:05,155 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8213, 4.7365, 4.5235, 3.9062, 4.5687, 1.7886, 4.4238, 4.5713], device='cuda:3'), covar=tensor([0.0049, 0.0051, 0.0083, 0.0348, 0.0058, 0.1630, 0.0089, 0.0128], device='cuda:3'), in_proj_covar=tensor([0.0089, 0.0078, 0.0117, 0.0128, 0.0090, 0.0137, 0.0105, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:55:36,417 INFO [train.py:904] (3/8) Epoch 5, batch 5350, loss[loss=0.2078, simple_loss=0.2971, pruned_loss=0.05931, over 16815.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2968, pruned_loss=0.06718, over 3210321.49 frames. ], batch size: 83, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:55:53,055 INFO [optim.py:368] (3/8) 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 07:56:14,701 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3867, 4.1618, 4.3833, 4.6339, 4.7366, 4.2770, 4.7191, 4.7386], device='cuda:3'), covar=tensor([0.0759, 0.0667, 0.1034, 0.0431, 0.0322, 0.0566, 0.0319, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0461, 0.0585, 0.0470, 0.0352, 0.0353, 0.0372, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 07:56:52,489 INFO [train.py:904] (3/8) Epoch 5, batch 5400, loss[loss=0.2049, simple_loss=0.2888, pruned_loss=0.0605, over 17202.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2993, pruned_loss=0.0681, over 3195532.91 frames. ], batch size: 46, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:56:57,453 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:58:09,282 INFO [train.py:904] (3/8) Epoch 5, batch 5450, loss[loss=0.2421, simple_loss=0.3248, pruned_loss=0.07972, over 16855.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3038, pruned_loss=0.07075, over 3205195.85 frames. ], batch size: 96, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:58:11,188 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:58:19,238 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:58:24,986 INFO [optim.py:368] (3/8) 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 07:59:22,160 INFO [train.py:904] (3/8) Epoch 5, batch 5500, loss[loss=0.2354, simple_loss=0.316, pruned_loss=0.07742, over 17127.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3127, pruned_loss=0.07786, over 3156507.02 frames. ], batch size: 47, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:07,593 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:00:39,140 INFO [train.py:904] (3/8) Epoch 5, batch 5550, loss[loss=0.3678, simple_loss=0.3965, pruned_loss=0.1695, over 11190.00 frames. ], tot_loss[loss=0.2457, simple_loss=0.322, pruned_loss=0.08475, over 3143249.66 frames. ], batch size: 247, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:56,874 INFO [optim.py:368] (3/8) 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 08:01:54,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9088, 4.2054, 3.8797, 3.8945, 3.1168, 4.1503, 3.8905, 3.6967], device='cuda:3'), covar=tensor([0.0786, 0.0322, 0.0344, 0.0264, 0.1538, 0.0357, 0.0707, 0.0489], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0183, 0.0207, 0.0182, 0.0239, 0.0208, 0.0150, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:01:58,727 INFO [train.py:904] (3/8) Epoch 5, batch 5600, loss[loss=0.3581, simple_loss=0.3969, pruned_loss=0.1597, over 11332.00 frames. ], tot_loss[loss=0.2558, simple_loss=0.3289, pruned_loss=0.09134, over 3101214.78 frames. ], batch size: 246, lr: 1.33e-02, grad_scale: 16.0 2023-04-28 08:02:23,966 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 5650, loss[loss=0.2821, simple_loss=0.3409, pruned_loss=0.1116, over 15321.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3344, pruned_loss=0.09571, over 3099298.30 frames. ], batch size: 190, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:03:42,415 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.653e+02 4.676e+02 6.240e+02 7.908e+02 1.367e+03, threshold=1.248e+03, percent-clipped=1.0 2023-04-28 08:04:04,191 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:04:38,190 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7198, 6.0231, 5.7161, 5.9160, 5.3337, 5.1014, 5.6455, 6.1378], device='cuda:3'), covar=tensor([0.0638, 0.0634, 0.0916, 0.0382, 0.0642, 0.0545, 0.0507, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0468, 0.0407, 0.0306, 0.0291, 0.0310, 0.0379, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:04:43,823 INFO [train.py:904] (3/8) Epoch 5, batch 5700, loss[loss=0.3264, simple_loss=0.3679, pruned_loss=0.1424, over 11203.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3372, pruned_loss=0.09859, over 3059671.77 frames. ], batch size: 249, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:05:41,848 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 5750, loss[loss=0.2513, simple_loss=0.3319, pruned_loss=0.08534, over 16286.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3392, pruned_loss=0.0995, over 3046878.24 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:06:16,583 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:23,496 INFO [optim.py:368] (3/8) 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 08:06:29,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5037, 2.1016, 2.1229, 4.0811, 1.8303, 2.9925, 2.2598, 2.2087], device='cuda:3'), covar=tensor([0.0571, 0.2098, 0.1156, 0.0303, 0.2949, 0.1019, 0.1977, 0.2245], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0313, 0.0252, 0.0301, 0.0363, 0.0298, 0.0279, 0.0372], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:06:50,144 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6632, 5.9983, 5.6150, 5.8012, 5.2583, 5.0852, 5.4955, 6.0858], device='cuda:3'), covar=tensor([0.0630, 0.0537, 0.0986, 0.0387, 0.0597, 0.0488, 0.0542, 0.0569], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0467, 0.0410, 0.0308, 0.0294, 0.0311, 0.0379, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:07:20,512 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-28 08:07:26,288 INFO [train.py:904] (3/8) Epoch 5, batch 5800, loss[loss=0.2873, simple_loss=0.3662, pruned_loss=0.1042, over 16629.00 frames. ], tot_loss[loss=0.2657, simple_loss=0.3377, pruned_loss=0.09681, over 3064796.20 frames. ], batch size: 134, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:07:35,757 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:13,832 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:46,734 INFO [train.py:904] (3/8) Epoch 5, batch 5850, loss[loss=0.2601, simple_loss=0.3387, pruned_loss=0.09073, over 16413.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3346, pruned_loss=0.09458, over 3074796.36 frames. ], batch size: 146, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:09:06,684 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.458e+02 3.799e+02 4.852e+02 6.197e+02 1.255e+03, threshold=9.704e+02, percent-clipped=4.0 2023-04-28 08:09:29,288 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:10:09,068 INFO [train.py:904] (3/8) Epoch 5, batch 5900, loss[loss=0.2463, simple_loss=0.3294, pruned_loss=0.08162, over 16385.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3332, pruned_loss=0.09385, over 3082718.24 frames. ], batch size: 35, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:10:36,174 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:06,227 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4049, 4.4322, 4.2854, 4.1808, 3.7950, 4.3759, 4.2429, 4.0222], device='cuda:3'), covar=tensor([0.0514, 0.0328, 0.0205, 0.0177, 0.0891, 0.0310, 0.0337, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0180, 0.0202, 0.0176, 0.0231, 0.0203, 0.0147, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:11:21,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2248, 4.0567, 4.1937, 2.7787, 3.6580, 4.0305, 3.9238, 2.2467], device='cuda:3'), covar=tensor([0.0297, 0.0019, 0.0020, 0.0205, 0.0037, 0.0052, 0.0025, 0.0262], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0053, 0.0057, 0.0113, 0.0058, 0.0068, 0.0063, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 08:11:31,951 INFO [train.py:904] (3/8) Epoch 5, batch 5950, loss[loss=0.2808, simple_loss=0.3493, pruned_loss=0.1061, over 11395.00 frames. ], tot_loss[loss=0.2592, simple_loss=0.3343, pruned_loss=0.09209, over 3106252.50 frames. ], batch size: 246, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:11:52,723 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:53,494 INFO [optim.py:368] (3/8) 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 08:12:03,477 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8687, 4.1378, 3.8540, 3.9388, 3.6123, 3.7135, 3.8184, 4.0722], device='cuda:3'), covar=tensor([0.0804, 0.0797, 0.1057, 0.0504, 0.0712, 0.1380, 0.0720, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0364, 0.0480, 0.0415, 0.0314, 0.0301, 0.0319, 0.0390, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:12:43,382 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 6000, loss[loss=0.25, simple_loss=0.324, pruned_loss=0.08795, over 16308.00 frames. ], tot_loss[loss=0.2584, simple_loss=0.3335, pruned_loss=0.09169, over 3109119.02 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:12:51,974 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 08:13:04,053 INFO [train.py:938] (3/8) Epoch 5, validation: loss=0.1879, simple_loss=0.2992, pruned_loss=0.03826, over 944034.00 frames. 2023-04-28 08:13:04,054 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 08:13:05,746 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:23,131 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5405, 1.9922, 2.2518, 4.1720, 1.8332, 2.9671, 2.2903, 2.1830], device='cuda:3'), covar=tensor([0.0564, 0.2119, 0.1132, 0.0265, 0.3119, 0.1045, 0.1814, 0.2431], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0313, 0.0256, 0.0301, 0.0368, 0.0300, 0.0281, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:13:51,746 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:53,155 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3677, 2.0129, 1.4986, 1.7371, 2.4572, 2.1364, 2.5580, 2.6258], device='cuda:3'), covar=tensor([0.0046, 0.0159, 0.0229, 0.0211, 0.0084, 0.0174, 0.0072, 0.0091], device='cuda:3'), in_proj_covar=tensor([0.0077, 0.0153, 0.0154, 0.0151, 0.0146, 0.0155, 0.0125, 0.0134], device='cuda:3'), out_proj_covar=tensor([9.6459e-05, 1.8759e-04, 1.8486e-04, 1.8198e-04, 1.8147e-04, 1.9060e-04, 1.4975e-04, 1.6639e-04], device='cuda:3') 2023-04-28 08:14:27,392 INFO [train.py:904] (3/8) Epoch 5, batch 6050, loss[loss=0.2384, simple_loss=0.3282, pruned_loss=0.07426, over 17262.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3318, pruned_loss=0.09104, over 3095666.96 frames. ], batch size: 52, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:14:37,522 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 08:14:48,340 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:48,995 INFO [optim.py:368] (3/8) 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 08:15:33,667 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2418, 3.1097, 3.2302, 3.4121, 3.4450, 3.1354, 3.4388, 3.4654], device='cuda:3'), covar=tensor([0.0707, 0.0675, 0.1014, 0.0504, 0.0518, 0.1756, 0.0640, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0367, 0.0455, 0.0573, 0.0464, 0.0348, 0.0345, 0.0368, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:15:45,453 INFO [train.py:904] (3/8) Epoch 5, batch 6100, loss[loss=0.2306, simple_loss=0.308, pruned_loss=0.07661, over 17197.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3313, pruned_loss=0.08968, over 3108501.55 frames. ], batch size: 46, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:16:38,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0199, 4.2440, 1.9743, 4.7422, 2.8252, 4.5797, 2.2257, 3.1785], device='cuda:3'), covar=tensor([0.0100, 0.0167, 0.1500, 0.0021, 0.0660, 0.0225, 0.1331, 0.0473], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0145, 0.0176, 0.0080, 0.0159, 0.0183, 0.0184, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 08:17:01,199 INFO [train.py:904] (3/8) Epoch 5, batch 6150, loss[loss=0.2308, simple_loss=0.3027, pruned_loss=0.07941, over 17243.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3292, pruned_loss=0.08941, over 3099854.19 frames. ], batch size: 52, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:17:22,716 INFO [optim.py:368] (3/8) 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 08:17:48,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8806, 3.9294, 1.8149, 4.4911, 2.5998, 4.2791, 2.2582, 2.9272], device='cuda:3'), covar=tensor([0.0125, 0.0282, 0.1825, 0.0031, 0.0799, 0.0382, 0.1313, 0.0650], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0146, 0.0178, 0.0080, 0.0160, 0.0182, 0.0185, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 08:18:21,097 INFO [train.py:904] (3/8) Epoch 5, batch 6200, loss[loss=0.2525, simple_loss=0.3229, pruned_loss=0.09105, over 16716.00 frames. ], tot_loss[loss=0.2519, simple_loss=0.327, pruned_loss=0.08842, over 3098168.69 frames. ], batch size: 134, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:18:36,384 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:19:10,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6392, 3.8126, 2.9906, 2.2335, 2.7384, 2.2538, 3.9947, 3.7923], device='cuda:3'), covar=tensor([0.2279, 0.0635, 0.1377, 0.1791, 0.2005, 0.1461, 0.0453, 0.0707], device='cuda:3'), in_proj_covar=tensor([0.0279, 0.0251, 0.0271, 0.0244, 0.0294, 0.0202, 0.0241, 0.0254], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:19:20,430 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7264, 1.4305, 1.9658, 2.5871, 2.4896, 2.8926, 1.6266, 2.7388], device='cuda:3'), covar=tensor([0.0068, 0.0261, 0.0160, 0.0123, 0.0113, 0.0063, 0.0231, 0.0043], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0141, 0.0125, 0.0120, 0.0124, 0.0088, 0.0138, 0.0077], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 08:19:37,501 INFO [train.py:904] (3/8) Epoch 5, batch 6250, loss[loss=0.225, simple_loss=0.3147, pruned_loss=0.06762, over 17143.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3268, pruned_loss=0.08836, over 3110172.60 frames. ], batch size: 48, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:19:57,338 INFO [optim.py:368] (3/8) 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 08:19:58,244 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 08:19:59,881 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8739, 2.1567, 2.2393, 4.4976, 1.8790, 3.1312, 2.3282, 2.3833], device='cuda:3'), covar=tensor([0.0569, 0.2212, 0.1189, 0.0284, 0.3226, 0.1061, 0.1983, 0.2438], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0311, 0.0258, 0.0305, 0.0367, 0.0299, 0.0281, 0.0374], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:20:09,426 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:20:29,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9247, 3.9447, 3.8433, 3.2887, 3.8566, 1.7431, 3.6860, 3.6020], device='cuda:3'), covar=tensor([0.0087, 0.0059, 0.0105, 0.0270, 0.0070, 0.1822, 0.0094, 0.0137], device='cuda:3'), in_proj_covar=tensor([0.0091, 0.0078, 0.0119, 0.0125, 0.0091, 0.0140, 0.0105, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:20:38,832 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-28 08:20:55,067 INFO [train.py:904] (3/8) Epoch 5, batch 6300, loss[loss=0.239, simple_loss=0.3136, pruned_loss=0.08225, over 16637.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3276, pruned_loss=0.08828, over 3110255.30 frames. ], batch size: 57, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:21:43,816 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 6350, loss[loss=0.2582, simple_loss=0.3324, pruned_loss=0.09202, over 16665.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3293, pruned_loss=0.09074, over 3077278.78 frames. ], batch size: 134, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:22:13,437 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:22:22,795 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:31,679 INFO [optim.py:368] (3/8) 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 08:22:38,754 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1155, 3.8406, 3.0642, 1.5869, 2.6334, 2.0726, 3.4258, 3.7089], device='cuda:3'), covar=tensor([0.0265, 0.0418, 0.0708, 0.1995, 0.0963, 0.1104, 0.0702, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0127, 0.0154, 0.0141, 0.0134, 0.0126, 0.0141, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 08:22:56,443 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 6400, loss[loss=0.2434, simple_loss=0.3252, pruned_loss=0.08076, over 16719.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3286, pruned_loss=0.0911, over 3077619.90 frames. ], batch size: 89, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:24:42,540 INFO [train.py:904] (3/8) Epoch 5, batch 6450, loss[loss=0.2106, simple_loss=0.2999, pruned_loss=0.06064, over 16804.00 frames. ], tot_loss[loss=0.2527, simple_loss=0.3274, pruned_loss=0.08904, over 3090124.37 frames. ], batch size: 102, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:24:53,182 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0963, 4.0314, 4.0450, 2.7732, 3.9994, 1.4751, 3.7158, 3.6619], device='cuda:3'), covar=tensor([0.0139, 0.0109, 0.0131, 0.0627, 0.0107, 0.2528, 0.0147, 0.0292], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0077, 0.0118, 0.0125, 0.0090, 0.0141, 0.0105, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:24:55,576 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5960, 2.7283, 2.2178, 4.1688, 3.4315, 3.9026, 1.6314, 2.7334], device='cuda:3'), covar=tensor([0.1392, 0.0564, 0.1217, 0.0070, 0.0245, 0.0371, 0.1343, 0.0792], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0144, 0.0170, 0.0086, 0.0179, 0.0181, 0.0164, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 08:25:00,017 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1502, 3.0975, 2.6844, 2.0013, 2.5332, 2.0347, 2.8398, 2.9819], device='cuda:3'), covar=tensor([0.0327, 0.0417, 0.0482, 0.1359, 0.0664, 0.0933, 0.0452, 0.0551], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0125, 0.0152, 0.0139, 0.0132, 0.0124, 0.0138, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 08:25:01,516 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 08:25:02,352 INFO [optim.py:368] (3/8) 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 08:25:26,427 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6708, 4.1207, 4.3812, 1.7506, 4.6515, 4.6257, 3.1519, 3.3365], device='cuda:3'), covar=tensor([0.0760, 0.0127, 0.0128, 0.1213, 0.0035, 0.0040, 0.0338, 0.0353], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0089, 0.0080, 0.0141, 0.0071, 0.0077, 0.0116, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 08:25:59,699 INFO [train.py:904] (3/8) Epoch 5, batch 6500, loss[loss=0.244, simple_loss=0.3192, pruned_loss=0.08441, over 16519.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3252, pruned_loss=0.08799, over 3101306.41 frames. ], batch size: 75, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:26:48,141 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3471, 1.9778, 2.2453, 3.8006, 1.8916, 2.9714, 2.2307, 2.0969], device='cuda:3'), covar=tensor([0.0548, 0.1965, 0.1034, 0.0296, 0.2814, 0.0891, 0.1853, 0.2132], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0313, 0.0259, 0.0304, 0.0367, 0.0303, 0.0283, 0.0378], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:27:18,007 INFO [train.py:904] (3/8) Epoch 5, batch 6550, loss[loss=0.2523, simple_loss=0.3507, pruned_loss=0.07698, over 16484.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3281, pruned_loss=0.08907, over 3104779.42 frames. ], batch size: 75, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:27:37,105 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:28:33,656 INFO [train.py:904] (3/8) Epoch 5, batch 6600, loss[loss=0.3541, simple_loss=0.3933, pruned_loss=0.1575, over 11267.00 frames. ], tot_loss[loss=0.2554, simple_loss=0.3306, pruned_loss=0.09011, over 3091848.06 frames. ], batch size: 246, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:35,136 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:29:51,329 INFO [train.py:904] (3/8) Epoch 5, batch 6650, loss[loss=0.2326, simple_loss=0.3118, pruned_loss=0.07666, over 16808.00 frames. ], tot_loss[loss=0.2575, simple_loss=0.3318, pruned_loss=0.0916, over 3085633.57 frames. ], batch size: 83, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:52,485 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:02,738 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:11,437 INFO [optim.py:368] (3/8) 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 08:31:04,888 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 6700, loss[loss=0.2866, simple_loss=0.3442, pruned_loss=0.1145, over 15430.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.3307, pruned_loss=0.0919, over 3086191.19 frames. ], batch size: 191, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:31:09,290 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 5, batch 6750, loss[loss=0.2278, simple_loss=0.3101, pruned_loss=0.07277, over 16464.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.33, pruned_loss=0.09211, over 3073084.91 frames. ], batch size: 146, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:32:45,840 INFO [optim.py:368] (3/8) 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 08:33:08,543 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 6800, loss[loss=0.2819, simple_loss=0.3413, pruned_loss=0.1113, over 11406.00 frames. ], tot_loss[loss=0.2569, simple_loss=0.3303, pruned_loss=0.09181, over 3065325.30 frames. ], batch size: 246, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:34:40,939 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:34:57,633 INFO [train.py:904] (3/8) Epoch 5, batch 6850, loss[loss=0.2506, simple_loss=0.3417, pruned_loss=0.07978, over 16675.00 frames. ], tot_loss[loss=0.2569, simple_loss=0.3306, pruned_loss=0.0916, over 3074465.38 frames. ], batch size: 57, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:35:17,575 INFO [optim.py:368] (3/8) 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 08:35:21,077 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:35:29,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8101, 2.3422, 2.3313, 4.6222, 1.8707, 3.2513, 2.4083, 2.4538], device='cuda:3'), covar=tensor([0.0586, 0.2029, 0.1137, 0.0194, 0.3021, 0.0899, 0.1896, 0.2207], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0316, 0.0261, 0.0307, 0.0372, 0.0308, 0.0286, 0.0379], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:35:44,335 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8449, 2.9877, 2.6397, 4.9558, 4.0262, 4.4554, 1.9375, 3.2741], device='cuda:3'), covar=tensor([0.1339, 0.0565, 0.1035, 0.0052, 0.0250, 0.0221, 0.1198, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0141, 0.0167, 0.0085, 0.0179, 0.0176, 0.0161, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 08:35:56,727 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 08:36:12,141 INFO [train.py:904] (3/8) Epoch 5, batch 6900, loss[loss=0.2129, simple_loss=0.3015, pruned_loss=0.06219, over 17093.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.3328, pruned_loss=0.09087, over 3101215.22 frames. ], batch size: 49, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:36:13,837 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-28 08:36:23,471 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9366, 2.7057, 2.5953, 1.6876, 2.8196, 2.8123, 2.4387, 2.3283], device='cuda:3'), covar=tensor([0.0714, 0.0143, 0.0194, 0.0946, 0.0100, 0.0115, 0.0334, 0.0413], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0086, 0.0080, 0.0140, 0.0071, 0.0077, 0.0115, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 08:36:33,890 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:37:30,502 INFO [train.py:904] (3/8) Epoch 5, batch 6950, loss[loss=0.2567, simple_loss=0.3316, pruned_loss=0.09088, over 16420.00 frames. ], tot_loss[loss=0.261, simple_loss=0.335, pruned_loss=0.09348, over 3101519.08 frames. ], batch size: 146, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:37:48,336 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4544, 3.8713, 3.3054, 3.6593, 3.3063, 3.5367, 3.5627, 3.8420], device='cuda:3'), covar=tensor([0.2016, 0.1332, 0.2678, 0.1113, 0.1595, 0.2016, 0.1421, 0.1390], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0468, 0.0412, 0.0306, 0.0295, 0.0320, 0.0386, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:37:50,869 INFO [optim.py:368] (3/8) 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 08:38:42,822 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 7000, loss[loss=0.2881, simple_loss=0.3433, pruned_loss=0.1165, over 11419.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3353, pruned_loss=0.09312, over 3077088.73 frames. ], batch size: 246, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:38:54,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7671, 3.6035, 3.7172, 3.9543, 4.0048, 3.5859, 3.9516, 4.0046], device='cuda:3'), covar=tensor([0.0783, 0.0738, 0.1324, 0.0573, 0.0482, 0.1358, 0.0705, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0362, 0.0450, 0.0572, 0.0462, 0.0347, 0.0338, 0.0371, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:40:06,953 INFO [train.py:904] (3/8) Epoch 5, batch 7050, loss[loss=0.3006, simple_loss=0.3523, pruned_loss=0.1244, over 11542.00 frames. ], tot_loss[loss=0.2604, simple_loss=0.3353, pruned_loss=0.09279, over 3076844.27 frames. ], batch size: 247, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:40:26,882 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.724e+02 4.498e+02 5.703e+02 1.501e+03, threshold=8.996e+02, percent-clipped=4.0 2023-04-28 08:41:10,859 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5925, 2.2217, 1.5116, 1.9021, 2.9505, 2.6514, 3.4732, 3.3177], device='cuda:3'), covar=tensor([0.0019, 0.0207, 0.0293, 0.0243, 0.0099, 0.0174, 0.0067, 0.0083], device='cuda:3'), in_proj_covar=tensor([0.0075, 0.0152, 0.0155, 0.0151, 0.0146, 0.0157, 0.0127, 0.0134], device='cuda:3'), out_proj_covar=tensor([9.2622e-05, 1.8558e-04, 1.8512e-04, 1.8133e-04, 1.7958e-04, 1.9214e-04, 1.5039e-04, 1.6430e-04], device='cuda:3') 2023-04-28 08:41:26,196 INFO [train.py:904] (3/8) Epoch 5, batch 7100, loss[loss=0.2322, simple_loss=0.3081, pruned_loss=0.07812, over 16542.00 frames. ], tot_loss[loss=0.2605, simple_loss=0.3345, pruned_loss=0.09324, over 3065254.20 frames. ], batch size: 68, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:42:19,693 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:24,241 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:44,668 INFO [train.py:904] (3/8) Epoch 5, batch 7150, loss[loss=0.3038, simple_loss=0.3541, pruned_loss=0.1268, over 11519.00 frames. ], tot_loss[loss=0.258, simple_loss=0.3317, pruned_loss=0.09218, over 3069046.72 frames. ], batch size: 247, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:43:03,968 INFO [optim.py:368] (3/8) 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 08:43:39,996 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:43:57,147 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:44:00,012 INFO [train.py:904] (3/8) Epoch 5, batch 7200, loss[loss=0.2531, simple_loss=0.3201, pruned_loss=0.09309, over 11696.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3286, pruned_loss=0.08917, over 3078845.36 frames. ], batch size: 246, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:44:14,037 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:45:22,033 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:45:24,167 INFO [train.py:904] (3/8) Epoch 5, batch 7250, loss[loss=0.233, simple_loss=0.3068, pruned_loss=0.07961, over 16501.00 frames. ], tot_loss[loss=0.2524, simple_loss=0.3271, pruned_loss=0.08881, over 3059371.86 frames. ], batch size: 68, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:45:42,198 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9616, 2.9699, 2.6403, 1.9953, 2.5249, 1.9803, 2.7286, 2.8786], device='cuda:3'), covar=tensor([0.0260, 0.0408, 0.0488, 0.1366, 0.0670, 0.0903, 0.0552, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0125, 0.0154, 0.0141, 0.0133, 0.0124, 0.0140, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 08:45:43,496 INFO [optim.py:368] (3/8) 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 08:45:53,855 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:45:56,740 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3086, 2.8962, 2.5613, 2.2393, 2.2872, 2.1248, 2.8304, 2.9694], device='cuda:3'), covar=tensor([0.1741, 0.0652, 0.1176, 0.1434, 0.1742, 0.1361, 0.0410, 0.0639], device='cuda:3'), in_proj_covar=tensor([0.0279, 0.0248, 0.0270, 0.0243, 0.0288, 0.0200, 0.0241, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:46:34,243 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 7300, loss[loss=0.2366, simple_loss=0.3222, pruned_loss=0.07552, over 16671.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3255, pruned_loss=0.08794, over 3068392.86 frames. ], batch size: 134, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:47:13,928 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 08:47:49,460 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:47:56,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6161, 3.7031, 3.0145, 2.3578, 2.8121, 2.3699, 3.9732, 3.7178], device='cuda:3'), covar=tensor([0.2284, 0.0773, 0.1283, 0.1470, 0.2018, 0.1384, 0.0415, 0.0679], device='cuda:3'), in_proj_covar=tensor([0.0281, 0.0249, 0.0269, 0.0244, 0.0292, 0.0202, 0.0243, 0.0250], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:47:58,197 INFO [train.py:904] (3/8) Epoch 5, batch 7350, loss[loss=0.2826, simple_loss=0.3438, pruned_loss=0.1107, over 11125.00 frames. ], tot_loss[loss=0.2509, simple_loss=0.3254, pruned_loss=0.08816, over 3049295.30 frames. ], batch size: 247, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:48:17,667 INFO [optim.py:368] (3/8) 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 08:49:18,342 INFO [train.py:904] (3/8) Epoch 5, batch 7400, loss[loss=0.2993, simple_loss=0.347, pruned_loss=0.1258, over 11063.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3266, pruned_loss=0.08868, over 3060910.70 frames. ], batch size: 247, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:49:37,193 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 08:50:02,991 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1388, 2.9911, 2.8517, 3.3040, 3.2960, 3.0803, 3.1727, 3.3474], device='cuda:3'), covar=tensor([0.0942, 0.1088, 0.2390, 0.1015, 0.0985, 0.2061, 0.1366, 0.1060], device='cuda:3'), in_proj_covar=tensor([0.0368, 0.0455, 0.0579, 0.0470, 0.0347, 0.0342, 0.0370, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:50:13,270 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 7450, loss[loss=0.2536, simple_loss=0.3276, pruned_loss=0.08977, over 16602.00 frames. ], tot_loss[loss=0.2547, simple_loss=0.3286, pruned_loss=0.09037, over 3064895.73 frames. ], batch size: 57, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:51:00,812 INFO [optim.py:368] (3/8) 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 08:51:14,559 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:30,381 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 08:51:31,925 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:49,933 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:50,125 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6627, 1.3249, 1.4865, 1.5664, 1.8181, 1.8293, 1.3878, 1.6120], device='cuda:3'), covar=tensor([0.0100, 0.0157, 0.0087, 0.0129, 0.0088, 0.0057, 0.0169, 0.0037], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0140, 0.0123, 0.0121, 0.0125, 0.0089, 0.0137, 0.0078], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 08:52:00,138 INFO [train.py:904] (3/8) Epoch 5, batch 7500, loss[loss=0.2448, simple_loss=0.3201, pruned_loss=0.08475, over 16542.00 frames. ], tot_loss[loss=0.2541, simple_loss=0.3286, pruned_loss=0.08978, over 3054480.82 frames. ], batch size: 62, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:52:50,878 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:08,107 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 7550, loss[loss=0.2325, simple_loss=0.3085, pruned_loss=0.0783, over 16609.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3273, pruned_loss=0.08943, over 3062273.07 frames. ], batch size: 57, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:53:38,522 INFO [optim.py:368] (3/8) 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 08:53:40,714 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:54:34,589 INFO [train.py:904] (3/8) Epoch 5, batch 7600, loss[loss=0.2331, simple_loss=0.3202, pruned_loss=0.07303, over 17193.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3258, pruned_loss=0.08872, over 3075551.77 frames. ], batch size: 46, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:54:51,964 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:55:51,253 INFO [train.py:904] (3/8) Epoch 5, batch 7650, loss[loss=0.2476, simple_loss=0.3301, pruned_loss=0.08248, over 16575.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3277, pruned_loss=0.09049, over 3062878.79 frames. ], batch size: 75, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:56:12,528 INFO [optim.py:368] (3/8) 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 08:56:25,935 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 08:57:08,654 INFO [train.py:904] (3/8) Epoch 5, batch 7700, loss[loss=0.3175, simple_loss=0.3589, pruned_loss=0.138, over 11488.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.3286, pruned_loss=0.09193, over 3044959.87 frames. ], batch size: 250, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:57:44,771 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 7750, loss[loss=0.224, simple_loss=0.3068, pruned_loss=0.07063, over 16406.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.328, pruned_loss=0.09087, over 3071554.96 frames. ], batch size: 146, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:58:47,829 INFO [optim.py:368] (3/8) 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 08:59:20,076 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:28,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2694, 3.1713, 3.2412, 3.4420, 3.4177, 3.1965, 3.3981, 3.4431], device='cuda:3'), covar=tensor([0.0772, 0.0693, 0.1219, 0.0546, 0.0647, 0.1763, 0.0797, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0462, 0.0585, 0.0479, 0.0354, 0.0349, 0.0374, 0.0396], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 08:59:34,706 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:39,594 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 7800, loss[loss=0.2956, simple_loss=0.3454, pruned_loss=0.1229, over 11432.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3288, pruned_loss=0.09185, over 3063300.32 frames. ], batch size: 247, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:00:15,715 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 09:00:21,104 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:28,134 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:52,285 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:02,288 INFO [train.py:904] (3/8) Epoch 5, batch 7850, loss[loss=0.2528, simple_loss=0.3282, pruned_loss=0.08872, over 16816.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3292, pruned_loss=0.09091, over 3072824.97 frames. ], batch size: 124, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:01:13,115 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.309e+02 3.753e+02 4.508e+02 5.540e+02 1.129e+03, threshold=9.017e+02, percent-clipped=6.0 2023-04-28 09:01:23,990 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:01:52,505 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:04,081 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:16,621 INFO [train.py:904] (3/8) Epoch 5, batch 7900, loss[loss=0.3008, simple_loss=0.3494, pruned_loss=0.1261, over 11565.00 frames. ], tot_loss[loss=0.2531, simple_loss=0.3274, pruned_loss=0.08942, over 3078439.99 frames. ], batch size: 248, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:02:35,254 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:03:36,606 INFO [train.py:904] (3/8) Epoch 5, batch 7950, loss[loss=0.2224, simple_loss=0.3023, pruned_loss=0.07128, over 16808.00 frames. ], tot_loss[loss=0.254, simple_loss=0.3279, pruned_loss=0.0901, over 3083726.68 frames. ], batch size: 96, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:03:56,964 INFO [optim.py:368] (3/8) 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 09:04:02,930 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:04:50,730 INFO [train.py:904] (3/8) Epoch 5, batch 8000, loss[loss=0.2358, simple_loss=0.3152, pruned_loss=0.07817, over 16730.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3286, pruned_loss=0.09081, over 3085799.65 frames. ], batch size: 134, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:05:34,674 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8229, 1.2138, 1.6257, 1.6891, 1.8797, 1.8764, 1.4101, 1.6059], device='cuda:3'), covar=tensor([0.0077, 0.0162, 0.0085, 0.0111, 0.0084, 0.0054, 0.0155, 0.0040], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0145, 0.0126, 0.0125, 0.0130, 0.0094, 0.0143, 0.0081], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 09:05:48,682 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 09:06:05,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7189, 2.9061, 2.5153, 4.5730, 3.6218, 4.1938, 1.7560, 3.1716], device='cuda:3'), covar=tensor([0.1355, 0.0582, 0.1144, 0.0087, 0.0303, 0.0293, 0.1299, 0.0655], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0144, 0.0169, 0.0087, 0.0184, 0.0178, 0.0160, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 09:06:06,082 INFO [train.py:904] (3/8) Epoch 5, batch 8050, loss[loss=0.254, simple_loss=0.3341, pruned_loss=0.08699, over 16867.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3281, pruned_loss=0.09026, over 3088479.69 frames. ], batch size: 83, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:06:26,963 INFO [optim.py:368] (3/8) 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 09:06:47,915 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8763, 4.2393, 1.8619, 4.6605, 2.5297, 4.6076, 2.1054, 2.9955], device='cuda:3'), covar=tensor([0.0136, 0.0248, 0.1818, 0.0035, 0.0872, 0.0318, 0.1550, 0.0644], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0148, 0.0178, 0.0079, 0.0160, 0.0178, 0.0184, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 09:06:49,657 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:06:56,233 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6802, 3.8683, 3.0888, 2.3468, 2.9527, 2.2948, 4.1106, 4.0320], device='cuda:3'), covar=tensor([0.2181, 0.0616, 0.1229, 0.1577, 0.1819, 0.1417, 0.0359, 0.0651], device='cuda:3'), in_proj_covar=tensor([0.0280, 0.0246, 0.0269, 0.0242, 0.0290, 0.0201, 0.0240, 0.0253], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:07:21,588 INFO [train.py:904] (3/8) Epoch 5, batch 8100, loss[loss=0.3054, simple_loss=0.3562, pruned_loss=0.1273, over 11672.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3271, pruned_loss=0.08932, over 3078649.54 frames. ], batch size: 246, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:06,006 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 8150, loss[loss=0.2233, simple_loss=0.302, pruned_loss=0.07233, over 16641.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3248, pruned_loss=0.0883, over 3066231.13 frames. ], batch size: 62, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:44,092 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 09:09:20,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:24,936 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:00,528 INFO [train.py:904] (3/8) Epoch 5, batch 8200, loss[loss=0.2623, simple_loss=0.3348, pruned_loss=0.09492, over 16350.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3219, pruned_loss=0.08741, over 3065229.12 frames. ], batch size: 146, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:10:32,482 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:54,417 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4853, 3.4080, 3.1091, 1.8465, 2.6671, 2.1946, 2.8957, 3.2931], device='cuda:3'), covar=tensor([0.0423, 0.0509, 0.0437, 0.1541, 0.0751, 0.0904, 0.0860, 0.0609], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0123, 0.0151, 0.0138, 0.0130, 0.0123, 0.0138, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 09:11:00,853 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:11:18,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8169, 3.6693, 3.9100, 4.1026, 4.1672, 3.7311, 4.1494, 4.1329], device='cuda:3'), covar=tensor([0.1118, 0.0823, 0.1196, 0.0548, 0.0498, 0.1287, 0.0519, 0.0530], device='cuda:3'), in_proj_covar=tensor([0.0372, 0.0461, 0.0578, 0.0482, 0.0363, 0.0351, 0.0377, 0.0405], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:11:22,107 INFO [train.py:904] (3/8) Epoch 5, batch 8250, loss[loss=0.2231, simple_loss=0.3121, pruned_loss=0.06711, over 16836.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3215, pruned_loss=0.08525, over 3059736.97 frames. ], batch size: 96, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:11:27,348 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6562, 1.2652, 1.5492, 1.6430, 1.8172, 1.8318, 1.4367, 1.7359], device='cuda:3'), covar=tensor([0.0090, 0.0205, 0.0094, 0.0131, 0.0112, 0.0084, 0.0190, 0.0059], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0147, 0.0129, 0.0126, 0.0132, 0.0094, 0.0143, 0.0079], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 09:11:44,535 INFO [optim.py:368] (3/8) 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 09:11:50,741 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:12:13,763 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:40,537 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 8300, loss[loss=0.1996, simple_loss=0.2788, pruned_loss=0.06025, over 12179.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3181, pruned_loss=0.08166, over 3059363.37 frames. ], batch size: 246, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:12:54,762 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8246, 4.0521, 4.1103, 3.1285, 3.9052, 4.1828, 3.9710, 2.3192], device='cuda:3'), covar=tensor([0.0240, 0.0016, 0.0018, 0.0170, 0.0026, 0.0029, 0.0026, 0.0273], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0051, 0.0056, 0.0109, 0.0057, 0.0065, 0.0060, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 09:13:08,659 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:13:25,759 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-28 09:14:04,025 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9686, 3.2816, 2.9401, 4.7219, 3.8041, 4.4060, 1.7362, 3.2898], device='cuda:3'), covar=tensor([0.1330, 0.0511, 0.0886, 0.0073, 0.0186, 0.0254, 0.1347, 0.0608], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0138, 0.0163, 0.0083, 0.0174, 0.0172, 0.0158, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 09:14:06,483 INFO [train.py:904] (3/8) Epoch 5, batch 8350, loss[loss=0.241, simple_loss=0.3229, pruned_loss=0.07956, over 15299.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3165, pruned_loss=0.07872, over 3063192.17 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:14:30,518 INFO [optim.py:368] (3/8) 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 09:14:54,753 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 8400, loss[loss=0.203, simple_loss=0.2945, pruned_loss=0.05577, over 16731.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3129, pruned_loss=0.07613, over 3042665.43 frames. ], batch size: 124, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:15:48,895 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5242, 3.4991, 3.4723, 2.9798, 3.4305, 2.0704, 3.0740, 3.0018], device='cuda:3'), covar=tensor([0.0079, 0.0070, 0.0079, 0.0197, 0.0057, 0.1539, 0.0086, 0.0126], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0077, 0.0121, 0.0121, 0.0089, 0.0143, 0.0104, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:15:54,198 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2598, 4.2610, 4.1426, 3.5779, 4.1499, 1.6274, 3.8941, 3.9885], device='cuda:3'), covar=tensor([0.0057, 0.0049, 0.0087, 0.0236, 0.0059, 0.1790, 0.0091, 0.0126], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0077, 0.0121, 0.0121, 0.0089, 0.0143, 0.0105, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:16:11,874 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7324, 3.6619, 3.8476, 3.9890, 4.0350, 3.6050, 4.0279, 4.0340], device='cuda:3'), covar=tensor([0.0840, 0.0666, 0.0944, 0.0487, 0.0410, 0.1285, 0.0451, 0.0423], device='cuda:3'), in_proj_covar=tensor([0.0358, 0.0450, 0.0559, 0.0469, 0.0351, 0.0341, 0.0368, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:16:13,311 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 09:16:14,934 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:16:51,009 INFO [train.py:904] (3/8) Epoch 5, batch 8450, loss[loss=0.215, simple_loss=0.3035, pruned_loss=0.06327, over 16200.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3101, pruned_loss=0.07372, over 3042351.98 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:54,723 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:17:13,779 INFO [optim.py:368] (3/8) 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 09:17:20,247 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7375, 2.7645, 1.7164, 2.8212, 2.1658, 2.7938, 2.0625, 2.4973], device='cuda:3'), covar=tensor([0.0173, 0.0284, 0.1178, 0.0072, 0.0665, 0.0451, 0.1108, 0.0504], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0147, 0.0175, 0.0076, 0.0159, 0.0174, 0.0183, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 09:17:30,177 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9873, 3.6824, 3.4983, 1.7568, 2.8696, 2.3465, 3.3437, 3.5334], device='cuda:3'), covar=tensor([0.0274, 0.0567, 0.0427, 0.1757, 0.0689, 0.1005, 0.0718, 0.0842], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0120, 0.0147, 0.0136, 0.0128, 0.0123, 0.0134, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 09:17:38,430 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:13,181 INFO [train.py:904] (3/8) Epoch 5, batch 8500, loss[loss=0.2161, simple_loss=0.2967, pruned_loss=0.06772, over 15203.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3056, pruned_loss=0.07081, over 3030317.09 frames. ], batch size: 190, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:18:13,692 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:57,907 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:19:38,543 INFO [train.py:904] (3/8) Epoch 5, batch 8550, loss[loss=0.2289, simple_loss=0.3072, pruned_loss=0.07532, over 16714.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3026, pruned_loss=0.06903, over 3034680.36 frames. ], batch size: 57, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:20:04,106 INFO [optim.py:368] (3/8) 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 09:20:29,053 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:20:45,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.11 vs. limit=2.0 2023-04-28 09:21:04,488 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:17,317 INFO [train.py:904] (3/8) Epoch 5, batch 8600, loss[loss=0.2135, simple_loss=0.2882, pruned_loss=0.06935, over 12370.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3029, pruned_loss=0.06823, over 3014129.17 frames. ], batch size: 249, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:22:28,802 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:22:48,980 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 09:22:58,656 INFO [train.py:904] (3/8) Epoch 5, batch 8650, loss[loss=0.2198, simple_loss=0.3111, pruned_loss=0.06424, over 16433.00 frames. ], tot_loss[loss=0.217, simple_loss=0.301, pruned_loss=0.06649, over 3013343.11 frames. ], batch size: 146, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:23:33,201 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1081, 3.3543, 3.5875, 3.5468, 3.5463, 3.2870, 3.3570, 3.3821], device='cuda:3'), covar=tensor([0.0297, 0.0439, 0.0356, 0.0441, 0.0446, 0.0367, 0.0721, 0.0380], device='cuda:3'), in_proj_covar=tensor([0.0226, 0.0218, 0.0225, 0.0222, 0.0272, 0.0237, 0.0332, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 09:23:33,946 INFO [optim.py:368] (3/8) 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 09:23:39,014 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3416, 3.6654, 3.5978, 2.7039, 3.4571, 3.6431, 3.5194, 2.1921], device='cuda:3'), covar=tensor([0.0264, 0.0016, 0.0022, 0.0177, 0.0033, 0.0026, 0.0029, 0.0263], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0051, 0.0056, 0.0109, 0.0057, 0.0063, 0.0060, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 09:24:39,628 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:24:47,014 INFO [train.py:904] (3/8) Epoch 5, batch 8700, loss[loss=0.211, simple_loss=0.2991, pruned_loss=0.06148, over 16720.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2973, pruned_loss=0.06409, over 3029582.61 frames. ], batch size: 134, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:26:24,150 INFO [train.py:904] (3/8) Epoch 5, batch 8750, loss[loss=0.2406, simple_loss=0.3227, pruned_loss=0.07921, over 15516.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2974, pruned_loss=0.06343, over 3050302.69 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:27:05,487 INFO [optim.py:368] (3/8) 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 09:27:50,426 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8068, 3.7819, 4.1966, 4.1881, 4.1651, 3.8305, 3.8960, 3.8173], device='cuda:3'), covar=tensor([0.0231, 0.0326, 0.0314, 0.0398, 0.0381, 0.0305, 0.0716, 0.0328], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0213, 0.0221, 0.0218, 0.0263, 0.0233, 0.0324, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 09:27:50,467 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0836, 4.1844, 3.9244, 3.8269, 3.5858, 4.0745, 3.8156, 3.7432], device='cuda:3'), covar=tensor([0.0376, 0.0225, 0.0220, 0.0171, 0.0696, 0.0259, 0.0451, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0173, 0.0199, 0.0169, 0.0220, 0.0196, 0.0140, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:28:17,602 INFO [train.py:904] (3/8) Epoch 5, batch 8800, loss[loss=0.2315, simple_loss=0.3158, pruned_loss=0.07362, over 16319.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2961, pruned_loss=0.06215, over 3066063.88 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:28:21,225 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 09:28:26,847 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-28 09:28:51,772 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2849, 5.1939, 4.9694, 4.4563, 5.0738, 1.6206, 4.8077, 4.9693], device='cuda:3'), covar=tensor([0.0041, 0.0040, 0.0077, 0.0178, 0.0041, 0.1759, 0.0064, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0076, 0.0120, 0.0115, 0.0089, 0.0141, 0.0103, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:29:01,950 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:29:24,455 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:30:04,511 INFO [train.py:904] (3/8) Epoch 5, batch 8850, loss[loss=0.2113, simple_loss=0.3093, pruned_loss=0.05659, over 16594.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2988, pruned_loss=0.0614, over 3069364.11 frames. ], batch size: 62, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:30:27,398 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 09:30:38,663 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 3.290e+02 4.056e+02 5.025e+02 1.173e+03, threshold=8.111e+02, percent-clipped=7.0 2023-04-28 09:31:02,522 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:16,941 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:37,611 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:31:39,700 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:54,238 INFO [train.py:904] (3/8) Epoch 5, batch 8900, loss[loss=0.2029, simple_loss=0.2945, pruned_loss=0.05564, over 16910.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2986, pruned_loss=0.06061, over 3070240.21 frames. ], batch size: 96, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:32:20,019 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 09:32:48,142 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:39,420 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:54,499 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 09:33:59,797 INFO [train.py:904] (3/8) Epoch 5, batch 8950, loss[loss=0.1923, simple_loss=0.281, pruned_loss=0.0518, over 17253.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2998, pruned_loss=0.06237, over 3044994.11 frames. ], batch size: 52, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:34:28,499 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 09:34:35,629 INFO [optim.py:368] (3/8) 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 09:34:36,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7017, 3.6864, 4.0818, 4.0304, 4.0120, 3.7503, 3.7773, 3.7881], device='cuda:3'), covar=tensor([0.0226, 0.0389, 0.0272, 0.0358, 0.0381, 0.0272, 0.0663, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0219, 0.0211, 0.0217, 0.0216, 0.0265, 0.0233, 0.0320, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 09:34:42,546 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0713, 3.2428, 3.3025, 1.6394, 3.4828, 3.5312, 3.0216, 2.7909], device='cuda:3'), covar=tensor([0.0817, 0.0141, 0.0152, 0.1227, 0.0060, 0.0063, 0.0277, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0087, 0.0077, 0.0141, 0.0067, 0.0075, 0.0114, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 09:35:33,159 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:35:51,805 INFO [train.py:904] (3/8) Epoch 5, batch 9000, loss[loss=0.2015, simple_loss=0.2863, pruned_loss=0.05837, over 15389.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2956, pruned_loss=0.06033, over 3044895.59 frames. ], batch size: 191, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:35:51,806 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 09:36:02,103 INFO [train.py:938] (3/8) Epoch 5, validation: loss=0.1735, simple_loss=0.2766, pruned_loss=0.0352, over 944034.00 frames. 2023-04-28 09:36:02,104 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 09:37:44,738 INFO [train.py:904] (3/8) Epoch 5, batch 9050, loss[loss=0.1974, simple_loss=0.2798, pruned_loss=0.05752, over 16802.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2966, pruned_loss=0.06117, over 3048679.91 frames. ], batch size: 124, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:38:11,777 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 09:38:18,678 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.141e+02 3.840e+02 5.023e+02 8.628e+02, threshold=7.679e+02, percent-clipped=5.0 2023-04-28 09:38:19,717 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-28 09:39:06,971 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9614, 3.1862, 3.1225, 1.7327, 3.3548, 3.3837, 2.7342, 2.6336], device='cuda:3'), covar=tensor([0.0781, 0.0133, 0.0150, 0.1100, 0.0058, 0.0085, 0.0344, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0086, 0.0076, 0.0139, 0.0067, 0.0075, 0.0114, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 09:39:29,262 INFO [train.py:904] (3/8) Epoch 5, batch 9100, loss[loss=0.2263, simple_loss=0.3159, pruned_loss=0.0683, over 15476.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2962, pruned_loss=0.06168, over 3047126.83 frames. ], batch size: 191, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:39:49,426 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8772, 3.7458, 4.2399, 4.1836, 4.2018, 3.8952, 3.9204, 3.8545], device='cuda:3'), covar=tensor([0.0207, 0.0410, 0.0287, 0.0370, 0.0289, 0.0262, 0.0627, 0.0328], device='cuda:3'), in_proj_covar=tensor([0.0222, 0.0218, 0.0223, 0.0222, 0.0273, 0.0239, 0.0329, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 09:41:26,341 INFO [train.py:904] (3/8) Epoch 5, batch 9150, loss[loss=0.1861, simple_loss=0.2773, pruned_loss=0.04742, over 16504.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2968, pruned_loss=0.0611, over 3052021.20 frames. ], batch size: 68, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:42:00,581 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:47,441 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 09:42:52,196 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3198, 3.5774, 3.4430, 2.4374, 3.3245, 3.4789, 3.4473, 1.9718], device='cuda:3'), covar=tensor([0.0244, 0.0014, 0.0030, 0.0209, 0.0034, 0.0040, 0.0026, 0.0281], device='cuda:3'), in_proj_covar=tensor([0.0112, 0.0052, 0.0057, 0.0112, 0.0056, 0.0063, 0.0060, 0.0105], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 09:42:58,556 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 09:43:09,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0316, 3.8895, 3.9494, 2.8774, 3.9255, 1.4511, 3.6023, 3.6276], device='cuda:3'), covar=tensor([0.0134, 0.0112, 0.0147, 0.0527, 0.0113, 0.2643, 0.0162, 0.0250], device='cuda:3'), in_proj_covar=tensor([0.0090, 0.0076, 0.0121, 0.0114, 0.0089, 0.0143, 0.0103, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:43:09,755 INFO [train.py:904] (3/8) Epoch 5, batch 9200, loss[loss=0.2078, simple_loss=0.2959, pruned_loss=0.0598, over 16264.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2919, pruned_loss=0.05972, over 3066148.47 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:43:58,023 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 9250, loss[loss=0.2098, simple_loss=0.303, pruned_loss=0.05837, over 16690.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2908, pruned_loss=0.0594, over 3046641.22 frames. ], batch size: 134, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:45:18,929 INFO [optim.py:368] (3/8) 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 09:46:16,149 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8493, 3.0721, 3.0444, 2.0641, 2.9876, 3.0835, 2.8922, 1.6380], device='cuda:3'), covar=tensor([0.0322, 0.0024, 0.0038, 0.0246, 0.0040, 0.0049, 0.0043, 0.0345], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0052, 0.0056, 0.0111, 0.0056, 0.0062, 0.0060, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 09:46:21,458 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:46:40,052 INFO [train.py:904] (3/8) Epoch 5, batch 9300, loss[loss=0.1946, simple_loss=0.2706, pruned_loss=0.05926, over 12530.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2886, pruned_loss=0.05848, over 3037704.45 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:48:05,855 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:48:26,418 INFO [train.py:904] (3/8) Epoch 5, batch 9350, loss[loss=0.1895, simple_loss=0.2812, pruned_loss=0.04889, over 16902.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2884, pruned_loss=0.05812, over 3060122.37 frames. ], batch size: 116, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:49:00,529 INFO [optim.py:368] (3/8) 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 09:50:11,379 INFO [train.py:904] (3/8) Epoch 5, batch 9400, loss[loss=0.2221, simple_loss=0.3145, pruned_loss=0.06488, over 16228.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2887, pruned_loss=0.05784, over 3065449.13 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:51:06,440 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8427, 1.2450, 1.5177, 1.7154, 1.8154, 1.8694, 1.4272, 1.6496], device='cuda:3'), covar=tensor([0.0100, 0.0175, 0.0103, 0.0137, 0.0117, 0.0085, 0.0205, 0.0041], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0141, 0.0127, 0.0122, 0.0127, 0.0088, 0.0139, 0.0075], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 09:51:51,672 INFO [train.py:904] (3/8) Epoch 5, batch 9450, loss[loss=0.1941, simple_loss=0.2827, pruned_loss=0.05279, over 15235.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2908, pruned_loss=0.05867, over 3054295.21 frames. ], batch size: 191, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:52:21,841 INFO [optim.py:368] (3/8) 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 09:52:33,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5814, 3.8507, 4.0234, 2.9496, 3.7870, 4.0448, 3.7963, 2.3276], device='cuda:3'), covar=tensor([0.0276, 0.0019, 0.0025, 0.0204, 0.0035, 0.0038, 0.0038, 0.0295], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0052, 0.0056, 0.0111, 0.0057, 0.0063, 0.0060, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 09:52:47,306 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:53:09,067 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:53:33,285 INFO [train.py:904] (3/8) Epoch 5, batch 9500, loss[loss=0.1878, simple_loss=0.2793, pruned_loss=0.04818, over 16777.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2902, pruned_loss=0.05833, over 3047815.38 frames. ], batch size: 83, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:53:39,422 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 09:54:18,247 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:25,574 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:29,300 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:44,424 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:55:18,505 INFO [train.py:904] (3/8) Epoch 5, batch 9550, loss[loss=0.2401, simple_loss=0.3283, pruned_loss=0.07598, over 16176.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2902, pruned_loss=0.05829, over 3074407.79 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:55:53,351 INFO [optim.py:368] (3/8) 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 09:56:38,112 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:56:45,210 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0159, 2.2694, 2.4441, 4.7736, 2.0277, 3.2890, 2.5803, 2.5675], device='cuda:3'), covar=tensor([0.0469, 0.2044, 0.1082, 0.0190, 0.2983, 0.0976, 0.1761, 0.2280], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0306, 0.0259, 0.0302, 0.0364, 0.0305, 0.0285, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 09:56:50,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9725, 3.1306, 3.1704, 1.5330, 3.3675, 3.3847, 2.7316, 2.5165], device='cuda:3'), covar=tensor([0.0834, 0.0140, 0.0114, 0.1254, 0.0058, 0.0069, 0.0346, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0089, 0.0078, 0.0141, 0.0068, 0.0076, 0.0114, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 09:56:52,892 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:56:59,038 INFO [train.py:904] (3/8) Epoch 5, batch 9600, loss[loss=0.1738, simple_loss=0.2646, pruned_loss=0.04153, over 17134.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2917, pruned_loss=0.05915, over 3074866.95 frames. ], batch size: 49, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:57:46,791 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-04-28 09:58:43,805 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 09:58:47,432 INFO [train.py:904] (3/8) Epoch 5, batch 9650, loss[loss=0.2072, simple_loss=0.2971, pruned_loss=0.05863, over 16615.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2937, pruned_loss=0.05946, over 3082864.79 frames. ], batch size: 134, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:59:05,757 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:59:27,461 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 2.991e+02 3.657e+02 4.626e+02 9.582e+02, threshold=7.315e+02, percent-clipped=7.0 2023-04-28 10:00:34,986 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7606, 3.6085, 3.8165, 3.9794, 4.0142, 3.6149, 4.0122, 4.0244], device='cuda:3'), covar=tensor([0.0844, 0.0709, 0.1028, 0.0447, 0.0440, 0.1252, 0.0478, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0431, 0.0540, 0.0446, 0.0331, 0.0329, 0.0354, 0.0371], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:00:35,786 INFO [train.py:904] (3/8) Epoch 5, batch 9700, loss[loss=0.2008, simple_loss=0.292, pruned_loss=0.05483, over 15284.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2921, pruned_loss=0.0588, over 3066946.44 frames. ], batch size: 190, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:18,557 INFO [train.py:904] (3/8) Epoch 5, batch 9750, loss[loss=0.2033, simple_loss=0.2955, pruned_loss=0.05553, over 16695.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2914, pruned_loss=0.05902, over 3062918.87 frames. ], batch size: 134, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:50,575 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 3.120e+02 3.818e+02 4.621e+02 8.897e+02, threshold=7.636e+02, percent-clipped=1.0 2023-04-28 10:03:32,217 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 10:03:56,536 INFO [train.py:904] (3/8) Epoch 5, batch 9800, loss[loss=0.2043, simple_loss=0.2994, pruned_loss=0.05458, over 16989.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2916, pruned_loss=0.05799, over 3076423.04 frames. ], batch size: 109, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:04:36,919 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:05:22,397 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9174, 3.8731, 4.4502, 4.3707, 4.3810, 4.0292, 4.0206, 3.8992], device='cuda:3'), covar=tensor([0.0247, 0.0511, 0.0274, 0.0382, 0.0380, 0.0280, 0.0751, 0.0387], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0210, 0.0211, 0.0212, 0.0255, 0.0226, 0.0313, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-04-28 10:05:41,977 INFO [train.py:904] (3/8) Epoch 5, batch 9850, loss[loss=0.1929, simple_loss=0.2859, pruned_loss=0.04993, over 15391.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2922, pruned_loss=0.05744, over 3058728.03 frames. ], batch size: 191, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:06:14,680 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:06:52,948 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 9900, loss[loss=0.2128, simple_loss=0.3047, pruned_loss=0.06049, over 15350.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2925, pruned_loss=0.05732, over 3064028.60 frames. ], batch size: 191, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:08:24,199 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 10:09:17,062 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 9950, loss[loss=0.2167, simple_loss=0.3092, pruned_loss=0.06208, over 16485.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2957, pruned_loss=0.05849, over 3072436.26 frames. ], batch size: 147, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:33,660 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:10:04,528 INFO [optim.py:368] (3/8) 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 10:10:22,572 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7470, 4.9987, 5.0612, 5.0706, 5.0230, 5.5444, 5.1683, 4.8972], device='cuda:3'), covar=tensor([0.0706, 0.1392, 0.1369, 0.1435, 0.2230, 0.0907, 0.0947, 0.1837], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0355, 0.0348, 0.0305, 0.0409, 0.0380, 0.0287, 0.0405], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:10:42,981 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2541, 4.0199, 4.2654, 4.4573, 4.6090, 4.1024, 4.6386, 4.4437], device='cuda:3'), covar=tensor([0.0877, 0.0753, 0.1171, 0.0525, 0.0399, 0.0765, 0.0364, 0.0524], device='cuda:3'), in_proj_covar=tensor([0.0352, 0.0446, 0.0558, 0.0462, 0.0342, 0.0337, 0.0360, 0.0382], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:11:29,333 INFO [train.py:904] (3/8) Epoch 5, batch 10000, loss[loss=0.203, simple_loss=0.2814, pruned_loss=0.0623, over 12791.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2939, pruned_loss=0.05784, over 3074715.58 frames. ], batch size: 250, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:11:33,774 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2356, 3.1428, 3.0423, 3.4250, 3.3793, 3.2110, 3.4164, 3.3949], device='cuda:3'), covar=tensor([0.0970, 0.0939, 0.1781, 0.0938, 0.0961, 0.1940, 0.1016, 0.0992], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0450, 0.0561, 0.0465, 0.0346, 0.0338, 0.0362, 0.0385], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:11:43,865 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 5, batch 10050, loss[loss=0.1991, simple_loss=0.2888, pruned_loss=0.05474, over 16526.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2933, pruned_loss=0.05733, over 3080133.77 frames. ], batch size: 68, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:13:38,880 INFO [optim.py:368] (3/8) 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 10:13:55,827 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8840, 1.9727, 2.1364, 3.1484, 1.9457, 2.4169, 2.2080, 1.9571], device='cuda:3'), covar=tensor([0.0533, 0.2051, 0.1129, 0.0378, 0.2937, 0.1245, 0.1964, 0.2338], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0305, 0.0261, 0.0298, 0.0359, 0.0308, 0.0286, 0.0357], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:14:26,223 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8825, 2.6971, 2.6128, 1.9297, 2.6171, 2.6038, 2.6181, 1.7595], device='cuda:3'), covar=tensor([0.0266, 0.0028, 0.0041, 0.0199, 0.0056, 0.0058, 0.0046, 0.0303], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0054, 0.0058, 0.0115, 0.0060, 0.0065, 0.0063, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 10:14:38,998 INFO [train.py:904] (3/8) Epoch 5, batch 10100, loss[loss=0.2044, simple_loss=0.2816, pruned_loss=0.06361, over 12304.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2935, pruned_loss=0.05736, over 3077915.33 frames. ], batch size: 248, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:15:47,863 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2882, 4.4043, 4.5038, 4.4260, 4.4208, 4.8859, 4.6226, 4.3647], device='cuda:3'), covar=tensor([0.1046, 0.1527, 0.1189, 0.1569, 0.2105, 0.0983, 0.1040, 0.2098], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0355, 0.0351, 0.0308, 0.0406, 0.0381, 0.0289, 0.0409], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:16:20,269 INFO [train.py:904] (3/8) Epoch 6, batch 0, loss[loss=0.3562, simple_loss=0.3716, pruned_loss=0.1704, over 16748.00 frames. ], tot_loss[loss=0.3562, simple_loss=0.3716, pruned_loss=0.1704, over 16748.00 frames. ], batch size: 83, lr: 1.19e-02, grad_scale: 8.0 2023-04-28 10:16:20,269 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 10:16:27,647 INFO [train.py:938] (3/8) Epoch 6, validation: loss=0.1727, simple_loss=0.2755, pruned_loss=0.03501, over 944034.00 frames. 2023-04-28 10:16:27,648 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 10:16:52,395 INFO [optim.py:368] (3/8) 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 10:17:10,704 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:13,002 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 50, loss[loss=0.2295, simple_loss=0.3132, pruned_loss=0.07289, over 17065.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3167, pruned_loss=0.09102, over 752795.85 frames. ], batch size: 50, lr: 1.19e-02, grad_scale: 2.0 2023-04-28 10:17:46,427 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:18:20,193 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:34,193 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:47,052 INFO [train.py:904] (3/8) Epoch 6, batch 100, loss[loss=0.2151, simple_loss=0.3058, pruned_loss=0.06216, over 16696.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3099, pruned_loss=0.08526, over 1320114.90 frames. ], batch size: 57, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:18:49,540 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:19:11,789 INFO [optim.py:368] (3/8) 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 10:19:12,240 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 10:19:54,850 INFO [train.py:904] (3/8) Epoch 6, batch 150, loss[loss=0.2069, simple_loss=0.2759, pruned_loss=0.06888, over 16801.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3046, pruned_loss=0.08029, over 1766525.80 frames. ], batch size: 83, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:19:55,126 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:56,132 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:20:10,579 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 10:20:28,866 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 10:20:57,624 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:21:05,282 INFO [train.py:904] (3/8) Epoch 6, batch 200, loss[loss=0.2101, simple_loss=0.2998, pruned_loss=0.06022, over 16661.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3037, pruned_loss=0.07858, over 2096365.91 frames. ], batch size: 57, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:21:28,625 INFO [optim.py:368] (3/8) 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 10:21:51,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9657, 4.8826, 4.7702, 4.5587, 4.2481, 4.7539, 4.7602, 4.3889], device='cuda:3'), covar=tensor([0.0469, 0.0322, 0.0213, 0.0195, 0.0963, 0.0344, 0.0308, 0.0548], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0195, 0.0214, 0.0184, 0.0245, 0.0222, 0.0150, 0.0249], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 10:22:12,665 INFO [train.py:904] (3/8) Epoch 6, batch 250, loss[loss=0.1755, simple_loss=0.2645, pruned_loss=0.04328, over 16862.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.3011, pruned_loss=0.07778, over 2373885.85 frames. ], batch size: 42, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:22:20,835 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:20,591 INFO [train.py:904] (3/8) Epoch 6, batch 300, loss[loss=0.235, simple_loss=0.2963, pruned_loss=0.08686, over 16810.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.2979, pruned_loss=0.07527, over 2584784.78 frames. ], batch size: 124, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:23:37,438 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:45,701 INFO [optim.py:368] (3/8) 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 10:23:49,996 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 10:24:30,407 INFO [train.py:904] (3/8) Epoch 6, batch 350, loss[loss=0.1994, simple_loss=0.2786, pruned_loss=0.06011, over 17213.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.295, pruned_loss=0.0739, over 2730527.94 frames. ], batch size: 45, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:25:01,869 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:19,519 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 400, loss[loss=0.2065, simple_loss=0.2775, pruned_loss=0.0677, over 16532.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2917, pruned_loss=0.07136, over 2853160.54 frames. ], batch size: 68, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:25:55,420 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.109e+02 2.940e+02 3.610e+02 4.239e+02 7.005e+02, threshold=7.220e+02, percent-clipped=1.0 2023-04-28 10:26:05,630 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:26:45,054 INFO [train.py:904] (3/8) Epoch 6, batch 450, loss[loss=0.2088, simple_loss=0.2756, pruned_loss=0.07103, over 16867.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.29, pruned_loss=0.07073, over 2953027.56 frames. ], batch size: 116, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:26:47,062 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:29,546 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:27:52,965 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 500, loss[loss=0.2196, simple_loss=0.2828, pruned_loss=0.07825, over 15552.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2866, pruned_loss=0.0683, over 3036198.79 frames. ], batch size: 191, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:28:17,360 INFO [optim.py:368] (3/8) 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 10:28:57,694 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9254, 3.7715, 3.3710, 5.2366, 4.6435, 4.8101, 1.9712, 3.6388], device='cuda:3'), covar=tensor([0.1319, 0.0455, 0.0836, 0.0083, 0.0296, 0.0306, 0.1213, 0.0585], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0144, 0.0171, 0.0093, 0.0177, 0.0184, 0.0164, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 10:29:01,774 INFO [train.py:904] (3/8) Epoch 6, batch 550, loss[loss=0.1917, simple_loss=0.2806, pruned_loss=0.05142, over 16702.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2852, pruned_loss=0.06734, over 3101708.78 frames. ], batch size: 57, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:29:03,263 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:29:11,042 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4743, 4.5172, 4.6364, 4.5796, 4.5350, 5.1817, 4.7689, 4.4285], device='cuda:3'), covar=tensor([0.1347, 0.1795, 0.1535, 0.1884, 0.2979, 0.1068, 0.1448, 0.2763], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0417, 0.0410, 0.0359, 0.0484, 0.0439, 0.0336, 0.0482], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 10:29:16,079 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 10:29:30,768 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2351, 2.2174, 2.4069, 4.8522, 2.1476, 3.2483, 2.5386, 2.5279], device='cuda:3'), covar=tensor([0.0493, 0.2452, 0.1248, 0.0234, 0.3007, 0.1109, 0.1902, 0.2663], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0324, 0.0272, 0.0316, 0.0373, 0.0334, 0.0298, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:30:13,391 INFO [train.py:904] (3/8) Epoch 6, batch 600, loss[loss=0.1846, simple_loss=0.2513, pruned_loss=0.05898, over 16840.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2856, pruned_loss=0.06788, over 3156170.56 frames. ], batch size: 102, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:30:38,608 INFO [optim.py:368] (3/8) 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 10:30:55,734 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:10,958 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:23,783 INFO [train.py:904] (3/8) Epoch 6, batch 650, loss[loss=0.2275, simple_loss=0.2984, pruned_loss=0.07832, over 15611.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2846, pruned_loss=0.0677, over 3191002.45 frames. ], batch size: 191, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:31:39,709 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1452, 2.2548, 2.3501, 4.7087, 2.0777, 3.2040, 2.4289, 2.4255], device='cuda:3'), covar=tensor([0.0486, 0.2207, 0.1214, 0.0234, 0.2880, 0.1242, 0.1921, 0.2606], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0327, 0.0273, 0.0318, 0.0375, 0.0338, 0.0300, 0.0397], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:31:47,318 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:14,543 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:21,262 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0024, 1.7427, 2.4272, 2.8982, 2.7515, 3.0428, 1.7641, 3.1453], device='cuda:3'), covar=tensor([0.0085, 0.0232, 0.0145, 0.0133, 0.0105, 0.0097, 0.0234, 0.0056], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0145, 0.0131, 0.0128, 0.0131, 0.0095, 0.0144, 0.0080], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 10:32:21,272 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:21,394 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.52 vs. limit=5.0 2023-04-28 10:32:31,565 INFO [train.py:904] (3/8) Epoch 6, batch 700, loss[loss=0.2524, simple_loss=0.3145, pruned_loss=0.09519, over 16902.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2848, pruned_loss=0.06799, over 3216334.96 frames. ], batch size: 90, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:32:35,224 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:49,281 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:32:57,145 INFO [optim.py:368] (3/8) 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 10:33:06,286 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6802, 3.9589, 1.8572, 3.9819, 2.7016, 3.9814, 2.0696, 2.8210], device='cuda:3'), covar=tensor([0.0144, 0.0239, 0.1634, 0.0119, 0.0705, 0.0435, 0.1364, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0158, 0.0177, 0.0082, 0.0160, 0.0185, 0.0183, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 10:33:19,514 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:32,380 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:34,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3554, 2.2101, 1.7072, 1.9998, 2.7448, 2.5993, 2.8207, 2.8166], device='cuda:3'), covar=tensor([0.0078, 0.0169, 0.0234, 0.0217, 0.0083, 0.0134, 0.0095, 0.0113], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0164, 0.0161, 0.0157, 0.0156, 0.0163, 0.0142, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:33:39,717 INFO [train.py:904] (3/8) Epoch 6, batch 750, loss[loss=0.2002, simple_loss=0.2762, pruned_loss=0.06209, over 16613.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2849, pruned_loss=0.06738, over 3242523.98 frames. ], batch size: 68, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:33:48,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1245, 3.6877, 3.0221, 1.8796, 2.6491, 2.1813, 3.5118, 3.4508], device='cuda:3'), covar=tensor([0.0255, 0.0506, 0.0635, 0.1590, 0.0719, 0.0979, 0.0560, 0.0793], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0153, 0.0141, 0.0130, 0.0125, 0.0138, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 10:33:56,283 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:34:19,277 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:34:28,769 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-28 10:34:53,035 INFO [train.py:904] (3/8) Epoch 6, batch 800, loss[loss=0.2228, simple_loss=0.2889, pruned_loss=0.07834, over 16135.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2843, pruned_loss=0.06716, over 3254321.92 frames. ], batch size: 164, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:34:59,900 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:36:01,901 INFO [train.py:904] (3/8) Epoch 6, batch 850, loss[loss=0.1836, simple_loss=0.2665, pruned_loss=0.05041, over 16795.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2836, pruned_loss=0.06673, over 3252694.35 frames. ], batch size: 39, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:36:04,119 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:37:10,631 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 900, loss[loss=0.2196, simple_loss=0.2878, pruned_loss=0.0757, over 16835.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2828, pruned_loss=0.06606, over 3268549.50 frames. ], batch size: 90, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:37:39,517 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 950, loss[loss=0.1808, simple_loss=0.263, pruned_loss=0.04927, over 17038.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2832, pruned_loss=0.06655, over 3283099.31 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:38:46,320 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:12,961 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:25,681 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:30,782 INFO [train.py:904] (3/8) Epoch 6, batch 1000, loss[loss=0.1977, simple_loss=0.2695, pruned_loss=0.06291, over 16505.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2805, pruned_loss=0.0655, over 3283652.92 frames. ], batch size: 75, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:39:51,033 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 6, batch 1050, loss[loss=0.1956, simple_loss=0.2892, pruned_loss=0.05103, over 17141.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.281, pruned_loss=0.06599, over 3289618.26 frames. ], batch size: 48, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:40:49,348 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:41:17,917 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:41:23,401 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3260, 4.1524, 4.3184, 4.5376, 4.6345, 4.1324, 4.4328, 4.5973], device='cuda:3'), covar=tensor([0.0946, 0.0745, 0.1150, 0.0505, 0.0417, 0.0929, 0.1190, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0436, 0.0537, 0.0690, 0.0553, 0.0417, 0.0405, 0.0433, 0.0466], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:41:25,726 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-04-28 10:41:49,452 INFO [train.py:904] (3/8) Epoch 6, batch 1100, loss[loss=0.2121, simple_loss=0.2744, pruned_loss=0.07491, over 16322.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2809, pruned_loss=0.06548, over 3298935.21 frames. ], batch size: 145, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:41:49,769 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:42:16,638 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:42:59,412 INFO [train.py:904] (3/8) Epoch 6, batch 1150, loss[loss=0.2335, simple_loss=0.2882, pruned_loss=0.08941, over 16748.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2795, pruned_loss=0.0643, over 3306458.37 frames. ], batch size: 124, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:44:08,004 INFO [train.py:904] (3/8) Epoch 6, batch 1200, loss[loss=0.2145, simple_loss=0.2825, pruned_loss=0.07325, over 16356.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2789, pruned_loss=0.06375, over 3308017.15 frames. ], batch size: 146, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:44:33,628 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.935e+02 3.395e+02 4.046e+02 1.078e+03, threshold=6.791e+02, percent-clipped=4.0 2023-04-28 10:45:20,380 INFO [train.py:904] (3/8) Epoch 6, batch 1250, loss[loss=0.2322, simple_loss=0.3016, pruned_loss=0.08141, over 16476.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2799, pruned_loss=0.06485, over 3318900.62 frames. ], batch size: 146, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:46:13,389 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:15,824 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:26,699 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:30,470 INFO [train.py:904] (3/8) Epoch 6, batch 1300, loss[loss=0.2127, simple_loss=0.2789, pruned_loss=0.07323, over 16454.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2798, pruned_loss=0.06421, over 3329028.23 frames. ], batch size: 146, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:46:50,347 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6522, 3.5868, 2.9215, 5.1214, 4.6325, 4.6002, 1.6164, 3.6293], device='cuda:3'), covar=tensor([0.1375, 0.0479, 0.1020, 0.0093, 0.0235, 0.0333, 0.1417, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0144, 0.0169, 0.0093, 0.0185, 0.0187, 0.0162, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 10:46:58,326 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.918e+02 3.490e+02 4.202e+02 7.834e+02, threshold=6.979e+02, percent-clipped=4.0 2023-04-28 10:47:20,305 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:27,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0290, 4.0129, 3.1432, 2.3962, 2.9658, 2.4089, 4.2977, 3.9097], device='cuda:3'), covar=tensor([0.1784, 0.0542, 0.1087, 0.1501, 0.1956, 0.1333, 0.0310, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0255, 0.0273, 0.0250, 0.0285, 0.0206, 0.0251, 0.0271], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:47:34,137 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 1350, loss[loss=0.2175, simple_loss=0.3101, pruned_loss=0.06241, over 17045.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.28, pruned_loss=0.0638, over 3328901.09 frames. ], batch size: 55, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:47:42,032 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:44,101 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:51,203 INFO [train.py:904] (3/8) Epoch 6, batch 1400, loss[loss=0.2195, simple_loss=0.2812, pruned_loss=0.07892, over 15431.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2798, pruned_loss=0.06417, over 3323630.25 frames. ], batch size: 191, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:48:51,477 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:49:31,133 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8986, 3.7830, 3.1472, 5.3500, 4.7912, 4.9414, 1.7396, 3.5462], device='cuda:3'), covar=tensor([0.1255, 0.0452, 0.0892, 0.0082, 0.0247, 0.0254, 0.1302, 0.0602], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0145, 0.0169, 0.0094, 0.0187, 0.0189, 0.0163, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 10:49:49,869 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:50:00,573 INFO [train.py:904] (3/8) Epoch 6, batch 1450, loss[loss=0.1778, simple_loss=0.273, pruned_loss=0.04134, over 17024.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2784, pruned_loss=0.06378, over 3315184.70 frames. ], batch size: 50, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:51:10,829 INFO [train.py:904] (3/8) Epoch 6, batch 1500, loss[loss=0.2128, simple_loss=0.3011, pruned_loss=0.06224, over 17095.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2792, pruned_loss=0.065, over 3318979.53 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:51:15,498 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:51:38,533 INFO [optim.py:368] (3/8) 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 10:51:50,899 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0650, 5.4792, 5.6262, 5.5185, 5.4440, 5.9989, 5.7440, 5.5209], device='cuda:3'), covar=tensor([0.0680, 0.1476, 0.1257, 0.1719, 0.2443, 0.0926, 0.0961, 0.1979], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0418, 0.0415, 0.0360, 0.0482, 0.0449, 0.0339, 0.0481], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 10:52:11,008 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 10:52:18,640 INFO [train.py:904] (3/8) Epoch 6, batch 1550, loss[loss=0.1794, simple_loss=0.2535, pruned_loss=0.05266, over 15778.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2811, pruned_loss=0.06616, over 3313524.89 frames. ], batch size: 35, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:52:33,862 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 1600, loss[loss=0.1997, simple_loss=0.289, pruned_loss=0.05517, over 17075.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.283, pruned_loss=0.06674, over 3312026.19 frames. ], batch size: 50, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:53:55,823 INFO [optim.py:368] (3/8) 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 10:53:59,305 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:18,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9133, 1.6278, 2.3946, 2.8235, 2.6961, 3.1107, 1.5830, 3.1570], device='cuda:3'), covar=tensor([0.0081, 0.0263, 0.0155, 0.0140, 0.0121, 0.0116, 0.0267, 0.0064], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0151, 0.0135, 0.0134, 0.0138, 0.0100, 0.0145, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 10:54:30,440 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:37,117 INFO [train.py:904] (3/8) Epoch 6, batch 1650, loss[loss=0.2227, simple_loss=0.2886, pruned_loss=0.07843, over 16427.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2845, pruned_loss=0.06728, over 3304775.55 frames. ], batch size: 146, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:54:40,194 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:43,277 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8225, 4.1417, 1.9707, 4.6188, 2.7132, 4.6060, 2.2627, 3.2511], device='cuda:3'), covar=tensor([0.0177, 0.0322, 0.1513, 0.0051, 0.0813, 0.0284, 0.1319, 0.0515], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0163, 0.0180, 0.0088, 0.0162, 0.0194, 0.0188, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 10:54:58,185 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5973, 4.3606, 4.0636, 2.1721, 3.2045, 2.7145, 3.9328, 4.1533], device='cuda:3'), covar=tensor([0.0241, 0.0427, 0.0403, 0.1419, 0.0648, 0.0849, 0.0621, 0.0818], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0133, 0.0152, 0.0139, 0.0130, 0.0125, 0.0139, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 10:55:31,411 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5772, 2.3433, 1.6901, 2.1973, 2.8610, 2.6635, 2.9541, 2.9867], device='cuda:3'), covar=tensor([0.0073, 0.0179, 0.0258, 0.0203, 0.0104, 0.0139, 0.0108, 0.0093], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0167, 0.0164, 0.0161, 0.0162, 0.0167, 0.0153, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:55:37,173 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 10:55:39,487 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8608, 1.7102, 1.4420, 1.4770, 1.8821, 1.6433, 1.6970, 1.9243], device='cuda:3'), covar=tensor([0.0060, 0.0132, 0.0189, 0.0166, 0.0093, 0.0137, 0.0099, 0.0097], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0167, 0.0165, 0.0162, 0.0163, 0.0168, 0.0153, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 10:55:45,565 INFO [train.py:904] (3/8) Epoch 6, batch 1700, loss[loss=0.1939, simple_loss=0.2758, pruned_loss=0.05604, over 17041.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2864, pruned_loss=0.06727, over 3315048.00 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:55:45,911 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:56:14,195 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.939e+02 3.544e+02 4.398e+02 9.866e+02, threshold=7.088e+02, percent-clipped=4.0 2023-04-28 10:56:57,692 INFO [train.py:904] (3/8) Epoch 6, batch 1750, loss[loss=0.2689, simple_loss=0.3352, pruned_loss=0.1013, over 15498.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.288, pruned_loss=0.0673, over 3311446.08 frames. ], batch size: 190, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:05,486 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 1800, loss[loss=0.2259, simple_loss=0.305, pruned_loss=0.0734, over 16413.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2886, pruned_loss=0.06748, over 3309612.12 frames. ], batch size: 146, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:14,101 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:58:36,301 INFO [optim.py:368] (3/8) 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 10:59:08,791 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7969, 3.4185, 2.7194, 4.7265, 4.1366, 4.3559, 1.5350, 3.1544], device='cuda:3'), covar=tensor([0.1319, 0.0456, 0.1003, 0.0080, 0.0284, 0.0367, 0.1317, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0144, 0.0167, 0.0095, 0.0186, 0.0186, 0.0161, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 10:59:17,733 INFO [train.py:904] (3/8) Epoch 6, batch 1850, loss[loss=0.199, simple_loss=0.2919, pruned_loss=0.05306, over 17213.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2891, pruned_loss=0.06744, over 3313667.56 frames. ], batch size: 45, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 10:59:38,331 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:00:27,089 INFO [train.py:904] (3/8) Epoch 6, batch 1900, loss[loss=0.2051, simple_loss=0.2858, pruned_loss=0.06219, over 17015.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2879, pruned_loss=0.06582, over 3318376.36 frames. ], batch size: 55, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:00:51,239 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:00:54,522 INFO [optim.py:368] (3/8) 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 11:01:07,557 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6130, 3.8091, 1.9094, 3.9091, 2.6851, 3.8975, 1.9740, 2.8294], device='cuda:3'), covar=tensor([0.0111, 0.0238, 0.1334, 0.0082, 0.0620, 0.0411, 0.1153, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0161, 0.0177, 0.0085, 0.0160, 0.0193, 0.0185, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 11:01:09,337 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3945, 1.3662, 2.0458, 2.2773, 2.3302, 2.5702, 1.4992, 2.5290], device='cuda:3'), covar=tensor([0.0091, 0.0231, 0.0170, 0.0122, 0.0118, 0.0084, 0.0216, 0.0041], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0149, 0.0133, 0.0133, 0.0138, 0.0099, 0.0143, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 11:01:30,546 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 1950, loss[loss=0.2569, simple_loss=0.3201, pruned_loss=0.09689, over 16843.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2875, pruned_loss=0.06496, over 3325041.18 frames. ], batch size: 116, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:01:47,355 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7922, 4.5322, 4.7629, 4.9665, 5.1201, 4.4997, 5.0200, 5.0708], device='cuda:3'), covar=tensor([0.0943, 0.0754, 0.1223, 0.0532, 0.0377, 0.0685, 0.0530, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0450, 0.0553, 0.0705, 0.0570, 0.0425, 0.0418, 0.0443, 0.0476], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:02:26,100 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8502, 3.5019, 3.3302, 5.1946, 4.6940, 4.7842, 1.8178, 3.6097], device='cuda:3'), covar=tensor([0.1257, 0.0501, 0.0822, 0.0069, 0.0202, 0.0245, 0.1255, 0.0537], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0144, 0.0167, 0.0095, 0.0187, 0.0186, 0.0162, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 11:02:36,908 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 2000, loss[loss=0.1973, simple_loss=0.2847, pruned_loss=0.05494, over 17103.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.287, pruned_loss=0.06467, over 3312742.43 frames. ], batch size: 47, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:02:49,251 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 11:03:20,336 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7187, 3.9345, 4.2578, 3.1044, 3.8237, 4.0435, 3.9195, 2.3720], device='cuda:3'), covar=tensor([0.0278, 0.0036, 0.0025, 0.0197, 0.0041, 0.0045, 0.0037, 0.0273], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0062, 0.0060, 0.0112, 0.0062, 0.0071, 0.0064, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 11:03:49,533 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9383, 3.8332, 3.8005, 3.0986, 3.8502, 1.6803, 3.5752, 3.4384], device='cuda:3'), covar=tensor([0.0105, 0.0079, 0.0134, 0.0375, 0.0081, 0.2197, 0.0122, 0.0203], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0091, 0.0141, 0.0137, 0.0106, 0.0151, 0.0121, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:03:57,170 INFO [train.py:904] (3/8) Epoch 6, batch 2050, loss[loss=0.2031, simple_loss=0.2862, pruned_loss=0.06003, over 17174.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2863, pruned_loss=0.065, over 3312219.76 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:04:14,416 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:04:26,441 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-04-28 11:05:06,133 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:05:08,351 INFO [train.py:904] (3/8) Epoch 6, batch 2100, loss[loss=0.2018, simple_loss=0.2932, pruned_loss=0.0552, over 17108.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2864, pruned_loss=0.06522, over 3313762.48 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:05:24,878 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 11:05:36,281 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.933e+02 3.490e+02 4.297e+02 7.985e+02, threshold=6.979e+02, percent-clipped=3.0 2023-04-28 11:05:59,841 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5392, 4.5379, 4.4504, 3.7565, 4.4547, 1.6428, 4.2382, 4.2990], device='cuda:3'), covar=tensor([0.0095, 0.0065, 0.0121, 0.0350, 0.0084, 0.2010, 0.0103, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0091, 0.0141, 0.0138, 0.0106, 0.0152, 0.0122, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:06:12,955 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 2150, loss[loss=0.2032, simple_loss=0.2872, pruned_loss=0.05956, over 17121.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2879, pruned_loss=0.06622, over 3310782.40 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:06:33,116 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:07:29,309 INFO [train.py:904] (3/8) Epoch 6, batch 2200, loss[loss=0.2446, simple_loss=0.3281, pruned_loss=0.08061, over 16731.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2889, pruned_loss=0.06712, over 3318232.70 frames. ], batch size: 57, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:07:42,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3739, 4.1755, 4.1929, 4.5388, 4.6607, 4.2074, 4.4729, 4.6313], device='cuda:3'), covar=tensor([0.0958, 0.0948, 0.1652, 0.0690, 0.0559, 0.0937, 0.1084, 0.0611], device='cuda:3'), in_proj_covar=tensor([0.0437, 0.0538, 0.0688, 0.0552, 0.0414, 0.0410, 0.0428, 0.0463], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:07:52,332 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:07:52,464 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6445, 3.5485, 2.8077, 2.1392, 2.6238, 2.2296, 3.7174, 3.5045], device='cuda:3'), covar=tensor([0.2029, 0.0660, 0.1149, 0.1717, 0.2026, 0.1506, 0.0420, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0256, 0.0271, 0.0250, 0.0288, 0.0204, 0.0247, 0.0270], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:07:56,705 INFO [optim.py:368] (3/8) 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 11:08:02,894 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 11:08:38,582 INFO [train.py:904] (3/8) Epoch 6, batch 2250, loss[loss=0.2134, simple_loss=0.2812, pruned_loss=0.07276, over 16801.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2899, pruned_loss=0.06848, over 3317539.12 frames. ], batch size: 102, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:08:59,895 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:09:03,504 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0695, 4.4419, 4.7593, 3.5691, 4.1915, 4.5934, 4.0622, 3.1363], device='cuda:3'), covar=tensor([0.0290, 0.0028, 0.0017, 0.0167, 0.0033, 0.0033, 0.0036, 0.0239], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0063, 0.0061, 0.0115, 0.0063, 0.0072, 0.0066, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 11:09:20,133 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5570, 1.5126, 2.0541, 2.5052, 2.5299, 2.4859, 1.6988, 2.6726], device='cuda:3'), covar=tensor([0.0066, 0.0207, 0.0153, 0.0109, 0.0095, 0.0109, 0.0193, 0.0045], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0149, 0.0132, 0.0135, 0.0139, 0.0100, 0.0143, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 11:09:47,159 INFO [train.py:904] (3/8) Epoch 6, batch 2300, loss[loss=0.22, simple_loss=0.2864, pruned_loss=0.07674, over 16684.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2899, pruned_loss=0.06832, over 3321679.65 frames. ], batch size: 134, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:09:54,597 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 6, batch 2350, loss[loss=0.2305, simple_loss=0.3142, pruned_loss=0.0734, over 17029.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2911, pruned_loss=0.06888, over 3312930.62 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:11:08,083 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:20,220 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:12:08,722 INFO [train.py:904] (3/8) Epoch 6, batch 2400, loss[loss=0.2028, simple_loss=0.2919, pruned_loss=0.05687, over 17114.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2919, pruned_loss=0.06886, over 3306155.18 frames. ], batch size: 47, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:12:36,944 INFO [optim.py:368] (3/8) 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 11:12:57,654 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 11:13:19,222 INFO [train.py:904] (3/8) Epoch 6, batch 2450, loss[loss=0.2118, simple_loss=0.2992, pruned_loss=0.06222, over 17019.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2918, pruned_loss=0.06819, over 3312113.90 frames. ], batch size: 55, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:13:33,702 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:13:56,586 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 11:14:16,031 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:14:22,821 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0048, 2.1416, 2.2728, 4.6727, 1.9135, 3.1606, 2.2495, 2.3700], device='cuda:3'), covar=tensor([0.0547, 0.2575, 0.1304, 0.0249, 0.3252, 0.1241, 0.2179, 0.2682], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0333, 0.0274, 0.0315, 0.0373, 0.0349, 0.0302, 0.0402], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:14:24,909 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 11:14:29,168 INFO [train.py:904] (3/8) Epoch 6, batch 2500, loss[loss=0.1872, simple_loss=0.2702, pruned_loss=0.05206, over 17192.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2911, pruned_loss=0.06739, over 3311388.40 frames. ], batch size: 46, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:14:39,972 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:14:57,074 INFO [optim.py:368] (3/8) 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 11:15:35,658 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-28 11:15:38,468 INFO [train.py:904] (3/8) Epoch 6, batch 2550, loss[loss=0.2079, simple_loss=0.2863, pruned_loss=0.06474, over 16289.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2908, pruned_loss=0.06688, over 3319391.76 frames. ], batch size: 36, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:15:40,070 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:43,235 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:58,280 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:16:29,966 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2722, 3.8792, 2.8919, 5.3942, 4.9505, 4.7908, 2.0557, 3.3675], device='cuda:3'), covar=tensor([0.1075, 0.0396, 0.0968, 0.0076, 0.0309, 0.0310, 0.1157, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0144, 0.0168, 0.0097, 0.0193, 0.0188, 0.0161, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 11:16:48,720 INFO [train.py:904] (3/8) Epoch 6, batch 2600, loss[loss=0.2124, simple_loss=0.2934, pruned_loss=0.0657, over 16544.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2907, pruned_loss=0.06572, over 3317654.80 frames. ], batch size: 75, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:17:09,298 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:11,667 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:16,413 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.796e+02 3.424e+02 4.100e+02 9.306e+02, threshold=6.847e+02, percent-clipped=3.0 2023-04-28 11:17:24,099 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:57,444 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 11:17:59,456 INFO [train.py:904] (3/8) Epoch 6, batch 2650, loss[loss=0.2076, simple_loss=0.2987, pruned_loss=0.05822, over 17055.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.291, pruned_loss=0.06541, over 3321958.79 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:18:10,327 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:14,690 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:37,179 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 11:19:09,533 INFO [train.py:904] (3/8) Epoch 6, batch 2700, loss[loss=0.2084, simple_loss=0.284, pruned_loss=0.06636, over 16698.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2913, pruned_loss=0.06489, over 3329498.00 frames. ], batch size: 89, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:19:16,987 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.946e+02 3.581e+02 4.267e+02 8.371e+02, threshold=7.163e+02, percent-clipped=3.0 2023-04-28 11:20:13,227 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6169, 3.9243, 4.2071, 2.8788, 3.7168, 4.1383, 3.7506, 2.5624], device='cuda:3'), covar=tensor([0.0287, 0.0040, 0.0021, 0.0220, 0.0041, 0.0038, 0.0034, 0.0251], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0060, 0.0061, 0.0111, 0.0062, 0.0069, 0.0063, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 11:20:13,259 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8811, 3.9080, 2.0572, 4.3608, 2.5039, 4.2747, 2.1865, 2.9423], device='cuda:3'), covar=tensor([0.0141, 0.0295, 0.1479, 0.0053, 0.0866, 0.0251, 0.1344, 0.0603], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0161, 0.0175, 0.0086, 0.0160, 0.0193, 0.0184, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 11:20:19,909 INFO [train.py:904] (3/8) Epoch 6, batch 2750, loss[loss=0.2453, simple_loss=0.3116, pruned_loss=0.08946, over 12751.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2912, pruned_loss=0.06499, over 3326040.11 frames. ], batch size: 246, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:20:51,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9388, 4.1554, 4.3816, 3.1902, 3.8246, 4.2516, 3.8516, 2.7554], device='cuda:3'), covar=tensor([0.0241, 0.0026, 0.0020, 0.0193, 0.0045, 0.0042, 0.0038, 0.0238], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0062, 0.0062, 0.0114, 0.0063, 0.0071, 0.0065, 0.0109], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 11:21:22,474 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-28 11:21:31,891 INFO [train.py:904] (3/8) Epoch 6, batch 2800, loss[loss=0.2199, simple_loss=0.2859, pruned_loss=0.07696, over 16383.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2913, pruned_loss=0.065, over 3324971.38 frames. ], batch size: 146, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:22:01,718 INFO [optim.py:368] (3/8) 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 11:22:37,900 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:22:43,359 INFO [train.py:904] (3/8) Epoch 6, batch 2850, loss[loss=0.187, simple_loss=0.2595, pruned_loss=0.05732, over 15849.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2892, pruned_loss=0.06409, over 3331004.02 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:22:54,779 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 11:23:25,480 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0378, 2.0878, 2.2046, 4.6579, 1.9095, 3.1114, 2.3056, 2.3236], device='cuda:3'), covar=tensor([0.0531, 0.2514, 0.1340, 0.0251, 0.3230, 0.1385, 0.2063, 0.2788], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0334, 0.0274, 0.0315, 0.0374, 0.0353, 0.0302, 0.0401], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:23:51,285 INFO [train.py:904] (3/8) Epoch 6, batch 2900, loss[loss=0.2063, simple_loss=0.2753, pruned_loss=0.06862, over 15848.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2872, pruned_loss=0.06368, over 3332923.50 frames. ], batch size: 35, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:24:04,492 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:19,307 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 6, batch 2950, loss[loss=0.2327, simple_loss=0.2952, pruned_loss=0.08507, over 16843.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2876, pruned_loss=0.06519, over 3324472.84 frames. ], batch size: 96, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:25:15,493 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:25:30,125 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:26:08,565 INFO [train.py:904] (3/8) Epoch 6, batch 3000, loss[loss=0.183, simple_loss=0.2695, pruned_loss=0.0482, over 16862.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2875, pruned_loss=0.06571, over 3330832.86 frames. ], batch size: 42, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:26:08,566 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 11:26:17,402 INFO [train.py:938] (3/8) Epoch 6, validation: loss=0.1514, simple_loss=0.258, pruned_loss=0.02246, over 944034.00 frames. 2023-04-28 11:26:17,403 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 11:26:20,273 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6506, 2.5605, 2.1348, 2.3991, 2.9229, 2.8715, 3.6953, 3.2612], device='cuda:3'), covar=tensor([0.0032, 0.0190, 0.0225, 0.0207, 0.0128, 0.0162, 0.0095, 0.0106], device='cuda:3'), in_proj_covar=tensor([0.0095, 0.0164, 0.0163, 0.0160, 0.0162, 0.0167, 0.0155, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:26:29,470 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:26:45,969 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.839e+02 3.365e+02 4.010e+02 8.699e+02, threshold=6.731e+02, percent-clipped=2.0 2023-04-28 11:27:25,829 INFO [train.py:904] (3/8) Epoch 6, batch 3050, loss[loss=0.2193, simple_loss=0.2879, pruned_loss=0.07535, over 16804.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2876, pruned_loss=0.06595, over 3315191.37 frames. ], batch size: 124, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:28:35,659 INFO [train.py:904] (3/8) Epoch 6, batch 3100, loss[loss=0.2201, simple_loss=0.2853, pruned_loss=0.07742, over 16929.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2865, pruned_loss=0.06564, over 3325640.05 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:29:05,910 INFO [optim.py:368] (3/8) 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 11:29:40,274 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:29:45,138 INFO [train.py:904] (3/8) Epoch 6, batch 3150, loss[loss=0.2334, simple_loss=0.2975, pruned_loss=0.0847, over 15689.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2858, pruned_loss=0.06518, over 3335183.05 frames. ], batch size: 191, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:29:45,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7617, 3.9051, 4.0922, 3.2598, 3.6328, 4.0145, 3.7732, 2.5365], device='cuda:3'), covar=tensor([0.0248, 0.0037, 0.0030, 0.0165, 0.0051, 0.0053, 0.0044, 0.0255], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0061, 0.0062, 0.0112, 0.0063, 0.0070, 0.0064, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 11:30:49,056 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:30:56,670 INFO [train.py:904] (3/8) Epoch 6, batch 3200, loss[loss=0.2158, simple_loss=0.2843, pruned_loss=0.07358, over 16932.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2855, pruned_loss=0.06504, over 3327883.95 frames. ], batch size: 109, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:31:09,519 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 11:32:09,686 INFO [train.py:904] (3/8) Epoch 6, batch 3250, loss[loss=0.2359, simple_loss=0.2991, pruned_loss=0.08636, over 16444.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2859, pruned_loss=0.06529, over 3329476.95 frames. ], batch size: 146, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:32:19,694 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:35,405 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:40,074 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:33:17,063 INFO [train.py:904] (3/8) Epoch 6, batch 3300, loss[loss=0.2136, simple_loss=0.2845, pruned_loss=0.07133, over 16798.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2875, pruned_loss=0.06643, over 3327805.73 frames. ], batch size: 124, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:33:24,651 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7433, 6.1129, 5.8927, 5.9476, 5.2502, 5.0758, 5.7115, 6.2446], device='cuda:3'), covar=tensor([0.0783, 0.0665, 0.0759, 0.0541, 0.0741, 0.0534, 0.0623, 0.0700], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0571, 0.0473, 0.0362, 0.0352, 0.0354, 0.0456, 0.0399], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:33:45,159 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 11:34:09,901 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9309, 4.6276, 4.8409, 5.1095, 5.3007, 4.6091, 5.2492, 5.2923], device='cuda:3'), covar=tensor([0.1213, 0.0834, 0.1606, 0.0589, 0.0442, 0.0680, 0.0503, 0.0397], device='cuda:3'), in_proj_covar=tensor([0.0448, 0.0537, 0.0698, 0.0562, 0.0423, 0.0417, 0.0436, 0.0472], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:34:26,275 INFO [train.py:904] (3/8) Epoch 6, batch 3350, loss[loss=0.2111, simple_loss=0.2811, pruned_loss=0.0706, over 16424.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2875, pruned_loss=0.06547, over 3332647.83 frames. ], batch size: 146, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:34:35,717 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-28 11:35:17,736 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7009, 1.2084, 1.5629, 1.7079, 1.8879, 1.9389, 1.4114, 1.6878], device='cuda:3'), covar=tensor([0.0104, 0.0179, 0.0112, 0.0123, 0.0098, 0.0065, 0.0176, 0.0046], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0152, 0.0137, 0.0136, 0.0141, 0.0103, 0.0144, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 11:35:34,480 INFO [train.py:904] (3/8) Epoch 6, batch 3400, loss[loss=0.2179, simple_loss=0.2871, pruned_loss=0.07439, over 16893.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2883, pruned_loss=0.06558, over 3324349.64 frames. ], batch size: 109, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:36:05,329 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 3450, loss[loss=0.1864, simple_loss=0.2781, pruned_loss=0.0474, over 17037.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2864, pruned_loss=0.06484, over 3326017.29 frames. ], batch size: 50, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:37:30,159 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1472, 4.9092, 5.0845, 5.3683, 5.5547, 4.8006, 5.4899, 5.4146], device='cuda:3'), covar=tensor([0.0961, 0.0644, 0.1268, 0.0464, 0.0337, 0.0491, 0.0343, 0.0380], device='cuda:3'), in_proj_covar=tensor([0.0449, 0.0544, 0.0708, 0.0569, 0.0427, 0.0420, 0.0437, 0.0477], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:37:58,613 INFO [train.py:904] (3/8) Epoch 6, batch 3500, loss[loss=0.1602, simple_loss=0.2448, pruned_loss=0.03779, over 16987.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2849, pruned_loss=0.06452, over 3320188.19 frames. ], batch size: 41, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:38:28,904 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.700e+02 3.164e+02 3.995e+02 8.205e+02, threshold=6.329e+02, percent-clipped=2.0 2023-04-28 11:38:56,007 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-28 11:39:10,634 INFO [train.py:904] (3/8) Epoch 6, batch 3550, loss[loss=0.1902, simple_loss=0.2834, pruned_loss=0.04848, over 17086.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2846, pruned_loss=0.06427, over 3318136.48 frames. ], batch size: 53, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:40:21,755 INFO [train.py:904] (3/8) Epoch 6, batch 3600, loss[loss=0.1664, simple_loss=0.2442, pruned_loss=0.04432, over 16771.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.283, pruned_loss=0.06387, over 3324104.87 frames. ], batch size: 39, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:40:51,130 INFO [optim.py:368] (3/8) 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 11:41:15,925 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 11:41:33,847 INFO [train.py:904] (3/8) Epoch 6, batch 3650, loss[loss=0.1937, simple_loss=0.2686, pruned_loss=0.05941, over 16380.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2817, pruned_loss=0.06384, over 3314546.16 frames. ], batch size: 68, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:42:46,526 INFO [train.py:904] (3/8) Epoch 6, batch 3700, loss[loss=0.206, simple_loss=0.2701, pruned_loss=0.07091, over 16747.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.281, pruned_loss=0.06605, over 3282932.03 frames. ], batch size: 83, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:43:17,520 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 3750, loss[loss=0.214, simple_loss=0.2816, pruned_loss=0.0732, over 16857.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2822, pruned_loss=0.06771, over 3276720.03 frames. ], batch size: 102, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:13,202 INFO [train.py:904] (3/8) Epoch 6, batch 3800, loss[loss=0.2039, simple_loss=0.273, pruned_loss=0.06738, over 16510.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2834, pruned_loss=0.06926, over 3271555.56 frames. ], batch size: 68, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:46,019 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.146e+02 2.814e+02 3.370e+02 4.517e+02 7.386e+02, threshold=6.740e+02, percent-clipped=3.0 2023-04-28 11:46:28,642 INFO [train.py:904] (3/8) Epoch 6, batch 3850, loss[loss=0.2494, simple_loss=0.3143, pruned_loss=0.0922, over 12348.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2827, pruned_loss=0.0693, over 3270246.01 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:40,667 INFO [train.py:904] (3/8) Epoch 6, batch 3900, loss[loss=0.2086, simple_loss=0.2868, pruned_loss=0.0652, over 17120.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.282, pruned_loss=0.06959, over 3267005.94 frames. ], batch size: 48, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:59,622 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:48:10,862 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 3950, loss[loss=0.2082, simple_loss=0.2743, pruned_loss=0.07099, over 16669.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2804, pruned_loss=0.06951, over 3279793.53 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:49:27,765 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:50:04,343 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-28 11:50:04,726 INFO [train.py:904] (3/8) Epoch 6, batch 4000, loss[loss=0.2144, simple_loss=0.2867, pruned_loss=0.07102, over 16326.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2806, pruned_loss=0.06996, over 3279997.39 frames. ], batch size: 165, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:50:17,969 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 11:50:36,654 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 4050, loss[loss=0.1742, simple_loss=0.2531, pruned_loss=0.0477, over 16491.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2808, pruned_loss=0.06916, over 3267940.43 frames. ], batch size: 68, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:28,287 INFO [train.py:904] (3/8) Epoch 6, batch 4100, loss[loss=0.2203, simple_loss=0.3027, pruned_loss=0.06892, over 16642.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2815, pruned_loss=0.0676, over 3263059.45 frames. ], batch size: 76, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:45,902 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:53:01,849 INFO [optim.py:368] (3/8) 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 11:53:45,441 INFO [train.py:904] (3/8) Epoch 6, batch 4150, loss[loss=0.2625, simple_loss=0.3376, pruned_loss=0.09369, over 15435.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2895, pruned_loss=0.07052, over 3255539.79 frames. ], batch size: 190, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:53:54,651 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2041, 3.2689, 1.6824, 3.4162, 2.2531, 3.4089, 1.7609, 2.5000], device='cuda:3'), covar=tensor([0.0159, 0.0327, 0.1616, 0.0062, 0.0860, 0.0364, 0.1536, 0.0707], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0162, 0.0181, 0.0086, 0.0165, 0.0197, 0.0187, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 11:54:19,610 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:55:01,101 INFO [train.py:904] (3/8) Epoch 6, batch 4200, loss[loss=0.2389, simple_loss=0.3248, pruned_loss=0.07654, over 16589.00 frames. ], tot_loss[loss=0.222, simple_loss=0.2975, pruned_loss=0.07321, over 3215302.72 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:55:33,716 INFO [optim.py:368] (3/8) 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 11:56:09,065 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1253, 5.4469, 5.1722, 5.2131, 4.8521, 4.5704, 4.9219, 5.5179], device='cuda:3'), covar=tensor([0.0563, 0.0564, 0.0667, 0.0393, 0.0581, 0.0641, 0.0668, 0.0553], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0516, 0.0431, 0.0330, 0.0318, 0.0331, 0.0421, 0.0365], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:56:15,276 INFO [train.py:904] (3/8) Epoch 6, batch 4250, loss[loss=0.2193, simple_loss=0.3029, pruned_loss=0.0679, over 16278.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3007, pruned_loss=0.07289, over 3209244.58 frames. ], batch size: 165, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:56:44,666 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:56:45,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3969, 3.2605, 3.3948, 3.5299, 3.5775, 3.2480, 3.5191, 3.5882], device='cuda:3'), covar=tensor([0.0712, 0.0678, 0.1017, 0.0488, 0.0430, 0.1834, 0.0743, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0489, 0.0634, 0.0504, 0.0385, 0.0383, 0.0398, 0.0431], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 11:57:29,393 INFO [train.py:904] (3/8) Epoch 6, batch 4300, loss[loss=0.2414, simple_loss=0.3267, pruned_loss=0.07806, over 16245.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3017, pruned_loss=0.07202, over 3189753.89 frames. ], batch size: 165, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:58:01,801 INFO [optim.py:368] (3/8) 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 11:58:35,535 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6122, 1.4315, 2.0694, 2.5680, 2.4240, 2.8586, 1.4129, 2.6301], device='cuda:3'), covar=tensor([0.0094, 0.0282, 0.0149, 0.0131, 0.0120, 0.0080, 0.0270, 0.0060], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0151, 0.0134, 0.0133, 0.0139, 0.0100, 0.0145, 0.0087], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 11:58:43,132 INFO [train.py:904] (3/8) Epoch 6, batch 4350, loss[loss=0.2905, simple_loss=0.3431, pruned_loss=0.1189, over 11870.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3053, pruned_loss=0.07341, over 3174312.55 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:59:53,033 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:59:56,594 INFO [train.py:904] (3/8) Epoch 6, batch 4400, loss[loss=0.2469, simple_loss=0.3372, pruned_loss=0.07832, over 16926.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3075, pruned_loss=0.0743, over 3168903.76 frames. ], batch size: 90, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:00:27,147 INFO [optim.py:368] (3/8) 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 12:00:43,235 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 12:00:58,447 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 12:01:06,335 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 12:01:06,617 INFO [train.py:904] (3/8) Epoch 6, batch 4450, loss[loss=0.233, simple_loss=0.3217, pruned_loss=0.07211, over 16646.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3101, pruned_loss=0.07453, over 3179629.34 frames. ], batch size: 134, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:01:09,378 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 12:01:19,248 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:21,572 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:31,359 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:02:16,913 INFO [train.py:904] (3/8) Epoch 6, batch 4500, loss[loss=0.2471, simple_loss=0.3238, pruned_loss=0.08517, over 16634.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3105, pruned_loss=0.07454, over 3213562.93 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:02:19,492 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 12:02:47,677 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:02:48,405 INFO [optim.py:368] (3/8) 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 12:03:29,867 INFO [train.py:904] (3/8) Epoch 6, batch 4550, loss[loss=0.2763, simple_loss=0.3501, pruned_loss=0.1013, over 17052.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3106, pruned_loss=0.07517, over 3195305.83 frames. ], batch size: 53, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:03:57,487 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 4600, loss[loss=0.1985, simple_loss=0.2866, pruned_loss=0.05517, over 17181.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3107, pruned_loss=0.07477, over 3205930.42 frames. ], batch size: 44, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:05:07,048 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:05:13,323 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.321e+02 2.807e+02 3.689e+02 1.502e+03, threshold=5.614e+02, percent-clipped=3.0 2023-04-28 12:05:52,744 INFO [train.py:904] (3/8) Epoch 6, batch 4650, loss[loss=0.29, simple_loss=0.3481, pruned_loss=0.116, over 11700.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3092, pruned_loss=0.07436, over 3213367.24 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:07:03,015 INFO [train.py:904] (3/8) Epoch 6, batch 4700, loss[loss=0.234, simple_loss=0.3161, pruned_loss=0.07589, over 15252.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3054, pruned_loss=0.07254, over 3217856.52 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:07:34,099 INFO [optim.py:368] (3/8) 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 12:08:12,120 INFO [train.py:904] (3/8) Epoch 6, batch 4750, loss[loss=0.2105, simple_loss=0.2855, pruned_loss=0.06773, over 17009.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3023, pruned_loss=0.0714, over 3210397.03 frames. ], batch size: 50, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:08:17,430 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:08:35,503 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1802, 5.0718, 5.0341, 4.8121, 4.4761, 5.0518, 4.9448, 4.7479], device='cuda:3'), covar=tensor([0.0403, 0.0255, 0.0160, 0.0172, 0.0826, 0.0257, 0.0187, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0203, 0.0222, 0.0197, 0.0254, 0.0222, 0.0159, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:08:36,481 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:09:22,944 INFO [train.py:904] (3/8) Epoch 6, batch 4800, loss[loss=0.1773, simple_loss=0.2697, pruned_loss=0.04242, over 16457.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2993, pruned_loss=0.06936, over 3200789.10 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:09:34,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6631, 3.5417, 3.5755, 3.8641, 3.9467, 3.6281, 3.9213, 3.9329], device='cuda:3'), covar=tensor([0.0902, 0.0805, 0.1425, 0.0603, 0.0568, 0.1597, 0.0650, 0.0572], device='cuda:3'), in_proj_covar=tensor([0.0404, 0.0486, 0.0629, 0.0498, 0.0378, 0.0377, 0.0390, 0.0421], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:09:45,465 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:46,590 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:53,990 INFO [optim.py:368] (3/8) 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 12:10:17,773 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3032, 4.4263, 4.5358, 4.4053, 4.4345, 4.9379, 4.5346, 4.3024], device='cuda:3'), covar=tensor([0.1091, 0.1308, 0.1117, 0.1666, 0.1986, 0.0847, 0.1069, 0.2014], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0391, 0.0389, 0.0342, 0.0454, 0.0418, 0.0325, 0.0464], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 12:10:35,250 INFO [train.py:904] (3/8) Epoch 6, batch 4850, loss[loss=0.2149, simple_loss=0.3013, pruned_loss=0.06429, over 16883.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3011, pruned_loss=0.06919, over 3192269.06 frames. ], batch size: 109, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:11:33,255 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5580, 3.8243, 1.5778, 4.0902, 2.4668, 3.9644, 1.9865, 2.7179], device='cuda:3'), covar=tensor([0.0158, 0.0223, 0.1742, 0.0033, 0.0763, 0.0279, 0.1449, 0.0670], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0155, 0.0177, 0.0082, 0.0161, 0.0186, 0.0185, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 12:11:48,112 INFO [train.py:904] (3/8) Epoch 6, batch 4900, loss[loss=0.1877, simple_loss=0.2762, pruned_loss=0.04965, over 16802.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2999, pruned_loss=0.06741, over 3189509.29 frames. ], batch size: 83, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:12:05,754 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 12:12:19,090 INFO [optim.py:368] (3/8) 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 12:12:54,576 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 12:12:59,378 INFO [train.py:904] (3/8) Epoch 6, batch 4950, loss[loss=0.2496, simple_loss=0.3212, pruned_loss=0.08905, over 11626.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2989, pruned_loss=0.06642, over 3204935.85 frames. ], batch size: 246, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:05,623 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 5000, loss[loss=0.2046, simple_loss=0.2974, pruned_loss=0.05586, over 16552.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.3006, pruned_loss=0.06643, over 3218235.74 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:38,593 INFO [optim.py:368] (3/8) 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 12:14:41,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 12:14:47,887 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7058, 5.2816, 5.3200, 5.1285, 5.1791, 5.7710, 5.2755, 5.0385], device='cuda:3'), covar=tensor([0.0928, 0.1228, 0.1313, 0.1459, 0.2315, 0.0954, 0.1090, 0.1974], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0392, 0.0393, 0.0341, 0.0457, 0.0425, 0.0327, 0.0463], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 12:15:21,195 INFO [train.py:904] (3/8) Epoch 6, batch 5050, loss[loss=0.2464, simple_loss=0.3188, pruned_loss=0.08704, over 12346.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3019, pruned_loss=0.06684, over 3215168.51 frames. ], batch size: 247, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:15:25,467 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:15:33,251 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:28,565 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:31,556 INFO [train.py:904] (3/8) Epoch 6, batch 5100, loss[loss=0.2037, simple_loss=0.2934, pruned_loss=0.05694, over 16834.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2993, pruned_loss=0.06516, over 3229409.90 frames. ], batch size: 83, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:16:32,962 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:54,882 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:17:04,479 INFO [optim.py:368] (3/8) 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 12:17:39,619 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3757, 1.8846, 1.5666, 1.7389, 2.2440, 2.0755, 2.3569, 2.4691], device='cuda:3'), covar=tensor([0.0049, 0.0198, 0.0256, 0.0245, 0.0108, 0.0187, 0.0083, 0.0116], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0160, 0.0161, 0.0158, 0.0156, 0.0166, 0.0143, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:17:45,240 INFO [train.py:904] (3/8) Epoch 6, batch 5150, loss[loss=0.2172, simple_loss=0.3128, pruned_loss=0.06077, over 16756.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2997, pruned_loss=0.06506, over 3201175.70 frames. ], batch size: 89, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:17:57,087 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:05,648 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:10,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9415, 3.7722, 3.7063, 2.2458, 3.2970, 3.5885, 3.4871, 1.9897], device='cuda:3'), covar=tensor([0.0374, 0.0017, 0.0023, 0.0253, 0.0044, 0.0056, 0.0039, 0.0302], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0055, 0.0059, 0.0113, 0.0062, 0.0070, 0.0064, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 12:18:28,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8133, 1.6558, 2.2346, 2.8785, 2.6170, 3.1467, 1.6074, 3.0252], device='cuda:3'), covar=tensor([0.0079, 0.0278, 0.0153, 0.0116, 0.0120, 0.0065, 0.0285, 0.0051], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0150, 0.0129, 0.0132, 0.0135, 0.0097, 0.0143, 0.0086], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 12:18:58,948 INFO [train.py:904] (3/8) Epoch 6, batch 5200, loss[loss=0.2197, simple_loss=0.2959, pruned_loss=0.07169, over 16427.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.299, pruned_loss=0.06515, over 3199841.12 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:19:25,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5678, 4.0998, 3.7576, 2.0468, 3.0953, 2.4060, 3.7996, 4.0386], device='cuda:3'), covar=tensor([0.0248, 0.0498, 0.0552, 0.1594, 0.0703, 0.0960, 0.0686, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0125, 0.0152, 0.0138, 0.0129, 0.0123, 0.0137, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 12:19:30,989 INFO [optim.py:368] (3/8) 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 12:20:15,818 INFO [train.py:904] (3/8) Epoch 6, batch 5250, loss[loss=0.1885, simple_loss=0.2765, pruned_loss=0.05028, over 16889.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2955, pruned_loss=0.06425, over 3209895.75 frames. ], batch size: 96, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:21:28,014 INFO [train.py:904] (3/8) Epoch 6, batch 5300, loss[loss=0.1794, simple_loss=0.2623, pruned_loss=0.04822, over 16405.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2914, pruned_loss=0.06282, over 3216566.50 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:21:41,924 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 12:21:42,892 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5946, 2.9247, 2.3128, 4.3675, 3.0759, 4.0110, 1.5069, 2.7396], device='cuda:3'), covar=tensor([0.1711, 0.0698, 0.1457, 0.0113, 0.0401, 0.0366, 0.1858, 0.1012], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0143, 0.0168, 0.0094, 0.0188, 0.0187, 0.0163, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 12:21:59,801 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.725e+02 2.658e+02 3.030e+02 3.663e+02 6.971e+02, threshold=6.060e+02, percent-clipped=3.0 2023-04-28 12:22:40,507 INFO [train.py:904] (3/8) Epoch 6, batch 5350, loss[loss=0.1931, simple_loss=0.2807, pruned_loss=0.0528, over 16441.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2908, pruned_loss=0.06285, over 3211523.81 frames. ], batch size: 68, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:22:45,576 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:22,067 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5590, 2.6762, 2.5181, 4.1066, 3.1141, 3.9247, 1.2772, 2.9184], device='cuda:3'), covar=tensor([0.1332, 0.0566, 0.1020, 0.0083, 0.0157, 0.0344, 0.1438, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0144, 0.0170, 0.0095, 0.0188, 0.0188, 0.0164, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 12:23:33,719 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1558, 5.1059, 4.9822, 4.3544, 4.9608, 1.8580, 4.8285, 5.0080], device='cuda:3'), covar=tensor([0.0046, 0.0038, 0.0070, 0.0269, 0.0050, 0.1593, 0.0061, 0.0092], device='cuda:3'), in_proj_covar=tensor([0.0097, 0.0086, 0.0132, 0.0132, 0.0099, 0.0146, 0.0114, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:23:44,956 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:52,596 INFO [train.py:904] (3/8) Epoch 6, batch 5400, loss[loss=0.2012, simple_loss=0.2918, pruned_loss=0.0553, over 16858.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2935, pruned_loss=0.06366, over 3209839.23 frames. ], batch size: 96, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:24:26,634 INFO [optim.py:368] (3/8) 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 12:24:40,075 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1926, 2.9952, 2.7320, 1.9871, 2.5964, 2.1411, 2.7190, 2.8979], device='cuda:3'), covar=tensor([0.0208, 0.0368, 0.0396, 0.1179, 0.0550, 0.0759, 0.0462, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0124, 0.0150, 0.0137, 0.0129, 0.0122, 0.0135, 0.0136], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 12:24:49,801 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 12:25:08,990 INFO [train.py:904] (3/8) Epoch 6, batch 5450, loss[loss=0.2718, simple_loss=0.3356, pruned_loss=0.104, over 16438.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2971, pruned_loss=0.06604, over 3177976.37 frames. ], batch size: 146, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:25:14,956 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:17,611 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:27,411 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-28 12:26:27,344 INFO [train.py:904] (3/8) Epoch 6, batch 5500, loss[loss=0.2938, simple_loss=0.3458, pruned_loss=0.1209, over 11792.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3059, pruned_loss=0.0722, over 3148184.85 frames. ], batch size: 246, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:27:01,684 INFO [optim.py:368] (3/8) 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 12:27:28,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 12:27:42,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7705, 1.2609, 1.6464, 1.7050, 1.7649, 1.8719, 1.4344, 1.6910], device='cuda:3'), covar=tensor([0.0099, 0.0166, 0.0094, 0.0123, 0.0108, 0.0065, 0.0165, 0.0055], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0149, 0.0130, 0.0132, 0.0137, 0.0098, 0.0143, 0.0088], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 12:27:46,058 INFO [train.py:904] (3/8) Epoch 6, batch 5550, loss[loss=0.2817, simple_loss=0.3573, pruned_loss=0.103, over 16682.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.314, pruned_loss=0.07846, over 3128127.11 frames. ], batch size: 62, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:29:07,882 INFO [train.py:904] (3/8) Epoch 6, batch 5600, loss[loss=0.2224, simple_loss=0.305, pruned_loss=0.0699, over 16674.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3197, pruned_loss=0.08352, over 3109043.99 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:29:45,253 INFO [optim.py:368] (3/8) 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 12:30:01,227 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1430, 4.2092, 4.3173, 4.2620, 4.2953, 4.7442, 4.4558, 4.2014], device='cuda:3'), covar=tensor([0.1320, 0.1618, 0.1553, 0.1809, 0.2197, 0.1017, 0.1177, 0.2391], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0394, 0.0399, 0.0342, 0.0455, 0.0424, 0.0324, 0.0464], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 12:30:29,968 INFO [train.py:904] (3/8) Epoch 6, batch 5650, loss[loss=0.2489, simple_loss=0.3166, pruned_loss=0.0906, over 16714.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.327, pruned_loss=0.0898, over 3084699.56 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:30:33,048 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0111, 3.1188, 1.5737, 3.3120, 2.2759, 3.2611, 1.8362, 2.5249], device='cuda:3'), covar=tensor([0.0187, 0.0350, 0.1668, 0.0071, 0.0851, 0.0438, 0.1489, 0.0609], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0155, 0.0175, 0.0081, 0.0163, 0.0186, 0.0187, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 12:30:34,905 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:31:02,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2908, 4.2872, 4.1188, 4.0672, 3.7586, 4.2158, 4.0421, 3.9463], device='cuda:3'), covar=tensor([0.0500, 0.0268, 0.0251, 0.0201, 0.0900, 0.0327, 0.0501, 0.0496], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0214, 0.0231, 0.0202, 0.0258, 0.0232, 0.0161, 0.0258], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:31:48,511 INFO [train.py:904] (3/8) Epoch 6, batch 5700, loss[loss=0.2864, simple_loss=0.3574, pruned_loss=0.1077, over 16386.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3296, pruned_loss=0.09222, over 3084754.67 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:31:51,663 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 6, batch 5750, loss[loss=0.2786, simple_loss=0.33, pruned_loss=0.1135, over 10943.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.3319, pruned_loss=0.09369, over 3058077.07 frames. ], batch size: 247, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:33:09,308 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:13,990 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:33,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6710, 3.8425, 1.7820, 4.0628, 2.5838, 4.0512, 2.0123, 2.8969], device='cuda:3'), covar=tensor([0.0137, 0.0208, 0.1689, 0.0045, 0.0747, 0.0330, 0.1537, 0.0574], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0152, 0.0173, 0.0080, 0.0161, 0.0183, 0.0183, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 12:33:44,555 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:34:04,264 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2567, 1.8128, 2.0574, 3.7419, 1.7993, 2.6322, 2.0924, 1.9799], device='cuda:3'), covar=tensor([0.0685, 0.2662, 0.1432, 0.0346, 0.3287, 0.1342, 0.2361, 0.2737], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0336, 0.0277, 0.0310, 0.0381, 0.0348, 0.0300, 0.0396], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:34:30,321 INFO [train.py:904] (3/8) Epoch 6, batch 5800, loss[loss=0.2687, simple_loss=0.3288, pruned_loss=0.1043, over 11906.00 frames. ], tot_loss[loss=0.2583, simple_loss=0.3318, pruned_loss=0.09241, over 3050830.69 frames. ], batch size: 247, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:34:32,754 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:35:05,969 INFO [optim.py:368] (3/8) 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 12:35:10,482 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 12:35:23,448 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:35:49,148 INFO [train.py:904] (3/8) Epoch 6, batch 5850, loss[loss=0.241, simple_loss=0.3212, pruned_loss=0.08041, over 16264.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3285, pruned_loss=0.08948, over 3058195.32 frames. ], batch size: 165, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:36:52,364 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8378, 3.9108, 4.3287, 4.3371, 4.3178, 3.9377, 3.9710, 3.8512], device='cuda:3'), covar=tensor([0.0294, 0.0407, 0.0350, 0.0374, 0.0364, 0.0334, 0.0820, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0250, 0.0257, 0.0255, 0.0307, 0.0276, 0.0376, 0.0229], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 12:37:11,643 INFO [train.py:904] (3/8) Epoch 6, batch 5900, loss[loss=0.2537, simple_loss=0.3347, pruned_loss=0.08642, over 16860.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3271, pruned_loss=0.08843, over 3072704.07 frames. ], batch size: 42, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:37:52,136 INFO [optim.py:368] (3/8) 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 12:38:33,938 INFO [train.py:904] (3/8) Epoch 6, batch 5950, loss[loss=0.2494, simple_loss=0.3274, pruned_loss=0.08568, over 17054.00 frames. ], tot_loss[loss=0.2519, simple_loss=0.3286, pruned_loss=0.08763, over 3075591.24 frames. ], batch size: 55, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:52,035 INFO [train.py:904] (3/8) Epoch 6, batch 6000, loss[loss=0.2541, simple_loss=0.3365, pruned_loss=0.08584, over 16431.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3288, pruned_loss=0.08793, over 3073899.55 frames. ], batch size: 68, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:52,036 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 12:40:01,521 INFO [train.py:938] (3/8) Epoch 6, validation: loss=0.18, simple_loss=0.2922, pruned_loss=0.03386, over 944034.00 frames. 2023-04-28 12:40:01,522 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 12:40:36,510 INFO [optim.py:368] (3/8) 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 12:41:19,685 INFO [train.py:904] (3/8) Epoch 6, batch 6050, loss[loss=0.2373, simple_loss=0.3361, pruned_loss=0.06928, over 16686.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3267, pruned_loss=0.08689, over 3078826.54 frames. ], batch size: 76, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:41:20,602 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:41:23,910 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:01,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8680, 3.7633, 2.7443, 5.1979, 4.1317, 4.6989, 1.9942, 3.2359], device='cuda:3'), covar=tensor([0.1223, 0.0415, 0.1014, 0.0087, 0.0407, 0.0235, 0.1158, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0145, 0.0170, 0.0095, 0.0198, 0.0191, 0.0164, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 12:42:13,294 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0049, 2.6937, 2.6468, 1.7738, 2.8129, 2.8359, 2.5026, 2.3008], device='cuda:3'), covar=tensor([0.0712, 0.0152, 0.0166, 0.0851, 0.0100, 0.0131, 0.0334, 0.0409], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0092, 0.0081, 0.0140, 0.0072, 0.0081, 0.0115, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 12:42:33,595 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 6100, loss[loss=0.2346, simple_loss=0.3223, pruned_loss=0.07345, over 16871.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3255, pruned_loss=0.08565, over 3102499.04 frames. ], batch size: 102, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:42:58,763 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:43:13,025 INFO [optim.py:368] (3/8) 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 12:43:20,732 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:43:35,237 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6296, 4.0097, 4.2453, 1.9153, 4.5729, 4.4858, 3.3057, 3.2929], device='cuda:3'), covar=tensor([0.0833, 0.0114, 0.0121, 0.1242, 0.0037, 0.0071, 0.0294, 0.0391], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0093, 0.0081, 0.0140, 0.0073, 0.0081, 0.0115, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 12:43:56,260 INFO [train.py:904] (3/8) Epoch 6, batch 6150, loss[loss=0.2052, simple_loss=0.2857, pruned_loss=0.06236, over 17044.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3228, pruned_loss=0.08444, over 3103165.91 frames. ], batch size: 55, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:44:08,557 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:17,264 INFO [train.py:904] (3/8) Epoch 6, batch 6200, loss[loss=0.2275, simple_loss=0.3057, pruned_loss=0.07468, over 16686.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3206, pruned_loss=0.08401, over 3092825.91 frames. ], batch size: 89, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:45:46,333 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:53,729 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.479e+02 4.387e+02 5.634e+02 1.000e+03, threshold=8.774e+02, percent-clipped=2.0 2023-04-28 12:46:08,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6308, 4.9321, 4.6375, 4.6341, 4.2807, 4.3240, 4.3761, 4.9596], device='cuda:3'), covar=tensor([0.0683, 0.0660, 0.0906, 0.0545, 0.0705, 0.0954, 0.0816, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0530, 0.0451, 0.0349, 0.0327, 0.0342, 0.0437, 0.0379], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:46:22,708 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:34,326 INFO [train.py:904] (3/8) Epoch 6, batch 6250, loss[loss=0.2292, simple_loss=0.3157, pruned_loss=0.07137, over 16711.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3196, pruned_loss=0.08295, over 3109194.98 frames. ], batch size: 83, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:50,596 INFO [train.py:904] (3/8) Epoch 6, batch 6300, loss[loss=0.2429, simple_loss=0.3157, pruned_loss=0.08504, over 15312.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3193, pruned_loss=0.08231, over 3117520.67 frames. ], batch size: 190, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:56,707 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:47:59,229 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:48:29,807 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.167e+02 3.660e+02 4.420e+02 5.766e+02 1.393e+03, threshold=8.841e+02, percent-clipped=5.0 2023-04-28 12:49:09,382 INFO [train.py:904] (3/8) Epoch 6, batch 6350, loss[loss=0.2553, simple_loss=0.3329, pruned_loss=0.08886, over 16469.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3204, pruned_loss=0.08392, over 3108296.79 frames. ], batch size: 75, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:49:14,807 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5132, 4.4803, 4.3395, 3.6559, 4.3090, 1.6191, 4.1040, 4.1399], device='cuda:3'), covar=tensor([0.0065, 0.0054, 0.0095, 0.0291, 0.0065, 0.1871, 0.0093, 0.0140], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0086, 0.0132, 0.0130, 0.0098, 0.0148, 0.0114, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:49:49,007 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5101, 4.4616, 4.3137, 3.6596, 4.2606, 1.6584, 4.1069, 4.1429], device='cuda:3'), covar=tensor([0.0062, 0.0060, 0.0099, 0.0310, 0.0072, 0.1887, 0.0092, 0.0149], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0086, 0.0133, 0.0131, 0.0098, 0.0150, 0.0115, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:50:26,696 INFO [train.py:904] (3/8) Epoch 6, batch 6400, loss[loss=0.2516, simple_loss=0.3206, pruned_loss=0.09133, over 15379.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3215, pruned_loss=0.08565, over 3093306.89 frames. ], batch size: 190, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:50:38,656 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:01,322 INFO [optim.py:368] (3/8) 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 12:51:07,810 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:51:19,394 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:36,856 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:42,172 INFO [train.py:904] (3/8) Epoch 6, batch 6450, loss[loss=0.2137, simple_loss=0.2992, pruned_loss=0.06409, over 15395.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3201, pruned_loss=0.08396, over 3105448.35 frames. ], batch size: 191, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:52:26,740 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:52:53,003 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:52:57,196 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:03,830 INFO [train.py:904] (3/8) Epoch 6, batch 6500, loss[loss=0.2451, simple_loss=0.3281, pruned_loss=0.08103, over 16356.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3183, pruned_loss=0.08354, over 3099335.93 frames. ], batch size: 35, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:53:15,508 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:53:25,363 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:41,778 INFO [optim.py:368] (3/8) 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 12:54:25,263 INFO [train.py:904] (3/8) Epoch 6, batch 6550, loss[loss=0.2526, simple_loss=0.342, pruned_loss=0.08164, over 16934.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3219, pruned_loss=0.08411, over 3117001.54 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:54:28,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5132, 4.2862, 4.1887, 2.9993, 4.0004, 4.3295, 4.0810, 2.4154], device='cuda:3'), covar=tensor([0.0313, 0.0018, 0.0025, 0.0198, 0.0028, 0.0040, 0.0024, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0054, 0.0058, 0.0113, 0.0061, 0.0071, 0.0064, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 12:54:31,725 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:40,093 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 6600, loss[loss=0.242, simple_loss=0.3224, pruned_loss=0.08082, over 16890.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3243, pruned_loss=0.08481, over 3114507.44 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:55:43,210 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:56:18,769 INFO [optim.py:368] (3/8) 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 12:56:29,968 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9299, 1.7194, 1.5049, 1.4978, 1.8751, 1.6069, 1.7524, 1.9156], device='cuda:3'), covar=tensor([0.0047, 0.0126, 0.0195, 0.0174, 0.0095, 0.0138, 0.0090, 0.0094], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0159, 0.0163, 0.0162, 0.0155, 0.0165, 0.0148, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 12:56:56,555 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 6650, loss[loss=0.217, simple_loss=0.2984, pruned_loss=0.06776, over 16844.00 frames. ], tot_loss[loss=0.248, simple_loss=0.3243, pruned_loss=0.08581, over 3103157.00 frames. ], batch size: 116, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:57:44,972 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 12:58:15,913 INFO [train.py:904] (3/8) Epoch 6, batch 6700, loss[loss=0.2262, simple_loss=0.3046, pruned_loss=0.07384, over 17037.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.323, pruned_loss=0.08627, over 3076821.22 frames. ], batch size: 50, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:58:20,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0022, 2.6345, 2.5969, 1.8193, 2.7126, 2.7522, 2.3743, 2.3476], device='cuda:3'), covar=tensor([0.0708, 0.0166, 0.0184, 0.0870, 0.0101, 0.0147, 0.0389, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0091, 0.0081, 0.0138, 0.0072, 0.0081, 0.0115, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 12:58:29,547 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:29,619 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:34,620 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:54,092 INFO [optim.py:368] (3/8) 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 12:59:24,851 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:59:34,109 INFO [train.py:904] (3/8) Epoch 6, batch 6750, loss[loss=0.2212, simple_loss=0.2966, pruned_loss=0.0729, over 16351.00 frames. ], tot_loss[loss=0.246, simple_loss=0.3211, pruned_loss=0.08542, over 3095066.34 frames. ], batch size: 146, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:59:43,068 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:07,993 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:13,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7420, 3.6167, 3.7466, 3.9448, 4.0019, 3.5786, 3.9463, 4.0241], device='cuda:3'), covar=tensor([0.0919, 0.0876, 0.1191, 0.0489, 0.0518, 0.1421, 0.0575, 0.0485], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0496, 0.0637, 0.0514, 0.0386, 0.0382, 0.0400, 0.0430], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:00:36,584 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 6800, loss[loss=0.2448, simple_loss=0.3284, pruned_loss=0.08067, over 16543.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.322, pruned_loss=0.08556, over 3083129.63 frames. ], batch size: 75, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:00:50,291 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3084, 3.2211, 2.5924, 2.1593, 2.3082, 2.0860, 3.2156, 3.2703], device='cuda:3'), covar=tensor([0.2174, 0.0756, 0.1284, 0.1622, 0.1804, 0.1486, 0.0454, 0.0713], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0250, 0.0267, 0.0251, 0.0283, 0.0202, 0.0248, 0.0259], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:00:52,598 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4668, 4.5384, 4.6379, 4.6428, 4.6211, 5.1525, 4.7974, 4.5487], device='cuda:3'), covar=tensor([0.1130, 0.1543, 0.1464, 0.1524, 0.2218, 0.0871, 0.1099, 0.2063], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0402, 0.0409, 0.0355, 0.0461, 0.0435, 0.0335, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 13:00:54,405 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 13:00:58,718 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:09,798 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.272e+02 3.624e+02 4.504e+02 5.898e+02 1.165e+03, threshold=9.008e+02, percent-clipped=3.0 2023-04-28 13:01:55,815 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5960, 2.8078, 2.5317, 4.3981, 3.1444, 4.1164, 1.3750, 3.0323], device='cuda:3'), covar=tensor([0.1410, 0.0607, 0.1104, 0.0072, 0.0270, 0.0310, 0.1524, 0.0707], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0144, 0.0170, 0.0094, 0.0196, 0.0189, 0.0164, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 13:02:06,899 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 6850, loss[loss=0.2381, simple_loss=0.3299, pruned_loss=0.07315, over 16842.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3232, pruned_loss=0.0857, over 3092042.70 frames. ], batch size: 116, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:02:23,509 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6073, 5.6202, 5.4289, 4.6148, 5.3846, 2.4585, 5.1078, 5.3395], device='cuda:3'), covar=tensor([0.0064, 0.0047, 0.0076, 0.0369, 0.0063, 0.1641, 0.0096, 0.0109], device='cuda:3'), in_proj_covar=tensor([0.0098, 0.0085, 0.0131, 0.0129, 0.0098, 0.0149, 0.0114, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:02:24,686 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:30,605 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 13:03:22,347 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:24,925 INFO [train.py:904] (3/8) Epoch 6, batch 6900, loss[loss=0.3486, simple_loss=0.3809, pruned_loss=0.1582, over 11409.00 frames. ], tot_loss[loss=0.2466, simple_loss=0.3244, pruned_loss=0.08442, over 3102897.73 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:03:25,270 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:02,709 INFO [optim.py:368] (3/8) 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 13:04:20,179 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0321, 2.9476, 2.7704, 2.0488, 2.5987, 2.1373, 2.7497, 2.9549], device='cuda:3'), covar=tensor([0.0263, 0.0435, 0.0446, 0.1267, 0.0623, 0.0802, 0.0548, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0126, 0.0153, 0.0139, 0.0133, 0.0124, 0.0138, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 13:04:37,886 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:40,788 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:43,531 INFO [train.py:904] (3/8) Epoch 6, batch 6950, loss[loss=0.2378, simple_loss=0.3205, pruned_loss=0.07752, over 16699.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3264, pruned_loss=0.08681, over 3084087.62 frames. ], batch size: 134, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:04:48,766 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 13:05:34,128 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:05:37,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2210, 4.5525, 2.4640, 5.0516, 3.0393, 4.8813, 2.5583, 3.2391], device='cuda:3'), covar=tensor([0.0124, 0.0204, 0.1406, 0.0029, 0.0650, 0.0264, 0.1426, 0.0611], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0154, 0.0177, 0.0083, 0.0161, 0.0189, 0.0187, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 13:05:37,614 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.18 vs. limit=5.0 2023-04-28 13:06:01,324 INFO [train.py:904] (3/8) Epoch 6, batch 7000, loss[loss=0.2485, simple_loss=0.3363, pruned_loss=0.0803, over 16200.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3264, pruned_loss=0.08553, over 3098086.06 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:06:06,474 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.231e+02 4.685e+02 6.535e+02 1.606e+03, threshold=9.370e+02, percent-clipped=8.0 2023-04-28 13:06:57,912 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9270, 5.1590, 4.8727, 4.8965, 4.5892, 4.4645, 4.6208, 5.2279], device='cuda:3'), covar=tensor([0.0789, 0.0782, 0.0996, 0.0609, 0.0694, 0.0822, 0.0789, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0521, 0.0451, 0.0341, 0.0322, 0.0346, 0.0425, 0.0377], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:07:07,541 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:07:10,604 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9658, 2.0686, 1.6550, 2.0313, 2.5647, 2.3308, 3.0275, 2.8596], device='cuda:3'), covar=tensor([0.0047, 0.0264, 0.0326, 0.0269, 0.0155, 0.0221, 0.0101, 0.0120], device='cuda:3'), in_proj_covar=tensor([0.0088, 0.0161, 0.0165, 0.0164, 0.0157, 0.0166, 0.0147, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:07:17,559 INFO [train.py:904] (3/8) Epoch 6, batch 7050, loss[loss=0.2421, simple_loss=0.3188, pruned_loss=0.08274, over 16739.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3267, pruned_loss=0.08503, over 3107641.12 frames. ], batch size: 124, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:07:44,457 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:07:57,071 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8784, 1.5623, 2.2586, 2.9412, 2.5931, 3.0429, 1.6935, 2.8927], device='cuda:3'), covar=tensor([0.0092, 0.0306, 0.0203, 0.0114, 0.0138, 0.0109, 0.0289, 0.0065], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0149, 0.0130, 0.0129, 0.0134, 0.0100, 0.0145, 0.0086], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 13:08:20,869 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:34,156 INFO [train.py:904] (3/8) Epoch 6, batch 7100, loss[loss=0.2202, simple_loss=0.307, pruned_loss=0.0667, over 16872.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3259, pruned_loss=0.08571, over 3091181.99 frames. ], batch size: 96, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:08:34,671 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:37,919 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 13:09:10,525 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 3.825e+02 4.803e+02 6.106e+02 1.425e+03, threshold=9.606e+02, percent-clipped=4.0 2023-04-28 13:09:32,898 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:48,232 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:49,598 INFO [train.py:904] (3/8) Epoch 6, batch 7150, loss[loss=0.3003, simple_loss=0.3447, pruned_loss=0.1279, over 11313.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.323, pruned_loss=0.08478, over 3093769.78 frames. ], batch size: 248, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:09:49,970 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:10:24,802 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5266, 2.0141, 2.3253, 4.2894, 2.0201, 2.7271, 2.3033, 2.2446], device='cuda:3'), covar=tensor([0.0721, 0.2796, 0.1470, 0.0301, 0.3372, 0.1502, 0.2298, 0.2683], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0339, 0.0278, 0.0312, 0.0385, 0.0352, 0.0303, 0.0399], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:11:00,232 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 7200, loss[loss=0.2029, simple_loss=0.2883, pruned_loss=0.05873, over 17098.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3201, pruned_loss=0.08258, over 3100998.78 frames. ], batch size: 49, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:11:24,131 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 13:11:41,800 INFO [optim.py:368] (3/8) 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 13:12:04,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9592, 5.0094, 4.6735, 4.1542, 4.7815, 1.7322, 4.5465, 4.7213], device='cuda:3'), covar=tensor([0.0045, 0.0038, 0.0110, 0.0260, 0.0050, 0.1774, 0.0070, 0.0105], device='cuda:3'), in_proj_covar=tensor([0.0096, 0.0083, 0.0129, 0.0128, 0.0097, 0.0147, 0.0112, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:12:28,079 INFO [train.py:904] (3/8) Epoch 6, batch 7250, loss[loss=0.2241, simple_loss=0.2978, pruned_loss=0.07519, over 16960.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3178, pruned_loss=0.0811, over 3119079.69 frames. ], batch size: 109, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:12:48,436 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 13:13:44,368 INFO [train.py:904] (3/8) Epoch 6, batch 7300, loss[loss=0.2318, simple_loss=0.3159, pruned_loss=0.07379, over 16679.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3167, pruned_loss=0.08053, over 3119123.16 frames. ], batch size: 134, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:13:49,514 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:14:21,942 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.373e+02 3.804e+02 4.653e+02 5.753e+02 1.061e+03, threshold=9.306e+02, percent-clipped=1.0 2023-04-28 13:14:43,399 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:02,139 INFO [train.py:904] (3/8) Epoch 6, batch 7350, loss[loss=0.2585, simple_loss=0.335, pruned_loss=0.09097, over 15573.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3171, pruned_loss=0.08118, over 3108185.65 frames. ], batch size: 191, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:15:03,735 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:28,181 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:04,421 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:09,181 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 7400, loss[loss=0.2259, simple_loss=0.3055, pruned_loss=0.07318, over 16499.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3184, pruned_loss=0.08192, over 3111766.13 frames. ], batch size: 68, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:16:19,780 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:43,737 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:57,163 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.315e+02 3.701e+02 4.363e+02 5.153e+02 1.099e+03, threshold=8.726e+02, percent-clipped=1.0 2023-04-28 13:17:13,360 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4200, 5.7338, 5.4714, 5.5510, 5.0978, 5.0770, 5.3452, 5.8427], device='cuda:3'), covar=tensor([0.0744, 0.0754, 0.0997, 0.0493, 0.0694, 0.0554, 0.0630, 0.0736], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0529, 0.0454, 0.0344, 0.0327, 0.0350, 0.0428, 0.0380], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:17:34,942 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:37,742 INFO [train.py:904] (3/8) Epoch 6, batch 7450, loss[loss=0.2368, simple_loss=0.3102, pruned_loss=0.08175, over 16715.00 frames. ], tot_loss[loss=0.2438, simple_loss=0.3202, pruned_loss=0.08373, over 3088895.02 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:17:39,887 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:45,811 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:18:59,287 INFO [train.py:904] (3/8) Epoch 6, batch 7500, loss[loss=0.2675, simple_loss=0.3356, pruned_loss=0.09972, over 15270.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.3208, pruned_loss=0.0829, over 3110399.18 frames. ], batch size: 191, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:19:39,439 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 7550, loss[loss=0.2209, simple_loss=0.2962, pruned_loss=0.07274, over 16595.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3201, pruned_loss=0.08305, over 3107293.78 frames. ], batch size: 68, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:21:33,625 INFO [train.py:904] (3/8) Epoch 6, batch 7600, loss[loss=0.2183, simple_loss=0.2971, pruned_loss=0.06977, over 17055.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3193, pruned_loss=0.08306, over 3113782.98 frames. ], batch size: 50, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:22:14,297 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.923e+02 3.777e+02 4.488e+02 5.757e+02 1.119e+03, threshold=8.975e+02, percent-clipped=2.0 2023-04-28 13:22:33,705 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:22:52,235 INFO [train.py:904] (3/8) Epoch 6, batch 7650, loss[loss=0.2496, simple_loss=0.323, pruned_loss=0.08813, over 16863.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3209, pruned_loss=0.08444, over 3103549.58 frames. ], batch size: 42, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:23:49,829 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:24:09,018 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5135, 4.4545, 4.9557, 4.8519, 4.9040, 4.5414, 4.5417, 4.2515], device='cuda:3'), covar=tensor([0.0205, 0.0341, 0.0249, 0.0379, 0.0349, 0.0264, 0.0704, 0.0413], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0245, 0.0251, 0.0246, 0.0296, 0.0264, 0.0363, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 13:24:11,082 INFO [train.py:904] (3/8) Epoch 6, batch 7700, loss[loss=0.3174, simple_loss=0.3654, pruned_loss=0.1347, over 11779.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3201, pruned_loss=0.08474, over 3098585.87 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:24:16,683 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1433, 4.1184, 4.2650, 4.1614, 4.2148, 4.7066, 4.3506, 4.0543], device='cuda:3'), covar=tensor([0.1371, 0.1641, 0.1491, 0.1822, 0.2172, 0.0932, 0.1218, 0.2292], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0407, 0.0415, 0.0360, 0.0473, 0.0444, 0.0341, 0.0482], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 13:24:51,994 INFO [optim.py:368] (3/8) 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 13:25:17,590 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8734, 3.1459, 3.0823, 1.9865, 2.8542, 3.0321, 3.0207, 1.7579], device='cuda:3'), covar=tensor([0.0360, 0.0027, 0.0033, 0.0259, 0.0060, 0.0073, 0.0046, 0.0315], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0054, 0.0058, 0.0115, 0.0062, 0.0074, 0.0064, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 13:25:23,661 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:28,785 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 7750, loss[loss=0.2369, simple_loss=0.3197, pruned_loss=0.07706, over 16973.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3209, pruned_loss=0.08474, over 3099264.45 frames. ], batch size: 41, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:26:07,949 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4633, 3.4993, 3.1911, 3.1249, 3.0841, 3.3741, 3.1953, 3.1582], device='cuda:3'), covar=tensor([0.0510, 0.0345, 0.0199, 0.0198, 0.0552, 0.0316, 0.1021, 0.0423], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0210, 0.0219, 0.0191, 0.0248, 0.0228, 0.0161, 0.0256], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:26:15,863 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1358, 1.7735, 2.0711, 3.7250, 1.8043, 2.5274, 2.0641, 1.9945], device='cuda:3'), covar=tensor([0.0743, 0.2566, 0.1404, 0.0313, 0.3162, 0.1335, 0.2168, 0.2457], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0343, 0.0281, 0.0315, 0.0387, 0.0354, 0.0305, 0.0399], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:26:46,072 INFO [train.py:904] (3/8) Epoch 6, batch 7800, loss[loss=0.2616, simple_loss=0.3456, pruned_loss=0.08875, over 16756.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3231, pruned_loss=0.08654, over 3095058.45 frames. ], batch size: 124, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:27:16,176 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2372, 4.2006, 4.0984, 4.0049, 3.7196, 4.1664, 3.9671, 3.8332], device='cuda:3'), covar=tensor([0.0514, 0.0421, 0.0213, 0.0191, 0.0876, 0.0396, 0.0517, 0.0549], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0212, 0.0221, 0.0194, 0.0251, 0.0230, 0.0164, 0.0258], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:27:26,211 INFO [optim.py:368] (3/8) 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 13:27:36,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8123, 1.6549, 1.4807, 1.4252, 1.7877, 1.5740, 1.6671, 1.8635], device='cuda:3'), covar=tensor([0.0050, 0.0135, 0.0193, 0.0181, 0.0095, 0.0134, 0.0103, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0086, 0.0161, 0.0167, 0.0164, 0.0157, 0.0168, 0.0149, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:28:01,703 INFO [train.py:904] (3/8) Epoch 6, batch 7850, loss[loss=0.2367, simple_loss=0.3182, pruned_loss=0.07757, over 16901.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3245, pruned_loss=0.08729, over 3070383.10 frames. ], batch size: 116, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:14,322 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 13:29:14,704 INFO [train.py:904] (3/8) Epoch 6, batch 7900, loss[loss=0.2999, simple_loss=0.3477, pruned_loss=0.1261, over 11265.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3229, pruned_loss=0.08607, over 3076653.59 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:53,321 INFO [optim.py:368] (3/8) 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 13:30:02,270 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:30:31,227 INFO [train.py:904] (3/8) Epoch 6, batch 7950, loss[loss=0.2985, simple_loss=0.349, pruned_loss=0.124, over 11625.00 frames. ], tot_loss[loss=0.2486, simple_loss=0.3234, pruned_loss=0.08691, over 3071617.96 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:31:33,507 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:31:46,443 INFO [train.py:904] (3/8) Epoch 6, batch 8000, loss[loss=0.2348, simple_loss=0.3186, pruned_loss=0.07549, over 16875.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3228, pruned_loss=0.08641, over 3085653.53 frames. ], batch size: 116, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:32:25,081 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 13:32:57,955 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:02,601 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 8050, loss[loss=0.2214, simple_loss=0.3037, pruned_loss=0.06953, over 15279.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3217, pruned_loss=0.08474, over 3108391.25 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:33:58,014 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:12,440 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:16,876 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:20,744 INFO [train.py:904] (3/8) Epoch 6, batch 8100, loss[loss=0.2514, simple_loss=0.3297, pruned_loss=0.08652, over 16669.00 frames. ], tot_loss[loss=0.245, simple_loss=0.3213, pruned_loss=0.08428, over 3093039.90 frames. ], batch size: 62, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:03,715 INFO [optim.py:368] (3/8) 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 13:35:13,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1863, 3.7418, 3.7293, 2.4078, 3.3633, 3.6923, 3.5381, 2.0551], device='cuda:3'), covar=tensor([0.0377, 0.0021, 0.0028, 0.0262, 0.0060, 0.0062, 0.0040, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0055, 0.0058, 0.0115, 0.0062, 0.0074, 0.0063, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 13:35:16,080 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0597, 5.3177, 5.0408, 5.0457, 4.7066, 4.5949, 4.8674, 5.3769], device='cuda:3'), covar=tensor([0.0752, 0.0736, 0.0910, 0.0534, 0.0660, 0.0722, 0.0638, 0.0740], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0521, 0.0448, 0.0337, 0.0322, 0.0346, 0.0426, 0.0374], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:35:24,715 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9063, 3.9601, 3.7945, 3.6526, 3.4675, 3.8273, 3.5843, 3.6547], device='cuda:3'), covar=tensor([0.0576, 0.0326, 0.0235, 0.0197, 0.0818, 0.0381, 0.0824, 0.0518], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0218, 0.0226, 0.0195, 0.0255, 0.0232, 0.0165, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:35:38,303 INFO [train.py:904] (3/8) Epoch 6, batch 8150, loss[loss=0.2401, simple_loss=0.3134, pruned_loss=0.08338, over 16694.00 frames. ], tot_loss[loss=0.2436, simple_loss=0.3195, pruned_loss=0.08391, over 3086345.62 frames. ], batch size: 134, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:46,339 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:36:03,669 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1024, 3.6618, 3.5499, 2.3136, 3.3481, 3.6491, 3.3719, 1.6962], device='cuda:3'), covar=tensor([0.0379, 0.0035, 0.0045, 0.0285, 0.0071, 0.0088, 0.0083, 0.0398], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0054, 0.0058, 0.0114, 0.0062, 0.0074, 0.0063, 0.0106], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 13:36:46,865 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 13:36:54,138 INFO [train.py:904] (3/8) Epoch 6, batch 8200, loss[loss=0.2274, simple_loss=0.3091, pruned_loss=0.07282, over 16677.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3174, pruned_loss=0.08335, over 3088798.86 frames. ], batch size: 134, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:36:56,703 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2291, 3.4015, 3.4965, 1.6162, 3.6921, 3.6412, 2.7970, 2.6785], device='cuda:3'), covar=tensor([0.0743, 0.0117, 0.0126, 0.1172, 0.0063, 0.0100, 0.0373, 0.0401], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0093, 0.0084, 0.0142, 0.0073, 0.0084, 0.0119, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 13:37:02,747 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 13:37:03,746 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8299, 1.3418, 1.6183, 1.6553, 1.7927, 1.8592, 1.4020, 1.7365], device='cuda:3'), covar=tensor([0.0111, 0.0185, 0.0108, 0.0135, 0.0107, 0.0077, 0.0180, 0.0049], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0147, 0.0129, 0.0128, 0.0134, 0.0099, 0.0145, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 13:37:21,733 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 13:37:46,317 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3562, 1.8313, 2.0531, 3.8398, 1.8164, 2.5358, 2.0372, 1.9730], device='cuda:3'), covar=tensor([0.0639, 0.2922, 0.1587, 0.0327, 0.3577, 0.1470, 0.2469, 0.2879], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0340, 0.0278, 0.0313, 0.0384, 0.0352, 0.0305, 0.0398], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:38:06,320 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3660, 1.8791, 2.0594, 3.8171, 1.9015, 2.5454, 2.1068, 2.0003], device='cuda:3'), covar=tensor([0.0630, 0.2854, 0.1585, 0.0347, 0.3434, 0.1564, 0.2383, 0.2837], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0339, 0.0278, 0.0313, 0.0383, 0.0352, 0.0305, 0.0397], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:38:17,129 INFO [train.py:904] (3/8) Epoch 6, batch 8250, loss[loss=0.2064, simple_loss=0.2814, pruned_loss=0.06564, over 12229.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3164, pruned_loss=0.08102, over 3075680.93 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:16,122 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:39:35,532 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 13:39:37,602 INFO [train.py:904] (3/8) Epoch 6, batch 8300, loss[loss=0.1994, simple_loss=0.2942, pruned_loss=0.05232, over 16866.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3128, pruned_loss=0.07761, over 3056511.85 frames. ], batch size: 96, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:39:52,476 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 13:40:05,967 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 13:40:22,347 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 8350, loss[loss=0.2018, simple_loss=0.2932, pruned_loss=0.05519, over 16435.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3107, pruned_loss=0.07442, over 3059836.47 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:41:04,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5944, 4.8182, 4.9380, 4.8307, 4.7887, 5.3243, 4.8661, 4.6053], device='cuda:3'), covar=tensor([0.0823, 0.1304, 0.1121, 0.1447, 0.2058, 0.0807, 0.1169, 0.2054], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0385, 0.0399, 0.0336, 0.0444, 0.0422, 0.0320, 0.0457], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:41:38,047 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-28 13:41:49,324 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:42:03,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0486, 5.3861, 5.1510, 5.2129, 4.6999, 4.6588, 4.9322, 5.4565], device='cuda:3'), covar=tensor([0.0827, 0.0867, 0.1003, 0.0489, 0.0675, 0.0704, 0.0665, 0.0766], device='cuda:3'), in_proj_covar=tensor([0.0396, 0.0513, 0.0441, 0.0335, 0.0318, 0.0344, 0.0422, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:42:21,175 INFO [train.py:904] (3/8) Epoch 6, batch 8400, loss[loss=0.1818, simple_loss=0.2791, pruned_loss=0.04225, over 16589.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3071, pruned_loss=0.0717, over 3049566.88 frames. ], batch size: 62, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:42:35,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3710, 3.3229, 2.7058, 2.1093, 2.3016, 2.1535, 3.2981, 3.2700], device='cuda:3'), covar=tensor([0.2368, 0.0656, 0.1219, 0.1715, 0.1915, 0.1518, 0.0373, 0.0730], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0236, 0.0260, 0.0243, 0.0263, 0.0196, 0.0238, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:42:42,355 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:42:54,370 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9763, 3.9909, 3.8666, 3.7125, 3.5258, 3.9682, 3.6504, 3.7071], device='cuda:3'), covar=tensor([0.0441, 0.0339, 0.0239, 0.0171, 0.0723, 0.0289, 0.0714, 0.0451], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0208, 0.0220, 0.0188, 0.0243, 0.0223, 0.0157, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:43:05,301 INFO [optim.py:368] (3/8) 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 13:43:35,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4563, 2.4598, 2.0105, 2.2795, 2.8542, 2.6246, 3.2292, 3.1814], device='cuda:3'), covar=tensor([0.0031, 0.0203, 0.0281, 0.0227, 0.0140, 0.0199, 0.0102, 0.0098], device='cuda:3'), in_proj_covar=tensor([0.0085, 0.0161, 0.0164, 0.0163, 0.0159, 0.0164, 0.0145, 0.0143], device='cuda:3'), out_proj_covar=tensor([9.8514e-05, 1.8688e-04, 1.8564e-04, 1.8535e-04, 1.8659e-04, 1.9012e-04, 1.6313e-04, 1.6549e-04], device='cuda:3') 2023-04-28 13:43:40,735 INFO [train.py:904] (3/8) Epoch 6, batch 8450, loss[loss=0.1789, simple_loss=0.2752, pruned_loss=0.04126, over 16843.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3054, pruned_loss=0.06992, over 3035775.37 frames. ], batch size: 102, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:43:46,913 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-28 13:44:10,019 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8003, 3.5246, 3.2587, 4.7570, 4.0172, 4.6703, 1.6517, 3.3290], device='cuda:3'), covar=tensor([0.1372, 0.0433, 0.0731, 0.0099, 0.0162, 0.0259, 0.1389, 0.0594], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0142, 0.0167, 0.0093, 0.0186, 0.0186, 0.0163, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 13:44:19,049 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1455, 1.3380, 1.8718, 1.9833, 2.2254, 2.2245, 1.4567, 2.1550], device='cuda:3'), covar=tensor([0.0108, 0.0278, 0.0132, 0.0156, 0.0122, 0.0110, 0.0245, 0.0069], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0150, 0.0131, 0.0130, 0.0137, 0.0098, 0.0147, 0.0085], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 13:44:19,058 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 8500, loss[loss=0.1942, simple_loss=0.2777, pruned_loss=0.0554, over 15199.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3008, pruned_loss=0.06701, over 3029192.83 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:45:16,429 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-28 13:45:19,655 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:45:46,051 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.693e+02 3.373e+02 4.229e+02 1.183e+03, threshold=6.745e+02, percent-clipped=1.0 2023-04-28 13:46:21,074 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 13:46:25,537 INFO [train.py:904] (3/8) Epoch 6, batch 8550, loss[loss=0.2263, simple_loss=0.3141, pruned_loss=0.06923, over 16063.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2976, pruned_loss=0.06574, over 2988350.89 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:46:50,489 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 13:47:23,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6741, 3.8037, 1.8235, 4.0416, 2.4286, 3.9822, 1.9771, 2.7658], device='cuda:3'), covar=tensor([0.0139, 0.0252, 0.1673, 0.0065, 0.0835, 0.0331, 0.1590, 0.0647], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0146, 0.0173, 0.0081, 0.0154, 0.0178, 0.0185, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 13:47:38,191 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 8600, loss[loss=0.2005, simple_loss=0.2713, pruned_loss=0.06481, over 12226.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2981, pruned_loss=0.0647, over 2989236.81 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:49:03,381 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:49:20,373 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 13:49:46,454 INFO [train.py:904] (3/8) Epoch 6, batch 8650, loss[loss=0.2002, simple_loss=0.2812, pruned_loss=0.05959, over 12258.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.296, pruned_loss=0.06283, over 2986703.08 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:50:55,058 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:51:16,106 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1634, 5.4714, 5.1629, 5.2527, 4.9464, 4.8244, 4.9583, 5.5233], device='cuda:3'), covar=tensor([0.0779, 0.0674, 0.0821, 0.0422, 0.0524, 0.0674, 0.0708, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0382, 0.0503, 0.0424, 0.0326, 0.0308, 0.0335, 0.0413, 0.0365], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:51:28,489 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-28 13:51:31,166 INFO [train.py:904] (3/8) Epoch 6, batch 8700, loss[loss=0.1832, simple_loss=0.2752, pruned_loss=0.04556, over 16806.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2928, pruned_loss=0.06099, over 2994687.84 frames. ], batch size: 83, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:51:49,204 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 13:52:21,221 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 6, batch 8750, loss[loss=0.211, simple_loss=0.3148, pruned_loss=0.05354, over 16828.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2918, pruned_loss=0.05969, over 3014388.56 frames. ], batch size: 102, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:53:08,781 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:54,427 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:54:55,917 INFO [train.py:904] (3/8) Epoch 6, batch 8800, loss[loss=0.1879, simple_loss=0.2837, pruned_loss=0.04608, over 16943.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2898, pruned_loss=0.05851, over 3005806.58 frames. ], batch size: 109, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:55:06,452 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6222, 3.5955, 4.0440, 3.9847, 4.0370, 3.7335, 3.7808, 3.6817], device='cuda:3'), covar=tensor([0.0258, 0.0462, 0.0373, 0.0487, 0.0340, 0.0327, 0.0673, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0238, 0.0237, 0.0241, 0.0238, 0.0281, 0.0256, 0.0347, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 13:55:17,679 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:19,835 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:52,199 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.649e+02 3.204e+02 3.980e+02 7.645e+02, threshold=6.408e+02, percent-clipped=1.0 2023-04-28 13:56:30,521 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4036, 4.3772, 4.2034, 3.7775, 4.2372, 1.5226, 4.0789, 3.9729], device='cuda:3'), covar=tensor([0.0049, 0.0042, 0.0090, 0.0191, 0.0056, 0.1908, 0.0080, 0.0130], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0078, 0.0125, 0.0117, 0.0093, 0.0146, 0.0108, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 13:56:37,975 INFO [train.py:904] (3/8) Epoch 6, batch 8850, loss[loss=0.197, simple_loss=0.276, pruned_loss=0.05899, over 12455.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.292, pruned_loss=0.05765, over 3010738.85 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:56:56,205 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:57:41,030 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 13:58:01,561 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 13:58:21,072 INFO [train.py:904] (3/8) Epoch 6, batch 8900, loss[loss=0.2117, simple_loss=0.3007, pruned_loss=0.06133, over 16817.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.292, pruned_loss=0.05678, over 3017562.46 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:58:34,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0249, 2.7929, 2.7322, 1.7256, 2.8987, 2.9575, 2.5019, 2.4756], device='cuda:3'), covar=tensor([0.0632, 0.0132, 0.0155, 0.0931, 0.0064, 0.0106, 0.0314, 0.0357], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0087, 0.0078, 0.0137, 0.0066, 0.0079, 0.0112, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 13:59:22,390 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.981e+02 2.897e+02 3.325e+02 4.292e+02 6.901e+02, threshold=6.651e+02, percent-clipped=1.0 2023-04-28 14:00:23,135 INFO [train.py:904] (3/8) Epoch 6, batch 8950, loss[loss=0.2088, simple_loss=0.2868, pruned_loss=0.06546, over 12669.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2923, pruned_loss=0.05744, over 3027518.62 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:02:12,299 INFO [train.py:904] (3/8) Epoch 6, batch 9000, loss[loss=0.1823, simple_loss=0.2731, pruned_loss=0.04579, over 15634.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2893, pruned_loss=0.05568, over 3046589.98 frames. ], batch size: 194, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:02:12,300 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 14:02:22,191 INFO [train.py:938] (3/8) Epoch 6, validation: loss=0.1682, simple_loss=0.2716, pruned_loss=0.03235, over 944034.00 frames. 2023-04-28 14:02:22,191 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 14:03:21,581 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 9050, loss[loss=0.215, simple_loss=0.2933, pruned_loss=0.06837, over 12600.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2904, pruned_loss=0.05646, over 3051851.08 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:04:07,158 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9853, 4.2677, 2.1644, 4.6249, 2.8661, 4.5068, 2.4359, 3.0868], device='cuda:3'), covar=tensor([0.0166, 0.0204, 0.1601, 0.0055, 0.0739, 0.0345, 0.1345, 0.0605], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0147, 0.0176, 0.0081, 0.0155, 0.0179, 0.0185, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 14:04:44,677 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:21,853 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 14:05:45,896 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 9100, loss[loss=0.198, simple_loss=0.2921, pruned_loss=0.05194, over 16954.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2894, pruned_loss=0.05658, over 3058315.54 frames. ], batch size: 116, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:06:04,208 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:24,265 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:37,079 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1781, 2.9140, 2.8180, 2.0327, 2.6116, 2.0751, 2.6935, 2.8541], device='cuda:3'), covar=tensor([0.0335, 0.0593, 0.0499, 0.1445, 0.0680, 0.0959, 0.0746, 0.0708], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0120, 0.0150, 0.0140, 0.0131, 0.0124, 0.0135, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 14:06:37,368 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 14:06:55,735 INFO [optim.py:368] (3/8) 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 14:07:19,847 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:07:46,829 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 14:07:47,008 INFO [train.py:904] (3/8) Epoch 6, batch 9150, loss[loss=0.1977, simple_loss=0.2855, pruned_loss=0.05497, over 16835.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2899, pruned_loss=0.05617, over 3057638.72 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:08:06,745 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4270, 3.7948, 3.9119, 1.7441, 4.1355, 4.1429, 3.1607, 3.0974], device='cuda:3'), covar=tensor([0.0714, 0.0128, 0.0124, 0.1124, 0.0042, 0.0052, 0.0259, 0.0355], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0086, 0.0076, 0.0136, 0.0065, 0.0077, 0.0111, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 14:08:06,782 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:08:18,624 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0666, 1.4420, 1.7727, 2.0965, 2.1337, 2.1526, 1.5492, 2.2171], device='cuda:3'), covar=tensor([0.0115, 0.0255, 0.0145, 0.0170, 0.0148, 0.0114, 0.0263, 0.0064], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0148, 0.0130, 0.0129, 0.0138, 0.0095, 0.0147, 0.0082], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 14:09:29,950 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 9200, loss[loss=0.1695, simple_loss=0.2598, pruned_loss=0.03956, over 17142.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2846, pruned_loss=0.05462, over 3056431.02 frames. ], batch size: 46, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:10:12,309 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 14:10:22,279 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 9250, loss[loss=0.193, simple_loss=0.2835, pruned_loss=0.05122, over 15430.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2844, pruned_loss=0.05484, over 3043642.85 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:11:11,569 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4965, 4.6438, 4.7167, 4.6604, 4.6895, 5.1864, 4.8297, 4.5168], device='cuda:3'), covar=tensor([0.0864, 0.1442, 0.1536, 0.1653, 0.2249, 0.0967, 0.1045, 0.2050], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0382, 0.0388, 0.0331, 0.0438, 0.0422, 0.0317, 0.0448], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:11:45,895 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3238, 2.4789, 1.9527, 2.2348, 2.9661, 2.6623, 3.2880, 3.1681], device='cuda:3'), covar=tensor([0.0030, 0.0219, 0.0257, 0.0237, 0.0108, 0.0174, 0.0075, 0.0096], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0163, 0.0164, 0.0162, 0.0158, 0.0164, 0.0142, 0.0143], device='cuda:3'), out_proj_covar=tensor([9.5012e-05, 1.8960e-04, 1.8484e-04, 1.8337e-04, 1.8413e-04, 1.9026e-04, 1.5855e-04, 1.6470e-04], device='cuda:3') 2023-04-28 14:12:33,901 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:12:55,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4944, 4.2890, 4.0626, 2.1807, 3.2018, 2.5659, 3.8129, 4.1509], device='cuda:3'), covar=tensor([0.0303, 0.0545, 0.0387, 0.1523, 0.0654, 0.0882, 0.0647, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0120, 0.0150, 0.0139, 0.0131, 0.0124, 0.0135, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 14:12:58,477 INFO [train.py:904] (3/8) Epoch 6, batch 9300, loss[loss=0.1979, simple_loss=0.2663, pruned_loss=0.06477, over 12397.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2821, pruned_loss=0.05401, over 3034254.52 frames. ], batch size: 247, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:13:01,056 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:33,149 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:51,439 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5344, 3.5601, 3.3802, 3.2194, 3.1927, 3.4594, 3.2716, 3.2822], device='cuda:3'), covar=tensor([0.0436, 0.0438, 0.0233, 0.0180, 0.0550, 0.0343, 0.0920, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0202, 0.0217, 0.0187, 0.0236, 0.0218, 0.0153, 0.0243], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:14:01,689 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:14:40,790 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1711, 3.1776, 3.1544, 1.6099, 3.3557, 3.3609, 2.7151, 2.7311], device='cuda:3'), covar=tensor([0.0725, 0.0157, 0.0171, 0.1234, 0.0069, 0.0082, 0.0355, 0.0357], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0088, 0.0078, 0.0138, 0.0067, 0.0079, 0.0112, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 14:14:44,853 INFO [train.py:904] (3/8) Epoch 6, batch 9350, loss[loss=0.2194, simple_loss=0.2946, pruned_loss=0.07214, over 12381.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2823, pruned_loss=0.05415, over 3046634.92 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:14:45,723 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:15:12,249 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:15:37,206 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:03,964 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 9400, loss[loss=0.1725, simple_loss=0.2553, pruned_loss=0.04487, over 12624.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2821, pruned_loss=0.05411, over 3025682.27 frames. ], batch size: 248, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:16:40,814 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:17:25,097 INFO [optim.py:368] (3/8) 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 14:18:05,598 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 9450, loss[loss=0.1821, simple_loss=0.2758, pruned_loss=0.04426, over 16723.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2838, pruned_loss=0.05425, over 3033903.73 frames. ], batch size: 83, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:18:14,581 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:19,280 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:26,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6211, 3.8127, 3.0172, 2.2773, 2.5570, 2.2971, 4.0050, 3.5978], device='cuda:3'), covar=tensor([0.2253, 0.0567, 0.1208, 0.1792, 0.1872, 0.1562, 0.0335, 0.0675], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0236, 0.0258, 0.0242, 0.0239, 0.0197, 0.0235, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:19:39,940 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 9500, loss[loss=0.1681, simple_loss=0.2613, pruned_loss=0.03747, over 16607.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2825, pruned_loss=0.05334, over 3050676.74 frames. ], batch size: 57, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:20:11,547 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:20:13,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4088, 1.4312, 1.8803, 2.4381, 2.3836, 2.4718, 1.6368, 2.4568], device='cuda:3'), covar=tensor([0.0096, 0.0277, 0.0195, 0.0136, 0.0128, 0.0120, 0.0260, 0.0079], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0148, 0.0131, 0.0129, 0.0136, 0.0096, 0.0146, 0.0082], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 14:20:40,247 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9085, 5.1931, 5.2677, 5.2351, 5.1348, 5.7856, 5.4094, 5.1085], device='cuda:3'), covar=tensor([0.0670, 0.1539, 0.1740, 0.1514, 0.2356, 0.0917, 0.1069, 0.2042], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0379, 0.0380, 0.0325, 0.0431, 0.0410, 0.0315, 0.0436], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:20:47,848 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9835, 3.8514, 3.8580, 3.3466, 3.8307, 1.7075, 3.5845, 3.6006], device='cuda:3'), covar=tensor([0.0073, 0.0070, 0.0101, 0.0211, 0.0072, 0.1954, 0.0106, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0092, 0.0078, 0.0122, 0.0113, 0.0093, 0.0146, 0.0108, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:20:51,246 INFO [optim.py:368] (3/8) 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 14:21:39,641 INFO [train.py:904] (3/8) Epoch 6, batch 9550, loss[loss=0.2003, simple_loss=0.2969, pruned_loss=0.05183, over 16399.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2825, pruned_loss=0.05328, over 3060323.89 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:22:14,246 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-04-28 14:22:19,358 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4160, 4.2134, 4.4982, 4.6087, 4.7798, 4.3248, 4.7998, 4.7605], device='cuda:3'), covar=tensor([0.0977, 0.0798, 0.0970, 0.0535, 0.0383, 0.0622, 0.0325, 0.0358], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0463, 0.0578, 0.0469, 0.0358, 0.0352, 0.0369, 0.0398], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:23:21,497 INFO [train.py:904] (3/8) Epoch 6, batch 9600, loss[loss=0.2439, simple_loss=0.3322, pruned_loss=0.07783, over 15406.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2853, pruned_loss=0.05452, over 3064763.04 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:24:17,215 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.768e+02 3.406e+02 4.197e+02 1.161e+03, threshold=6.812e+02, percent-clipped=3.0 2023-04-28 14:24:57,309 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 14:25:09,700 INFO [train.py:904] (3/8) Epoch 6, batch 9650, loss[loss=0.2198, simple_loss=0.3051, pruned_loss=0.06723, over 16752.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2875, pruned_loss=0.05513, over 3053296.48 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:25:28,080 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 14:25:57,584 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:13,529 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9462, 3.8800, 4.3940, 4.3518, 4.3268, 4.0202, 4.0903, 3.9083], device='cuda:3'), covar=tensor([0.0255, 0.0472, 0.0326, 0.0345, 0.0378, 0.0318, 0.0677, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0237, 0.0243, 0.0236, 0.0281, 0.0258, 0.0349, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 14:26:22,100 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:58,017 INFO [train.py:904] (3/8) Epoch 6, batch 9700, loss[loss=0.1967, simple_loss=0.2856, pruned_loss=0.05393, over 16867.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2862, pruned_loss=0.05465, over 3054207.85 frames. ], batch size: 116, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:27:59,429 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 6, batch 9750, loss[loss=0.1751, simple_loss=0.268, pruned_loss=0.04105, over 16797.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2847, pruned_loss=0.05465, over 3050117.52 frames. ], batch size: 83, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:28:47,809 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:11,545 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:19,565 INFO [train.py:904] (3/8) Epoch 6, batch 9800, loss[loss=0.2027, simple_loss=0.3053, pruned_loss=0.05, over 16777.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2851, pruned_loss=0.05385, over 3052312.00 frames. ], batch size: 83, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:30:22,403 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:27,280 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:31:14,366 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.802e+02 3.388e+02 4.163e+02 9.547e+02, threshold=6.775e+02, percent-clipped=1.0 2023-04-28 14:31:33,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1927, 4.1466, 4.0050, 3.4210, 4.0157, 1.6318, 3.8146, 3.7932], device='cuda:3'), covar=tensor([0.0066, 0.0057, 0.0104, 0.0229, 0.0066, 0.2015, 0.0100, 0.0181], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0080, 0.0124, 0.0114, 0.0093, 0.0148, 0.0108, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:31:48,142 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:32:03,897 INFO [train.py:904] (3/8) Epoch 6, batch 9850, loss[loss=0.2032, simple_loss=0.2958, pruned_loss=0.05526, over 16886.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2871, pruned_loss=0.0538, over 3055272.13 frames. ], batch size: 90, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:33:38,817 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 14:33:56,848 INFO [train.py:904] (3/8) Epoch 6, batch 9900, loss[loss=0.1968, simple_loss=0.2904, pruned_loss=0.05163, over 16362.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2872, pruned_loss=0.05315, over 3062991.99 frames. ], batch size: 166, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:35:04,552 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.757e+02 3.371e+02 4.013e+02 6.843e+02, threshold=6.741e+02, percent-clipped=1.0 2023-04-28 14:35:19,022 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 14:35:42,389 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:35:52,849 INFO [train.py:904] (3/8) Epoch 6, batch 9950, loss[loss=0.1972, simple_loss=0.2915, pruned_loss=0.05146, over 15415.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2901, pruned_loss=0.05394, over 3075944.10 frames. ], batch size: 191, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:36:09,985 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:28,986 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8844, 2.4408, 2.2384, 3.2666, 2.5198, 3.3612, 1.5581, 2.6987], device='cuda:3'), covar=tensor([0.1097, 0.0479, 0.0946, 0.0093, 0.0134, 0.0365, 0.1261, 0.0667], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0139, 0.0165, 0.0090, 0.0156, 0.0184, 0.0160, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 14:36:34,757 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:42,977 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:13,615 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:21,128 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-28 14:37:38,496 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 6, batch 10000, loss[loss=0.1849, simple_loss=0.2773, pruned_loss=0.04631, over 16588.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2885, pruned_loss=0.05364, over 3081841.65 frames. ], batch size: 62, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:38:06,644 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:29,852 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:50,641 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:51,219 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.599e+02 3.122e+02 4.143e+02 8.746e+02, threshold=6.243e+02, percent-clipped=4.0 2023-04-28 14:38:52,635 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1522, 1.3960, 1.6528, 2.2816, 2.2168, 2.3692, 1.5747, 2.2603], device='cuda:3'), covar=tensor([0.0101, 0.0299, 0.0178, 0.0139, 0.0142, 0.0099, 0.0261, 0.0064], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0148, 0.0132, 0.0129, 0.0136, 0.0095, 0.0144, 0.0081], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 14:38:57,162 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:39:12,742 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1826, 4.2207, 4.0372, 3.8727, 3.6611, 4.1247, 3.8969, 3.8337], device='cuda:3'), covar=tensor([0.0503, 0.0436, 0.0268, 0.0224, 0.0842, 0.0371, 0.0564, 0.0505], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0201, 0.0215, 0.0187, 0.0236, 0.0217, 0.0151, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:39:33,417 INFO [train.py:904] (3/8) Epoch 6, batch 10050, loss[loss=0.2333, simple_loss=0.3227, pruned_loss=0.07197, over 15458.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2883, pruned_loss=0.05336, over 3078860.19 frames. ], batch size: 191, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:05,270 INFO [train.py:904] (3/8) Epoch 6, batch 10100, loss[loss=0.2374, simple_loss=0.3199, pruned_loss=0.07739, over 16331.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2893, pruned_loss=0.05399, over 3088292.63 frames. ], batch size: 146, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:10,985 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:42:01,263 INFO [optim.py:368] (3/8) 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 14:42:01,530 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8564, 4.2113, 3.9968, 3.9941, 3.6302, 3.7548, 3.8535, 4.1374], device='cuda:3'), covar=tensor([0.0810, 0.0809, 0.0839, 0.0478, 0.0711, 0.1352, 0.0686, 0.0995], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0501, 0.0411, 0.0327, 0.0310, 0.0332, 0.0411, 0.0365], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:42:02,696 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9036, 3.8905, 3.7978, 3.1977, 3.8193, 1.6270, 3.6280, 3.6168], device='cuda:3'), covar=tensor([0.0091, 0.0071, 0.0126, 0.0242, 0.0083, 0.2138, 0.0115, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0094, 0.0080, 0.0125, 0.0112, 0.0093, 0.0149, 0.0109, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:42:22,731 INFO [train.py:904] (3/8) Epoch 6, batch 10150, loss[loss=0.202, simple_loss=0.2744, pruned_loss=0.06482, over 12535.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2874, pruned_loss=0.05415, over 3054306.01 frames. ], batch size: 250, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:42:48,314 INFO [train.py:904] (3/8) Epoch 7, batch 0, loss[loss=0.3287, simple_loss=0.3784, pruned_loss=0.1395, over 16316.00 frames. ], tot_loss[loss=0.3287, simple_loss=0.3784, pruned_loss=0.1395, over 16316.00 frames. ], batch size: 165, lr: 1.02e-02, grad_scale: 8.0 2023-04-28 14:42:48,314 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 14:42:55,783 INFO [train.py:938] (3/8) Epoch 7, validation: loss=0.1665, simple_loss=0.2702, pruned_loss=0.03141, over 944034.00 frames. 2023-04-28 14:42:55,784 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 14:42:55,981 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:43:13,837 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 14:43:38,149 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5717, 4.0683, 4.1670, 2.1674, 3.4158, 2.5552, 4.0005, 3.8873], device='cuda:3'), covar=tensor([0.0229, 0.0506, 0.0397, 0.1469, 0.0585, 0.0837, 0.0530, 0.0866], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0119, 0.0150, 0.0137, 0.0130, 0.0122, 0.0134, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 14:43:43,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7334, 2.6019, 2.2787, 2.4882, 3.0711, 2.9001, 3.5661, 3.2658], device='cuda:3'), covar=tensor([0.0028, 0.0225, 0.0280, 0.0232, 0.0143, 0.0190, 0.0100, 0.0131], device='cuda:3'), in_proj_covar=tensor([0.0083, 0.0165, 0.0164, 0.0162, 0.0159, 0.0165, 0.0142, 0.0145], device='cuda:3'), out_proj_covar=tensor([9.4271e-05, 1.9052e-04, 1.8500e-04, 1.8335e-04, 1.8436e-04, 1.8964e-04, 1.5666e-04, 1.6599e-04], device='cuda:3') 2023-04-28 14:44:05,495 INFO [train.py:904] (3/8) Epoch 7, batch 50, loss[loss=0.2395, simple_loss=0.2998, pruned_loss=0.08963, over 16765.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3061, pruned_loss=0.07732, over 746530.06 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:44:49,441 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 3.058e+02 3.620e+02 4.703e+02 1.122e+03, threshold=7.241e+02, percent-clipped=6.0 2023-04-28 14:45:08,496 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 14:45:10,881 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 14:45:15,337 INFO [train.py:904] (3/8) Epoch 7, batch 100, loss[loss=0.1786, simple_loss=0.2712, pruned_loss=0.04297, over 17192.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.2983, pruned_loss=0.07259, over 1318904.67 frames. ], batch size: 46, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:45:19,416 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0677, 3.1762, 1.7017, 3.2187, 2.2494, 3.2231, 1.8738, 2.5612], device='cuda:3'), covar=tensor([0.0194, 0.0335, 0.1599, 0.0160, 0.0797, 0.0447, 0.1347, 0.0624], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0154, 0.0182, 0.0088, 0.0159, 0.0188, 0.0192, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 14:45:26,279 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-28 14:45:52,790 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 14:46:12,691 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1465, 1.7798, 2.4265, 2.9600, 2.8078, 3.3612, 2.0008, 3.2845], device='cuda:3'), covar=tensor([0.0096, 0.0269, 0.0158, 0.0131, 0.0133, 0.0090, 0.0244, 0.0082], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0150, 0.0133, 0.0132, 0.0137, 0.0097, 0.0145, 0.0084], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 14:46:24,664 INFO [train.py:904] (3/8) Epoch 7, batch 150, loss[loss=0.202, simple_loss=0.2965, pruned_loss=0.0538, over 17099.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2959, pruned_loss=0.07019, over 1762678.86 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:37,278 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:47:07,162 INFO [optim.py:368] (3/8) 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 14:47:34,198 INFO [train.py:904] (3/8) Epoch 7, batch 200, loss[loss=0.1795, simple_loss=0.2715, pruned_loss=0.04374, over 16001.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.295, pruned_loss=0.06938, over 2111377.36 frames. ], batch size: 35, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:01,621 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:48:44,069 INFO [train.py:904] (3/8) Epoch 7, batch 250, loss[loss=0.2303, simple_loss=0.2963, pruned_loss=0.08216, over 16770.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2931, pruned_loss=0.06947, over 2376533.34 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:49:24,048 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3548, 4.0546, 4.3379, 4.5226, 4.6890, 4.1888, 4.4493, 4.6413], device='cuda:3'), covar=tensor([0.1099, 0.0869, 0.1205, 0.0569, 0.0419, 0.0833, 0.1084, 0.0398], device='cuda:3'), in_proj_covar=tensor([0.0430, 0.0525, 0.0659, 0.0525, 0.0399, 0.0396, 0.0420, 0.0451], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:49:24,841 INFO [optim.py:368] (3/8) 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 14:49:45,429 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2912, 4.5204, 4.7122, 2.6068, 5.0339, 5.0189, 3.5528, 3.7189], device='cuda:3'), covar=tensor([0.0623, 0.0132, 0.0116, 0.0958, 0.0052, 0.0077, 0.0310, 0.0349], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0092, 0.0080, 0.0141, 0.0069, 0.0085, 0.0117, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 14:49:52,014 INFO [train.py:904] (3/8) Epoch 7, batch 300, loss[loss=0.1872, simple_loss=0.2688, pruned_loss=0.05277, over 16860.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2904, pruned_loss=0.06838, over 2587096.06 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:50:38,696 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4265, 1.9431, 2.1764, 4.0046, 1.8935, 2.5353, 2.0119, 2.0989], device='cuda:3'), covar=tensor([0.0674, 0.2554, 0.1480, 0.0333, 0.3065, 0.1524, 0.2512, 0.2435], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0342, 0.0286, 0.0314, 0.0384, 0.0364, 0.0309, 0.0404], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:50:42,862 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7890, 4.0078, 4.1578, 3.2157, 3.9784, 3.8422, 3.9405, 2.0893], device='cuda:3'), covar=tensor([0.0322, 0.0049, 0.0058, 0.0218, 0.0050, 0.0105, 0.0066, 0.0418], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0060, 0.0060, 0.0116, 0.0062, 0.0075, 0.0065, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 14:50:52,524 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7319, 2.6637, 2.2618, 2.3520, 3.0122, 2.9581, 3.6913, 3.3075], device='cuda:3'), covar=tensor([0.0036, 0.0205, 0.0287, 0.0268, 0.0148, 0.0209, 0.0096, 0.0125], device='cuda:3'), in_proj_covar=tensor([0.0093, 0.0171, 0.0172, 0.0170, 0.0166, 0.0173, 0.0156, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:50:59,333 INFO [train.py:904] (3/8) Epoch 7, batch 350, loss[loss=0.1792, simple_loss=0.2618, pruned_loss=0.04825, over 16926.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2875, pruned_loss=0.06719, over 2756431.70 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:51:26,278 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 14:51:41,730 INFO [optim.py:368] (3/8) 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 14:51:54,176 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 14:51:55,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7826, 3.6060, 3.8206, 3.9421, 4.0607, 3.6272, 3.8513, 4.0473], device='cuda:3'), covar=tensor([0.1008, 0.0853, 0.1034, 0.0561, 0.0492, 0.1452, 0.1550, 0.0505], device='cuda:3'), in_proj_covar=tensor([0.0443, 0.0540, 0.0678, 0.0539, 0.0411, 0.0405, 0.0430, 0.0465], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 14:52:08,571 INFO [train.py:904] (3/8) Epoch 7, batch 400, loss[loss=0.2229, simple_loss=0.2948, pruned_loss=0.0755, over 15431.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2863, pruned_loss=0.06634, over 2877140.74 frames. ], batch size: 191, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:52:09,546 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 14:52:12,000 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3288, 5.2518, 5.1118, 4.8553, 4.6125, 5.1082, 5.1695, 4.7651], device='cuda:3'), covar=tensor([0.0491, 0.0294, 0.0244, 0.0211, 0.1081, 0.0359, 0.0238, 0.0523], device='cuda:3'), in_proj_covar=tensor([0.0206, 0.0231, 0.0241, 0.0214, 0.0277, 0.0247, 0.0171, 0.0279], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 14:53:17,203 INFO [train.py:904] (3/8) Epoch 7, batch 450, loss[loss=0.1773, simple_loss=0.2694, pruned_loss=0.04253, over 17194.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2837, pruned_loss=0.06411, over 2979932.20 frames. ], batch size: 46, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:48,641 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:54:00,435 INFO [optim.py:368] (3/8) 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 14:54:27,747 INFO [train.py:904] (3/8) Epoch 7, batch 500, loss[loss=0.224, simple_loss=0.2883, pruned_loss=0.07992, over 16719.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2817, pruned_loss=0.06276, over 3053018.09 frames. ], batch size: 134, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:54:47,690 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 7, batch 550, loss[loss=0.2245, simple_loss=0.2888, pruned_loss=0.08008, over 16741.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2805, pruned_loss=0.06187, over 3109866.19 frames. ], batch size: 134, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:17,396 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.888e+02 3.437e+02 4.263e+02 9.933e+02, threshold=6.875e+02, percent-clipped=6.0 2023-04-28 14:56:43,946 INFO [train.py:904] (3/8) Epoch 7, batch 600, loss[loss=0.1853, simple_loss=0.2559, pruned_loss=0.05734, over 16891.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.28, pruned_loss=0.06225, over 3160791.15 frames. ], batch size: 90, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:49,253 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-04-28 14:56:52,464 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 14:57:12,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7761, 3.0886, 2.6257, 5.0694, 4.4391, 4.6495, 1.3975, 3.5025], device='cuda:3'), covar=tensor([0.1246, 0.0598, 0.1149, 0.0103, 0.0218, 0.0276, 0.1464, 0.0588], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0145, 0.0168, 0.0098, 0.0181, 0.0194, 0.0164, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 14:57:53,637 INFO [train.py:904] (3/8) Epoch 7, batch 650, loss[loss=0.1747, simple_loss=0.2614, pruned_loss=0.04395, over 16798.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2786, pruned_loss=0.06241, over 3196210.13 frames. ], batch size: 42, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:58:35,762 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 700, loss[loss=0.2428, simple_loss=0.3027, pruned_loss=0.09142, over 12162.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2786, pruned_loss=0.06204, over 3213920.54 frames. ], batch size: 246, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 15:00:08,121 INFO [train.py:904] (3/8) Epoch 7, batch 750, loss[loss=0.2333, simple_loss=0.293, pruned_loss=0.08673, over 16875.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2792, pruned_loss=0.06248, over 3236216.66 frames. ], batch size: 116, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 15:00:39,814 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1348, 4.5868, 3.5380, 2.7694, 3.2788, 2.7001, 4.7919, 4.2853], device='cuda:3'), covar=tensor([0.2059, 0.0517, 0.1149, 0.1620, 0.2106, 0.1512, 0.0313, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0255, 0.0273, 0.0255, 0.0276, 0.0212, 0.0251, 0.0272], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:00:48,071 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9058, 3.4728, 3.1090, 5.2071, 4.6978, 4.9064, 1.7598, 3.6839], device='cuda:3'), covar=tensor([0.1241, 0.0511, 0.0913, 0.0090, 0.0224, 0.0222, 0.1313, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0146, 0.0168, 0.0101, 0.0185, 0.0196, 0.0165, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 15:00:51,555 INFO [optim.py:368] (3/8) 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 15:01:16,765 INFO [train.py:904] (3/8) Epoch 7, batch 800, loss[loss=0.2252, simple_loss=0.2872, pruned_loss=0.08154, over 16749.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2784, pruned_loss=0.06227, over 3256456.45 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:01:37,408 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:57,025 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:13,230 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 850, loss[loss=0.1927, simple_loss=0.2581, pruned_loss=0.06366, over 16254.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2774, pruned_loss=0.06107, over 3280258.12 frames. ], batch size: 165, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:02:40,427 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2656, 4.2525, 4.2261, 3.3330, 4.1694, 1.7129, 3.9096, 3.9332], device='cuda:3'), covar=tensor([0.0101, 0.0079, 0.0126, 0.0433, 0.0094, 0.2094, 0.0131, 0.0216], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0091, 0.0144, 0.0136, 0.0108, 0.0160, 0.0126, 0.0136], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:02:42,573 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:02,200 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0125, 5.4575, 5.5059, 5.4159, 5.4097, 6.0297, 5.6229, 5.3832], device='cuda:3'), covar=tensor([0.0777, 0.1599, 0.1502, 0.1793, 0.2662, 0.0896, 0.1046, 0.2127], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0443, 0.0444, 0.0377, 0.0504, 0.0472, 0.0359, 0.0507], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:03:05,349 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-28 15:03:06,806 INFO [optim.py:368] (3/8) 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 15:03:13,364 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0490, 5.4682, 5.5052, 5.4306, 5.4996, 6.0761, 5.6669, 5.3801], device='cuda:3'), covar=tensor([0.0721, 0.1675, 0.1649, 0.1959, 0.2934, 0.0965, 0.1137, 0.2404], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0443, 0.0444, 0.0377, 0.0505, 0.0472, 0.0358, 0.0508], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:03:19,435 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:33,690 INFO [train.py:904] (3/8) Epoch 7, batch 900, loss[loss=0.2284, simple_loss=0.2914, pruned_loss=0.08267, over 16663.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2759, pruned_loss=0.06006, over 3290347.43 frames. ], batch size: 134, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:03:36,390 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:41,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0961, 3.9042, 4.1362, 4.3451, 4.4224, 3.9869, 4.1407, 4.4252], device='cuda:3'), covar=tensor([0.1082, 0.0888, 0.1204, 0.0551, 0.0555, 0.1219, 0.1605, 0.0513], device='cuda:3'), in_proj_covar=tensor([0.0462, 0.0565, 0.0710, 0.0566, 0.0431, 0.0425, 0.0452, 0.0484], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:04:00,414 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4731, 4.4284, 4.3426, 4.2066, 3.9021, 4.3672, 4.2915, 4.0646], device='cuda:3'), covar=tensor([0.0539, 0.0385, 0.0238, 0.0216, 0.0890, 0.0372, 0.0489, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0216, 0.0242, 0.0253, 0.0223, 0.0290, 0.0258, 0.0178, 0.0289], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:04:40,613 INFO [train.py:904] (3/8) Epoch 7, batch 950, loss[loss=0.1997, simple_loss=0.2875, pruned_loss=0.05593, over 17084.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2757, pruned_loss=0.06032, over 3295479.53 frames. ], batch size: 55, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:20,755 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 1000, loss[loss=0.1929, simple_loss=0.2776, pruned_loss=0.05415, over 17174.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.275, pruned_loss=0.06042, over 3289067.88 frames. ], batch size: 46, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:56,999 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3415, 4.1329, 4.3053, 4.5431, 4.6523, 4.2079, 4.3876, 4.6069], device='cuda:3'), covar=tensor([0.1107, 0.0980, 0.1213, 0.0614, 0.0542, 0.1005, 0.1572, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0467, 0.0569, 0.0717, 0.0571, 0.0435, 0.0429, 0.0457, 0.0491], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:06:05,354 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:06:08,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1392, 5.1091, 4.8625, 4.2341, 4.9647, 1.8412, 4.7441, 4.9263], device='cuda:3'), covar=tensor([0.0056, 0.0044, 0.0116, 0.0336, 0.0060, 0.1962, 0.0092, 0.0127], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0092, 0.0145, 0.0137, 0.0109, 0.0159, 0.0127, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:06:40,642 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8828, 4.3511, 3.2234, 2.4814, 3.0980, 2.4156, 4.5920, 4.0502], device='cuda:3'), covar=tensor([0.2242, 0.0607, 0.1252, 0.1735, 0.2101, 0.1577, 0.0342, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0260, 0.0276, 0.0257, 0.0281, 0.0214, 0.0254, 0.0277], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:06:49,740 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 15:06:56,255 INFO [train.py:904] (3/8) Epoch 7, batch 1050, loss[loss=0.1978, simple_loss=0.2644, pruned_loss=0.06565, over 16910.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.275, pruned_loss=0.06058, over 3303705.78 frames. ], batch size: 102, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:07:28,141 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:07:39,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3105, 4.2952, 4.1688, 3.9808, 3.8132, 4.2415, 3.9927, 3.9993], device='cuda:3'), covar=tensor([0.0455, 0.0349, 0.0225, 0.0193, 0.0742, 0.0314, 0.0620, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0217, 0.0243, 0.0255, 0.0226, 0.0293, 0.0257, 0.0179, 0.0290], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:08:07,086 INFO [train.py:904] (3/8) Epoch 7, batch 1100, loss[loss=0.1813, simple_loss=0.2561, pruned_loss=0.0532, over 16458.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2746, pruned_loss=0.06036, over 3308562.47 frames. ], batch size: 75, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:08:34,909 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7554, 4.0918, 2.0898, 4.3173, 2.8127, 4.3998, 2.2944, 3.0999], device='cuda:3'), covar=tensor([0.0173, 0.0254, 0.1669, 0.0132, 0.0814, 0.0382, 0.1465, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0162, 0.0181, 0.0095, 0.0161, 0.0200, 0.0191, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 15:08:55,813 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6354, 2.4264, 1.9469, 2.0608, 2.8719, 2.7454, 3.0255, 2.9523], device='cuda:3'), covar=tensor([0.0094, 0.0218, 0.0291, 0.0278, 0.0122, 0.0165, 0.0136, 0.0136], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0174, 0.0173, 0.0173, 0.0170, 0.0176, 0.0165, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:09:15,542 INFO [train.py:904] (3/8) Epoch 7, batch 1150, loss[loss=0.203, simple_loss=0.2669, pruned_loss=0.06959, over 16705.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2737, pruned_loss=0.05991, over 3306283.79 frames. ], batch size: 134, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:57,383 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.657e+02 3.127e+02 3.921e+02 1.096e+03, threshold=6.253e+02, percent-clipped=5.0 2023-04-28 15:10:04,100 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:16,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3442, 5.7198, 5.4182, 5.5196, 4.9987, 4.9259, 5.1702, 5.7842], device='cuda:3'), covar=tensor([0.0807, 0.0811, 0.0984, 0.0519, 0.0762, 0.0640, 0.0800, 0.0885], device='cuda:3'), in_proj_covar=tensor([0.0447, 0.0579, 0.0485, 0.0386, 0.0361, 0.0376, 0.0479, 0.0425], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:10:18,763 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 1200, loss[loss=0.1992, simple_loss=0.2649, pruned_loss=0.06678, over 16747.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2728, pruned_loss=0.05924, over 3312390.10 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:10:51,414 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:11:30,498 INFO [train.py:904] (3/8) Epoch 7, batch 1250, loss[loss=0.1812, simple_loss=0.2667, pruned_loss=0.0479, over 17225.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2728, pruned_loss=0.05978, over 3325104.13 frames. ], batch size: 44, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:11:35,792 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 15:12:13,502 INFO [optim.py:368] (3/8) 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 15:12:13,953 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:12:39,183 INFO [train.py:904] (3/8) Epoch 7, batch 1300, loss[loss=0.1877, simple_loss=0.2755, pruned_loss=0.05, over 17124.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2725, pruned_loss=0.05956, over 3315728.31 frames. ], batch size: 47, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:13:30,617 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0030, 3.6901, 2.9696, 5.2790, 4.5117, 4.8105, 1.6000, 3.5337], device='cuda:3'), covar=tensor([0.1218, 0.0497, 0.1059, 0.0097, 0.0288, 0.0325, 0.1452, 0.0620], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0145, 0.0169, 0.0103, 0.0191, 0.0197, 0.0165, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 15:13:35,750 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9444, 1.6797, 2.2393, 2.7842, 2.7386, 3.0682, 1.8059, 2.9919], device='cuda:3'), covar=tensor([0.0103, 0.0289, 0.0189, 0.0164, 0.0131, 0.0097, 0.0275, 0.0062], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0154, 0.0138, 0.0139, 0.0143, 0.0101, 0.0147, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 15:13:49,052 INFO [train.py:904] (3/8) Epoch 7, batch 1350, loss[loss=0.1838, simple_loss=0.2793, pruned_loss=0.04414, over 16737.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2729, pruned_loss=0.05974, over 3313766.21 frames. ], batch size: 57, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:14:15,363 INFO [zipformer.py:625] (3/8) 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] (3/8) attn_weights_entropy = tensor([5.3015, 5.2022, 5.1017, 4.8331, 4.5784, 5.0765, 5.0960, 4.7351], device='cuda:3'), covar=tensor([0.0478, 0.0326, 0.0221, 0.0198, 0.1054, 0.0313, 0.0235, 0.0553], device='cuda:3'), in_proj_covar=tensor([0.0216, 0.0247, 0.0256, 0.0229, 0.0293, 0.0257, 0.0179, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:14:19,007 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:21,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0127, 4.4798, 2.0850, 4.7866, 2.9536, 4.7954, 2.4404, 3.3804], device='cuda:3'), covar=tensor([0.0159, 0.0180, 0.1583, 0.0061, 0.0691, 0.0271, 0.1362, 0.0563], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0163, 0.0182, 0.0098, 0.0161, 0.0203, 0.0191, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 15:14:31,591 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 1400, loss[loss=0.1654, simple_loss=0.2501, pruned_loss=0.0404, over 17213.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.273, pruned_loss=0.05916, over 3310352.27 frames. ], batch size: 44, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:15:42,010 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:15:56,506 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 15:15:59,784 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1701, 1.9934, 2.5368, 2.9763, 2.8448, 3.2848, 2.3382, 3.3751], device='cuda:3'), covar=tensor([0.0119, 0.0251, 0.0186, 0.0158, 0.0143, 0.0097, 0.0231, 0.0071], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0155, 0.0139, 0.0139, 0.0143, 0.0102, 0.0148, 0.0090], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 15:16:06,310 INFO [train.py:904] (3/8) Epoch 7, batch 1450, loss[loss=0.1815, simple_loss=0.2531, pruned_loss=0.05491, over 15610.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2715, pruned_loss=0.05902, over 3299912.97 frames. ], batch size: 190, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:16:47,643 INFO [optim.py:368] (3/8) 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 15:16:55,997 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:16:59,984 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:10,986 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:14,668 INFO [train.py:904] (3/8) Epoch 7, batch 1500, loss[loss=0.1972, simple_loss=0.2738, pruned_loss=0.06024, over 16116.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2713, pruned_loss=0.05884, over 3304209.06 frames. ], batch size: 36, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:17:45,675 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0874, 3.6789, 3.1465, 1.9875, 2.8348, 2.3222, 3.5559, 3.6085], device='cuda:3'), covar=tensor([0.0312, 0.0562, 0.0586, 0.1506, 0.0708, 0.0940, 0.0582, 0.0778], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0136, 0.0156, 0.0140, 0.0133, 0.0124, 0.0138, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 15:18:01,110 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:16,546 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:23,648 INFO [train.py:904] (3/8) Epoch 7, batch 1550, loss[loss=0.2203, simple_loss=0.2881, pruned_loss=0.07629, over 16206.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2738, pruned_loss=0.06106, over 3313816.50 frames. ], batch size: 165, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:18:24,106 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:59,838 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:19:06,018 INFO [optim.py:368] (3/8) 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 15:19:32,093 INFO [train.py:904] (3/8) Epoch 7, batch 1600, loss[loss=0.2003, simple_loss=0.2857, pruned_loss=0.05741, over 16436.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2759, pruned_loss=0.06122, over 3319046.89 frames. ], batch size: 68, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:20:31,616 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:20:40,967 INFO [train.py:904] (3/8) Epoch 7, batch 1650, loss[loss=0.1887, simple_loss=0.2762, pruned_loss=0.05061, over 17205.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2768, pruned_loss=0.06114, over 3311488.60 frames. ], batch size: 46, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:04,835 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:21:25,442 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:49,315 INFO [train.py:904] (3/8) Epoch 7, batch 1700, loss[loss=0.2206, simple_loss=0.2852, pruned_loss=0.07796, over 16886.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2793, pruned_loss=0.06196, over 3305250.05 frames. ], batch size: 96, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:55,169 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:12,978 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:27,850 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:51,344 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:58,544 INFO [train.py:904] (3/8) Epoch 7, batch 1750, loss[loss=0.2272, simple_loss=0.2948, pruned_loss=0.07983, over 16674.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2809, pruned_loss=0.06218, over 3305875.57 frames. ], batch size: 89, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:23:18,086 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:23:39,695 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5816, 3.8164, 4.0830, 2.0911, 4.2774, 4.2203, 3.2159, 3.0811], device='cuda:3'), covar=tensor([0.0716, 0.0130, 0.0127, 0.1025, 0.0049, 0.0090, 0.0310, 0.0370], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0096, 0.0085, 0.0140, 0.0071, 0.0091, 0.0120, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 15:23:40,825 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1338, 5.0122, 4.9252, 4.6420, 4.4447, 4.9097, 4.9971, 4.6027], device='cuda:3'), covar=tensor([0.0450, 0.0330, 0.0240, 0.0256, 0.1017, 0.0359, 0.0268, 0.0501], device='cuda:3'), in_proj_covar=tensor([0.0217, 0.0253, 0.0257, 0.0232, 0.0295, 0.0262, 0.0182, 0.0297], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:23:41,546 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.853e+02 3.319e+02 4.250e+02 7.879e+02, threshold=6.638e+02, percent-clipped=1.0 2023-04-28 15:24:08,263 INFO [train.py:904] (3/8) Epoch 7, batch 1800, loss[loss=0.2315, simple_loss=0.307, pruned_loss=0.078, over 15603.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2814, pruned_loss=0.06186, over 3310986.58 frames. ], batch size: 190, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:24:26,694 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 15:24:27,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2422, 3.3592, 3.5092, 1.8247, 3.6970, 3.6754, 2.8768, 2.7961], device='cuda:3'), covar=tensor([0.0689, 0.0127, 0.0125, 0.0989, 0.0061, 0.0098, 0.0371, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0096, 0.0085, 0.0140, 0.0072, 0.0091, 0.0120, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 15:24:42,429 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:05,649 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8802, 4.8266, 5.3627, 5.3448, 5.3687, 4.9520, 4.9544, 4.7737], device='cuda:3'), covar=tensor([0.0238, 0.0397, 0.0332, 0.0372, 0.0331, 0.0290, 0.0790, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0288, 0.0296, 0.0277, 0.0335, 0.0309, 0.0414, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 15:25:10,911 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:17,227 INFO [train.py:904] (3/8) Epoch 7, batch 1850, loss[loss=0.209, simple_loss=0.3015, pruned_loss=0.05827, over 16544.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2826, pruned_loss=0.06232, over 3315986.80 frames. ], batch size: 68, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:25:45,832 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:52,381 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.663e+02 2.724e+02 3.242e+02 3.816e+02 7.402e+02, threshold=6.484e+02, percent-clipped=2.0 2023-04-28 15:26:26,217 INFO [train.py:904] (3/8) Epoch 7, batch 1900, loss[loss=0.1923, simple_loss=0.2617, pruned_loss=0.06149, over 16746.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2823, pruned_loss=0.06227, over 3320826.75 frames. ], batch size: 124, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:26:38,982 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9655, 4.8936, 4.8045, 4.6323, 4.4068, 4.8467, 4.8153, 4.4786], device='cuda:3'), covar=tensor([0.0416, 0.0302, 0.0194, 0.0204, 0.0765, 0.0280, 0.0249, 0.0485], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0251, 0.0255, 0.0229, 0.0291, 0.0258, 0.0181, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:27:00,822 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:11,642 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:36,783 INFO [train.py:904] (3/8) Epoch 7, batch 1950, loss[loss=0.1516, simple_loss=0.23, pruned_loss=0.03656, over 16313.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2817, pruned_loss=0.06139, over 3329473.06 frames. ], batch size: 36, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:27:43,323 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 15:28:19,259 INFO [optim.py:368] (3/8) 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 15:28:44,381 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 2000, loss[loss=0.1995, simple_loss=0.2859, pruned_loss=0.05654, over 16624.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2814, pruned_loss=0.06119, over 3333135.86 frames. ], batch size: 57, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:29:24,401 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:39,596 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:55,689 INFO [train.py:904] (3/8) Epoch 7, batch 2050, loss[loss=0.2377, simple_loss=0.3002, pruned_loss=0.08758, over 16778.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2813, pruned_loss=0.06177, over 3325225.91 frames. ], batch size: 124, lr: 9.99e-03, grad_scale: 16.0 2023-04-28 15:30:30,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4907, 3.4390, 2.6333, 2.1546, 2.4181, 2.1147, 3.3707, 3.2327], device='cuda:3'), covar=tensor([0.2032, 0.0577, 0.1207, 0.1812, 0.1970, 0.1538, 0.0447, 0.0937], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0258, 0.0273, 0.0257, 0.0285, 0.0210, 0.0253, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:30:31,319 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:30:39,208 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.104e+02 2.968e+02 3.465e+02 4.226e+02 9.943e+02, threshold=6.931e+02, percent-clipped=3.0 2023-04-28 15:31:05,148 INFO [train.py:904] (3/8) Epoch 7, batch 2100, loss[loss=0.18, simple_loss=0.2603, pruned_loss=0.04986, over 15784.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2825, pruned_loss=0.06301, over 3314921.02 frames. ], batch size: 35, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:31:33,159 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:09,401 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:11,798 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8024, 4.7616, 4.6578, 4.4910, 4.2705, 4.7490, 4.6088, 4.3996], device='cuda:3'), covar=tensor([0.0443, 0.0348, 0.0210, 0.0220, 0.0827, 0.0288, 0.0441, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0256, 0.0260, 0.0234, 0.0298, 0.0262, 0.0185, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:32:16,868 INFO [train.py:904] (3/8) Epoch 7, batch 2150, loss[loss=0.2015, simple_loss=0.2894, pruned_loss=0.05677, over 16577.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2827, pruned_loss=0.06288, over 3320478.84 frames. ], batch size: 68, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:32:32,985 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3977, 4.3491, 4.2644, 3.8184, 4.3220, 1.7949, 4.1172, 4.1012], device='cuda:3'), covar=tensor([0.0075, 0.0062, 0.0110, 0.0202, 0.0060, 0.1739, 0.0078, 0.0126], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0095, 0.0148, 0.0142, 0.0113, 0.0158, 0.0129, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:32:34,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8003, 5.0026, 5.0980, 5.0271, 4.9862, 5.6053, 5.2561, 4.9351], device='cuda:3'), covar=tensor([0.1041, 0.1608, 0.1636, 0.1752, 0.2465, 0.0909, 0.1073, 0.2126], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0458, 0.0454, 0.0385, 0.0520, 0.0488, 0.0371, 0.0521], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:32:58,764 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2154, 1.9561, 2.0965, 3.7289, 1.9304, 2.4513, 2.0957, 2.0780], device='cuda:3'), covar=tensor([0.0802, 0.2514, 0.1522, 0.0366, 0.3016, 0.1593, 0.2384, 0.2431], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0350, 0.0290, 0.0319, 0.0385, 0.0381, 0.0317, 0.0417], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:33:00,561 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 7, batch 2200, loss[loss=0.2291, simple_loss=0.295, pruned_loss=0.08159, over 16650.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2839, pruned_loss=0.06376, over 3319744.48 frames. ], batch size: 134, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:34:03,305 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 2250, loss[loss=0.2158, simple_loss=0.2883, pruned_loss=0.07162, over 16281.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2851, pruned_loss=0.0647, over 3316938.01 frames. ], batch size: 165, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:35:18,521 INFO [optim.py:368] (3/8) 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 15:35:24,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8404, 4.1250, 4.2885, 3.2020, 3.7065, 4.1938, 4.0244, 2.6701], device='cuda:3'), covar=tensor([0.0285, 0.0025, 0.0024, 0.0189, 0.0057, 0.0044, 0.0034, 0.0288], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0063, 0.0062, 0.0114, 0.0065, 0.0076, 0.0066, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:35:42,254 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 2300, loss[loss=0.1787, simple_loss=0.2657, pruned_loss=0.04584, over 17187.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2853, pruned_loss=0.06404, over 3317424.77 frames. ], batch size: 46, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:36:10,184 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:36,756 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:41,378 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-04-28 15:36:48,217 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:51,523 INFO [train.py:904] (3/8) Epoch 7, batch 2350, loss[loss=0.2152, simple_loss=0.2866, pruned_loss=0.07189, over 16759.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2854, pruned_loss=0.06475, over 3310941.71 frames. ], batch size: 89, lr: 9.96e-03, grad_scale: 4.0 2023-04-28 15:37:34,013 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9877, 2.4187, 1.9625, 2.1464, 2.8802, 2.6604, 3.1868, 2.9984], device='cuda:3'), covar=tensor([0.0087, 0.0183, 0.0238, 0.0229, 0.0103, 0.0164, 0.0119, 0.0119], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0177, 0.0174, 0.0176, 0.0174, 0.0180, 0.0174, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:37:34,016 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:38:00,172 INFO [train.py:904] (3/8) Epoch 7, batch 2400, loss[loss=0.2103, simple_loss=0.2976, pruned_loss=0.06146, over 17118.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2859, pruned_loss=0.06442, over 3312004.09 frames. ], batch size: 49, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:38:27,624 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:38:59,485 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2274, 5.2041, 4.9611, 4.4433, 5.0716, 1.8974, 4.7920, 5.0192], device='cuda:3'), covar=tensor([0.0052, 0.0039, 0.0105, 0.0251, 0.0059, 0.1772, 0.0075, 0.0106], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0095, 0.0147, 0.0141, 0.0113, 0.0156, 0.0129, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:39:10,358 INFO [train.py:904] (3/8) Epoch 7, batch 2450, loss[loss=0.2238, simple_loss=0.2983, pruned_loss=0.07463, over 16710.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2859, pruned_loss=0.06395, over 3310065.28 frames. ], batch size: 134, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:39:34,203 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:43,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6364, 4.1442, 4.3243, 1.8899, 4.6698, 4.5724, 3.0234, 3.4966], device='cuda:3'), covar=tensor([0.0755, 0.0132, 0.0205, 0.1212, 0.0044, 0.0092, 0.0411, 0.0350], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0098, 0.0085, 0.0141, 0.0072, 0.0091, 0.0122, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 15:39:55,098 INFO [optim.py:368] (3/8) 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 15:40:19,248 INFO [train.py:904] (3/8) Epoch 7, batch 2500, loss[loss=0.2165, simple_loss=0.312, pruned_loss=0.06049, over 17021.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2864, pruned_loss=0.06378, over 3312975.95 frames. ], batch size: 50, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:40:20,849 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9908, 3.4957, 3.0638, 1.8565, 2.7266, 2.3020, 3.4003, 3.3997], device='cuda:3'), covar=tensor([0.0237, 0.0551, 0.0559, 0.1486, 0.0693, 0.0883, 0.0561, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0139, 0.0157, 0.0140, 0.0134, 0.0124, 0.0140, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 15:40:44,234 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:46,550 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:55,005 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:41:26,925 INFO [train.py:904] (3/8) Epoch 7, batch 2550, loss[loss=0.198, simple_loss=0.2899, pruned_loss=0.05309, over 17254.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2865, pruned_loss=0.06352, over 3316324.45 frames. ], batch size: 52, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:41:28,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7407, 4.0329, 4.2210, 2.0756, 4.5150, 4.4418, 3.0584, 3.4922], device='cuda:3'), covar=tensor([0.0686, 0.0129, 0.0140, 0.1023, 0.0044, 0.0103, 0.0374, 0.0314], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0097, 0.0084, 0.0139, 0.0072, 0.0090, 0.0121, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 15:41:30,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 15:42:01,822 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:08,256 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:11,026 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:42:35,897 INFO [train.py:904] (3/8) Epoch 7, batch 2600, loss[loss=0.2034, simple_loss=0.2896, pruned_loss=0.05859, over 17107.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2859, pruned_loss=0.06303, over 3319204.95 frames. ], batch size: 49, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:42:51,562 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2604, 4.2257, 4.6960, 4.6980, 4.7040, 4.3200, 4.3412, 4.2074], device='cuda:3'), covar=tensor([0.0283, 0.0467, 0.0297, 0.0334, 0.0379, 0.0306, 0.0824, 0.0452], device='cuda:3'), in_proj_covar=tensor([0.0282, 0.0287, 0.0289, 0.0273, 0.0329, 0.0305, 0.0410, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 15:43:42,047 INFO [train.py:904] (3/8) Epoch 7, batch 2650, loss[loss=0.225, simple_loss=0.2983, pruned_loss=0.0758, over 16687.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2859, pruned_loss=0.0626, over 3319978.15 frames. ], batch size: 134, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:44:06,175 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 15:44:27,966 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 2700, loss[loss=0.2189, simple_loss=0.3086, pruned_loss=0.06455, over 17065.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2859, pruned_loss=0.06176, over 3327814.41 frames. ], batch size: 53, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:45:28,060 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:47,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4999, 5.8710, 5.5976, 5.6916, 5.1895, 4.9023, 5.2872, 5.9110], device='cuda:3'), covar=tensor([0.0962, 0.0768, 0.0953, 0.0554, 0.0747, 0.0691, 0.0838, 0.0912], device='cuda:3'), in_proj_covar=tensor([0.0460, 0.0595, 0.0499, 0.0393, 0.0374, 0.0385, 0.0493, 0.0439], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:45:52,015 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 2750, loss[loss=0.1765, simple_loss=0.2524, pruned_loss=0.05025, over 16990.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2856, pruned_loss=0.06113, over 3323457.04 frames. ], batch size: 41, lr: 9.93e-03, grad_scale: 4.0 2023-04-28 15:46:09,307 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8466, 5.1119, 5.2712, 5.1021, 5.1354, 5.7072, 5.2771, 5.0118], device='cuda:3'), covar=tensor([0.0932, 0.1667, 0.1480, 0.1681, 0.2378, 0.0989, 0.1207, 0.2398], device='cuda:3'), in_proj_covar=tensor([0.0321, 0.0463, 0.0459, 0.0388, 0.0526, 0.0492, 0.0370, 0.0528], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 15:46:45,843 INFO [optim.py:368] (3/8) 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 15:47:08,528 INFO [train.py:904] (3/8) Epoch 7, batch 2800, loss[loss=0.2115, simple_loss=0.2772, pruned_loss=0.07284, over 16918.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2848, pruned_loss=0.06055, over 3323116.53 frames. ], batch size: 109, lr: 9.93e-03, grad_scale: 8.0 2023-04-28 15:48:12,843 INFO [train.py:904] (3/8) Epoch 7, batch 2850, loss[loss=0.1775, simple_loss=0.2541, pruned_loss=0.05048, over 17010.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2848, pruned_loss=0.06128, over 3325343.44 frames. ], batch size: 41, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:48:47,329 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.714e+02 3.034e+02 3.568e+02 4.214e+02 8.361e+02, threshold=7.135e+02, percent-clipped=4.0 2023-04-28 15:49:23,198 INFO [train.py:904] (3/8) Epoch 7, batch 2900, loss[loss=0.1747, simple_loss=0.26, pruned_loss=0.0447, over 17226.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2837, pruned_loss=0.0614, over 3329174.38 frames. ], batch size: 44, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:49:35,848 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:49:35,925 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9260, 1.8034, 2.2909, 2.9143, 2.8918, 2.8211, 1.8215, 3.0472], device='cuda:3'), covar=tensor([0.0096, 0.0257, 0.0174, 0.0120, 0.0097, 0.0109, 0.0246, 0.0059], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0159, 0.0141, 0.0142, 0.0147, 0.0105, 0.0149, 0.0095], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 15:50:13,416 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 15:50:22,290 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:32,279 INFO [train.py:904] (3/8) Epoch 7, batch 2950, loss[loss=0.1718, simple_loss=0.2489, pruned_loss=0.04733, over 15982.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2829, pruned_loss=0.06167, over 3327529.19 frames. ], batch size: 35, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:50:58,315 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:05,327 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:09,212 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 15:51:17,102 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.004e+02 3.573e+02 4.238e+02 9.088e+02, threshold=7.147e+02, percent-clipped=1.0 2023-04-28 15:51:38,107 INFO [train.py:904] (3/8) Epoch 7, batch 3000, loss[loss=0.1934, simple_loss=0.2741, pruned_loss=0.05634, over 17233.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2834, pruned_loss=0.06203, over 3323044.45 frames. ], batch size: 45, lr: 9.91e-03, grad_scale: 8.0 2023-04-28 15:51:38,107 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 15:51:46,842 INFO [train.py:938] (3/8) Epoch 7, validation: loss=0.1489, simple_loss=0.2553, pruned_loss=0.02124, over 944034.00 frames. 2023-04-28 15:51:46,843 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 15:51:54,287 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:06,710 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:15,567 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-28 15:52:19,936 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:19,950 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:21,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5649, 4.5479, 5.0086, 5.0238, 5.0208, 4.6548, 4.6582, 4.4297], device='cuda:3'), covar=tensor([0.0285, 0.0568, 0.0279, 0.0325, 0.0435, 0.0310, 0.0815, 0.0395], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0297, 0.0295, 0.0282, 0.0345, 0.0313, 0.0425, 0.0254], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 15:52:42,113 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:56,535 INFO [train.py:904] (3/8) Epoch 7, batch 3050, loss[loss=0.2788, simple_loss=0.3342, pruned_loss=0.1116, over 12125.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2831, pruned_loss=0.06191, over 3320655.97 frames. ], batch size: 246, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:53:30,231 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:53:43,706 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.789e+02 3.428e+02 3.925e+02 6.614e+02, threshold=6.856e+02, percent-clipped=0.0 2023-04-28 15:54:06,475 INFO [train.py:904] (3/8) Epoch 7, batch 3100, loss[loss=0.2024, simple_loss=0.2859, pruned_loss=0.05942, over 17054.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2827, pruned_loss=0.06153, over 3331156.08 frames. ], batch size: 55, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:54:12,686 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 2023-04-28 15:54:36,555 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:54:36,836 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 15:55:03,735 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8279, 4.0858, 2.2126, 4.4158, 2.8121, 4.4008, 2.4173, 3.1270], device='cuda:3'), covar=tensor([0.0153, 0.0254, 0.1393, 0.0096, 0.0710, 0.0413, 0.1242, 0.0578], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0162, 0.0178, 0.0101, 0.0161, 0.0205, 0.0189, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 15:55:16,374 INFO [train.py:904] (3/8) Epoch 7, batch 3150, loss[loss=0.1789, simple_loss=0.2543, pruned_loss=0.05176, over 17214.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2816, pruned_loss=0.06074, over 3334076.14 frames. ], batch size: 45, lr: 9.90e-03, grad_scale: 4.0 2023-04-28 15:55:49,729 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:52,131 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:00,081 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:03,681 INFO [optim.py:368] (3/8) 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 15:56:22,665 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 15:56:24,286 INFO [train.py:904] (3/8) Epoch 7, batch 3200, loss[loss=0.2081, simple_loss=0.2849, pruned_loss=0.06568, over 16439.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2804, pruned_loss=0.06041, over 3336427.93 frames. ], batch size: 146, lr: 9.90e-03, grad_scale: 8.0 2023-04-28 15:56:54,782 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:57,814 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:57:33,480 INFO [train.py:904] (3/8) Epoch 7, batch 3250, loss[loss=0.2406, simple_loss=0.3132, pruned_loss=0.08405, over 16723.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2811, pruned_loss=0.06066, over 3340239.31 frames. ], batch size: 134, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:57:53,162 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:20,556 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3046, 4.2532, 4.2121, 3.7567, 4.2450, 1.8519, 4.0027, 3.9960], device='cuda:3'), covar=tensor([0.0076, 0.0066, 0.0109, 0.0217, 0.0064, 0.1807, 0.0091, 0.0131], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0101, 0.0153, 0.0149, 0.0118, 0.0161, 0.0136, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:58:21,283 INFO [optim.py:368] (3/8) 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 15:58:42,120 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:43,054 INFO [train.py:904] (3/8) Epoch 7, batch 3300, loss[loss=0.1718, simple_loss=0.2524, pruned_loss=0.04557, over 16782.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2816, pruned_loss=0.06087, over 3334572.79 frames. ], batch size: 39, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:59:16,280 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:25,397 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:40,400 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:42,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2275, 4.9299, 5.1484, 5.4134, 5.5908, 4.8677, 5.4578, 5.5192], device='cuda:3'), covar=tensor([0.1085, 0.0893, 0.1404, 0.0548, 0.0412, 0.0627, 0.0486, 0.0409], device='cuda:3'), in_proj_covar=tensor([0.0493, 0.0604, 0.0774, 0.0615, 0.0465, 0.0473, 0.0477, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 15:59:51,973 INFO [train.py:904] (3/8) Epoch 7, batch 3350, loss[loss=0.2057, simple_loss=0.2925, pruned_loss=0.05942, over 16621.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2826, pruned_loss=0.06108, over 3337105.25 frames. ], batch size: 62, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 16:00:00,833 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0564, 4.7590, 5.0401, 5.2686, 5.4461, 4.7772, 5.3803, 5.4021], device='cuda:3'), covar=tensor([0.1106, 0.0990, 0.1361, 0.0540, 0.0434, 0.0647, 0.0408, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0494, 0.0604, 0.0774, 0.0615, 0.0465, 0.0473, 0.0476, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:00:20,544 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:22,690 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:32,379 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9082, 2.9404, 2.4526, 2.7399, 3.2127, 2.9477, 3.8478, 3.3733], device='cuda:3'), covar=tensor([0.0038, 0.0202, 0.0247, 0.0226, 0.0137, 0.0205, 0.0097, 0.0129], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0177, 0.0172, 0.0175, 0.0174, 0.0177, 0.0176, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:00:42,047 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.871e+02 3.440e+02 4.131e+02 8.469e+02, threshold=6.880e+02, percent-clipped=3.0 2023-04-28 16:00:45,938 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:49,520 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:51,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8664, 4.1132, 4.3402, 3.1791, 3.6778, 4.2669, 4.0330, 2.4545], device='cuda:3'), covar=tensor([0.0277, 0.0025, 0.0023, 0.0192, 0.0054, 0.0045, 0.0032, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0062, 0.0063, 0.0117, 0.0066, 0.0078, 0.0069, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:01:02,162 INFO [train.py:904] (3/8) Epoch 7, batch 3400, loss[loss=0.1766, simple_loss=0.2665, pruned_loss=0.04333, over 17200.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2815, pruned_loss=0.05996, over 3337286.51 frames. ], batch size: 46, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:01:09,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5547, 2.1063, 2.3903, 4.1596, 2.0477, 2.7418, 2.1677, 2.3141], device='cuda:3'), covar=tensor([0.0671, 0.2553, 0.1444, 0.0315, 0.2867, 0.1504, 0.2570, 0.2286], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0357, 0.0298, 0.0330, 0.0391, 0.0393, 0.0322, 0.0424], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:02:02,982 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:02:10,851 INFO [train.py:904] (3/8) Epoch 7, batch 3450, loss[loss=0.1805, simple_loss=0.2587, pruned_loss=0.05114, over 16848.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2806, pruned_loss=0.05984, over 3320488.25 frames. ], batch size: 42, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:02:49,999 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:03:01,780 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.744e+02 2.661e+02 3.216e+02 3.896e+02 8.084e+02, threshold=6.431e+02, percent-clipped=1.0 2023-04-28 16:03:21,642 INFO [train.py:904] (3/8) Epoch 7, batch 3500, loss[loss=0.2301, simple_loss=0.2962, pruned_loss=0.08201, over 16862.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2797, pruned_loss=0.05924, over 3328052.34 frames. ], batch size: 116, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:03:28,736 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:03:35,484 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7136, 4.7145, 4.9370, 4.7970, 4.8369, 5.4178, 5.0819, 4.7368], device='cuda:3'), covar=tensor([0.1206, 0.1890, 0.1375, 0.1832, 0.2569, 0.0918, 0.1113, 0.2403], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0455, 0.0452, 0.0383, 0.0517, 0.0485, 0.0365, 0.0519], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:04:13,300 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 16:04:30,074 INFO [train.py:904] (3/8) Epoch 7, batch 3550, loss[loss=0.1894, simple_loss=0.2702, pruned_loss=0.05425, over 16626.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2793, pruned_loss=0.05936, over 3323347.28 frames. ], batch size: 62, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:04:48,231 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:50,574 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:18,042 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 16:05:18,414 INFO [optim.py:368] (3/8) 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 16:05:28,779 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 16:05:38,479 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:39,311 INFO [train.py:904] (3/8) Epoch 7, batch 3600, loss[loss=0.1681, simple_loss=0.2476, pruned_loss=0.04426, over 16778.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2774, pruned_loss=0.05816, over 3310997.27 frames. ], batch size: 39, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:05:56,512 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:14,269 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:45,518 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 3650, loss[loss=0.1865, simple_loss=0.2565, pruned_loss=0.05823, over 16797.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.277, pruned_loss=0.05882, over 3296406.99 frames. ], batch size: 102, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:07:19,240 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:40,348 INFO [optim.py:368] (3/8) 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 16:07:42,459 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:44,503 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 3700, loss[loss=0.2531, simple_loss=0.3074, pruned_loss=0.0994, over 11321.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2763, pruned_loss=0.06111, over 3281863.23 frames. ], batch size: 248, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:08:27,963 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:09,893 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:12,312 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 3750, loss[loss=0.2143, simple_loss=0.2833, pruned_loss=0.07263, over 16466.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.277, pruned_loss=0.06276, over 3282710.67 frames. ], batch size: 68, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:09:55,073 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 2.843e+02 3.454e+02 4.284e+02 7.405e+02, threshold=6.908e+02, percent-clipped=2.0 2023-04-28 16:10:24,875 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 3800, loss[loss=0.2092, simple_loss=0.2859, pruned_loss=0.06627, over 15425.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2784, pruned_loss=0.06448, over 3287387.54 frames. ], batch size: 190, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:10:28,009 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:38,595 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:05,638 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:31,419 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:39,177 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9990, 3.7823, 4.0342, 4.1836, 4.2307, 3.8378, 3.9871, 4.2419], device='cuda:3'), covar=tensor([0.0945, 0.0758, 0.1055, 0.0473, 0.0483, 0.1367, 0.1153, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0464, 0.0568, 0.0726, 0.0579, 0.0438, 0.0441, 0.0450, 0.0497], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:11:39,992 INFO [train.py:904] (3/8) Epoch 7, batch 3850, loss[loss=0.2068, simple_loss=0.2754, pruned_loss=0.06911, over 16289.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.278, pruned_loss=0.06515, over 3274050.64 frames. ], batch size: 165, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:11:40,548 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:58,609 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:12:33,097 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 3900, loss[loss=0.1996, simple_loss=0.2726, pruned_loss=0.0633, over 17125.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2774, pruned_loss=0.0657, over 3272140.83 frames. ], batch size: 47, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:13:00,382 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:08,270 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:20,516 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:40,695 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9321, 5.0928, 4.8100, 4.7109, 4.2038, 4.9171, 4.7992, 4.5433], device='cuda:3'), covar=tensor([0.0501, 0.0297, 0.0273, 0.0238, 0.1073, 0.0362, 0.0311, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0212, 0.0251, 0.0257, 0.0230, 0.0294, 0.0259, 0.0179, 0.0292], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:13:58,697 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-28 16:14:02,900 INFO [train.py:904] (3/8) Epoch 7, batch 3950, loss[loss=0.19, simple_loss=0.2746, pruned_loss=0.05268, over 17123.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2773, pruned_loss=0.06645, over 3269263.90 frames. ], batch size: 49, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:14:19,388 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-28 16:14:34,395 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5748, 4.7708, 4.8642, 4.8316, 4.7946, 5.3301, 4.8782, 4.6748], device='cuda:3'), covar=tensor([0.1291, 0.1572, 0.1512, 0.1764, 0.2593, 0.0926, 0.1254, 0.2076], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0451, 0.0448, 0.0373, 0.0505, 0.0475, 0.0359, 0.0504], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:14:53,036 INFO [optim.py:368] (3/8) 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 16:14:53,490 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:15:12,235 INFO [train.py:904] (3/8) Epoch 7, batch 4000, loss[loss=0.2117, simple_loss=0.2811, pruned_loss=0.07114, over 16790.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2773, pruned_loss=0.06673, over 3274299.89 frames. ], batch size: 102, lr: 9.84e-03, grad_scale: 8.0 2023-04-28 16:16:02,045 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:06,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8705, 3.2841, 2.8046, 5.1518, 4.1936, 4.7288, 1.8884, 3.2556], device='cuda:3'), covar=tensor([0.1322, 0.0611, 0.1090, 0.0107, 0.0367, 0.0250, 0.1368, 0.0794], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0147, 0.0169, 0.0109, 0.0201, 0.0198, 0.0166, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 16:16:16,962 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:24,731 INFO [train.py:904] (3/8) Epoch 7, batch 4050, loss[loss=0.1992, simple_loss=0.2799, pruned_loss=0.05926, over 16943.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.277, pruned_loss=0.06524, over 3279929.62 frames. ], batch size: 90, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:16:59,742 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:17:16,150 INFO [optim.py:368] (3/8) 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 16:17:35,665 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4100, loss[loss=0.2157, simple_loss=0.2914, pruned_loss=0.06995, over 12318.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2782, pruned_loss=0.06427, over 3280339.61 frames. ], batch size: 248, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:17:39,312 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:29,694 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:18:47,166 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4150, loss[loss=0.2298, simple_loss=0.3135, pruned_loss=0.07306, over 15352.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2865, pruned_loss=0.0678, over 3259787.04 frames. ], batch size: 190, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:18:56,834 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:01,895 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:20:03,730 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:04,501 INFO [train.py:904] (3/8) Epoch 7, batch 4200, loss[loss=0.238, simple_loss=0.3198, pruned_loss=0.0781, over 16511.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2936, pruned_loss=0.07, over 3217645.96 frames. ], batch size: 68, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:20:13,288 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:24,638 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:32,664 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:16,725 INFO [train.py:904] (3/8) Epoch 7, batch 4250, loss[loss=0.2181, simple_loss=0.3052, pruned_loss=0.06551, over 17124.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2966, pruned_loss=0.06943, over 3210991.93 frames. ], batch size: 48, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:21:42,347 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:22:03,725 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8927, 2.7445, 2.6984, 1.8434, 2.5598, 2.6827, 2.6425, 1.7572], device='cuda:3'), covar=tensor([0.0284, 0.0039, 0.0048, 0.0225, 0.0060, 0.0064, 0.0048, 0.0321], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0058, 0.0059, 0.0114, 0.0064, 0.0073, 0.0065, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:22:08,030 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.671e+02 3.307e+02 3.991e+02 9.211e+02, threshold=6.614e+02, percent-clipped=3.0 2023-04-28 16:22:29,057 INFO [train.py:904] (3/8) Epoch 7, batch 4300, loss[loss=0.2009, simple_loss=0.2909, pruned_loss=0.05541, over 16418.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2977, pruned_loss=0.06858, over 3198368.09 frames. ], batch size: 35, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:23:31,786 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4350, loss[loss=0.2282, simple_loss=0.305, pruned_loss=0.07572, over 16871.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3009, pruned_loss=0.06954, over 3197857.77 frames. ], batch size: 109, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:01,869 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7241, 3.9224, 3.1932, 2.3479, 2.9677, 2.5241, 4.3133, 3.9114], device='cuda:3'), covar=tensor([0.2308, 0.0671, 0.1219, 0.1565, 0.2063, 0.1418, 0.0311, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0254, 0.0271, 0.0258, 0.0291, 0.0211, 0.0253, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:24:02,108 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 16:24:30,848 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:33,882 INFO [optim.py:368] (3/8) 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 16:24:42,764 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:54,958 INFO [train.py:904] (3/8) Epoch 7, batch 4400, loss[loss=0.2189, simple_loss=0.3035, pruned_loss=0.06711, over 16737.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3026, pruned_loss=0.07044, over 3187277.06 frames. ], batch size: 83, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:57,825 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:25:00,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4142, 4.0912, 4.0547, 2.5462, 3.5300, 3.9842, 3.7326, 2.4478], device='cuda:3'), covar=tensor([0.0315, 0.0015, 0.0020, 0.0249, 0.0045, 0.0045, 0.0030, 0.0258], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0058, 0.0059, 0.0115, 0.0064, 0.0074, 0.0066, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:25:39,690 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:25:59,688 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4450, loss[loss=0.2249, simple_loss=0.3098, pruned_loss=0.07004, over 16470.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.306, pruned_loss=0.07127, over 3196348.88 frames. ], batch size: 68, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:26:08,112 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:13,903 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 16:26:17,472 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:27:19,266 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4500, loss[loss=0.2288, simple_loss=0.308, pruned_loss=0.07482, over 16605.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.306, pruned_loss=0.07152, over 3198460.14 frames. ], batch size: 57, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:27:26,478 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:28,839 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:32,337 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:27,210 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4550, loss[loss=0.2314, simple_loss=0.2945, pruned_loss=0.08417, over 11989.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3063, pruned_loss=0.0716, over 3219101.46 frames. ], batch size: 248, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:28:38,294 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:29:21,409 INFO [optim.py:368] (3/8) 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 16:29:41,302 INFO [train.py:904] (3/8) Epoch 7, batch 4600, loss[loss=0.2532, simple_loss=0.3265, pruned_loss=0.08994, over 16721.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3073, pruned_loss=0.07165, over 3229648.61 frames. ], batch size: 124, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:14,546 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 16:30:24,215 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:30:50,581 INFO [train.py:904] (3/8) Epoch 7, batch 4650, loss[loss=0.218, simple_loss=0.3018, pruned_loss=0.06707, over 16671.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3057, pruned_loss=0.07093, over 3237764.47 frames. ], batch size: 134, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:59,686 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4853, 4.3334, 4.3613, 2.7185, 3.6274, 4.3498, 3.9668, 2.5851], device='cuda:3'), covar=tensor([0.0327, 0.0012, 0.0013, 0.0265, 0.0053, 0.0035, 0.0035, 0.0257], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0057, 0.0058, 0.0116, 0.0066, 0.0073, 0.0066, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:31:28,126 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-28 16:31:37,901 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-28 16:31:41,970 INFO [optim.py:368] (3/8) 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 16:31:49,760 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:32:03,152 INFO [train.py:904] (3/8) Epoch 7, batch 4700, loss[loss=0.212, simple_loss=0.2999, pruned_loss=0.06204, over 16394.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3028, pruned_loss=0.06973, over 3213942.21 frames. ], batch size: 146, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:32:07,233 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6244, 2.7087, 2.3484, 4.1355, 3.1451, 3.8959, 1.2899, 2.8916], device='cuda:3'), covar=tensor([0.1362, 0.0673, 0.1208, 0.0120, 0.0247, 0.0350, 0.1559, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0145, 0.0169, 0.0102, 0.0198, 0.0193, 0.0166, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 16:32:37,936 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8826, 3.4750, 3.1213, 1.9180, 2.6972, 2.1803, 3.3366, 3.3397], device='cuda:3'), covar=tensor([0.0245, 0.0562, 0.0563, 0.1600, 0.0804, 0.0865, 0.0628, 0.0784], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0134, 0.0154, 0.0140, 0.0132, 0.0124, 0.0136, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 16:32:39,018 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5510, 4.2930, 4.5969, 4.8105, 4.9101, 4.4400, 4.9223, 4.9161], device='cuda:3'), covar=tensor([0.0892, 0.0872, 0.1031, 0.0407, 0.0398, 0.0864, 0.0379, 0.0377], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0521, 0.0656, 0.0524, 0.0400, 0.0410, 0.0408, 0.0449], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:32:45,690 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:32:56,228 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:33:12,630 INFO [train.py:904] (3/8) Epoch 7, batch 4750, loss[loss=0.1873, simple_loss=0.2687, pruned_loss=0.05293, over 17217.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.298, pruned_loss=0.06746, over 3228729.26 frames. ], batch size: 45, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:33:22,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7819, 3.2596, 3.2055, 1.9319, 2.9109, 3.1484, 3.0236, 1.7565], device='cuda:3'), covar=tensor([0.0379, 0.0023, 0.0023, 0.0298, 0.0049, 0.0065, 0.0046, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0058, 0.0059, 0.0117, 0.0066, 0.0075, 0.0067, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:33:53,642 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:34:03,333 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 4800, loss[loss=0.2229, simple_loss=0.2962, pruned_loss=0.07483, over 11919.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2951, pruned_loss=0.066, over 3209976.93 frames. ], batch size: 246, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:34:36,323 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:35:21,847 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 16:35:25,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9529, 1.7323, 2.3649, 3.0207, 2.9893, 3.3179, 1.7608, 3.1701], device='cuda:3'), covar=tensor([0.0100, 0.0298, 0.0198, 0.0139, 0.0125, 0.0096, 0.0312, 0.0072], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0156, 0.0140, 0.0143, 0.0149, 0.0106, 0.0152, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 16:35:36,916 INFO [train.py:904] (3/8) Epoch 7, batch 4850, loss[loss=0.2401, simple_loss=0.3243, pruned_loss=0.07793, over 15532.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2964, pruned_loss=0.06564, over 3192655.08 frames. ], batch size: 191, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:35:45,498 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:36:07,386 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 16:36:28,268 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 16:36:28,553 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 4900, loss[loss=0.2002, simple_loss=0.2869, pruned_loss=0.05673, over 16529.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2953, pruned_loss=0.06416, over 3184507.69 frames. ], batch size: 68, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:37:10,067 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-28 16:37:44,000 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 4950, loss[loss=0.2183, simple_loss=0.3047, pruned_loss=0.06592, over 16743.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2949, pruned_loss=0.06396, over 3190074.06 frames. ], batch size: 124, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:38:50,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9857, 3.8481, 4.0575, 4.2575, 4.3322, 3.9148, 4.3125, 4.3102], device='cuda:3'), covar=tensor([0.1160, 0.0888, 0.1149, 0.0460, 0.0367, 0.1158, 0.0420, 0.0440], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0522, 0.0655, 0.0531, 0.0397, 0.0408, 0.0407, 0.0453], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:38:53,490 INFO [optim.py:368] (3/8) 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 16:38:53,919 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:39:08,380 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:39:12,682 INFO [train.py:904] (3/8) Epoch 7, batch 5000, loss[loss=0.2151, simple_loss=0.3142, pruned_loss=0.05796, over 16761.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2976, pruned_loss=0.0644, over 3200718.84 frames. ], batch size: 134, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:40:09,100 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:40:14,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1699, 4.3925, 4.6404, 4.6004, 4.5997, 4.2997, 3.8718, 4.1402], device='cuda:3'), covar=tensor([0.0472, 0.0635, 0.0461, 0.0652, 0.0674, 0.0472, 0.1425, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0265, 0.0265, 0.0262, 0.0316, 0.0286, 0.0383, 0.0228], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 16:40:19,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4761, 2.0336, 2.2495, 4.3163, 1.8426, 2.6828, 2.1837, 2.2772], device='cuda:3'), covar=tensor([0.0733, 0.2756, 0.1654, 0.0269, 0.3481, 0.1675, 0.2532, 0.2493], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0352, 0.0297, 0.0320, 0.0389, 0.0385, 0.0317, 0.0417], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:40:23,701 INFO [train.py:904] (3/8) Epoch 7, batch 5050, loss[loss=0.2295, simple_loss=0.3152, pruned_loss=0.07194, over 16440.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2979, pruned_loss=0.0642, over 3187474.35 frames. ], batch size: 146, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:41:14,577 INFO [optim.py:368] (3/8) 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 16:41:16,132 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 5100, loss[loss=0.1825, simple_loss=0.2614, pruned_loss=0.0518, over 16689.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2966, pruned_loss=0.06381, over 3185630.90 frames. ], batch size: 57, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:42:17,537 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-04-28 16:42:46,394 INFO [train.py:904] (3/8) Epoch 7, batch 5150, loss[loss=0.2257, simple_loss=0.3151, pruned_loss=0.06812, over 16914.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2967, pruned_loss=0.06305, over 3197957.72 frames. ], batch size: 96, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:43:37,177 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 5200, loss[loss=0.2072, simple_loss=0.2903, pruned_loss=0.06203, over 16889.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2958, pruned_loss=0.06284, over 3192862.03 frames. ], batch size: 116, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:44:56,209 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 5250, loss[loss=0.2111, simple_loss=0.2931, pruned_loss=0.06454, over 16436.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2939, pruned_loss=0.06269, over 3181294.07 frames. ], batch size: 68, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:45:18,973 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:46:00,025 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:46:01,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5533, 2.4300, 1.9460, 2.3659, 2.8984, 2.6654, 3.3247, 3.2507], device='cuda:3'), covar=tensor([0.0031, 0.0224, 0.0302, 0.0255, 0.0134, 0.0208, 0.0088, 0.0110], device='cuda:3'), in_proj_covar=tensor([0.0101, 0.0171, 0.0172, 0.0170, 0.0167, 0.0174, 0.0162, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:46:06,940 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:46:18,393 INFO [train.py:904] (3/8) Epoch 7, batch 5300, loss[loss=0.2381, simple_loss=0.305, pruned_loss=0.08561, over 12098.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2897, pruned_loss=0.06099, over 3196556.78 frames. ], batch size: 247, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:46:23,353 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:46:45,773 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:47:06,138 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:47:22,759 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2030, 3.9989, 4.2543, 4.4571, 4.5736, 4.1208, 4.5741, 4.5554], device='cuda:3'), covar=tensor([0.1057, 0.0881, 0.1228, 0.0461, 0.0413, 0.0893, 0.0371, 0.0427], device='cuda:3'), in_proj_covar=tensor([0.0445, 0.0540, 0.0679, 0.0552, 0.0414, 0.0412, 0.0420, 0.0468], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:47:28,949 INFO [train.py:904] (3/8) Epoch 7, batch 5350, loss[loss=0.2209, simple_loss=0.3055, pruned_loss=0.06815, over 16774.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2879, pruned_loss=0.06016, over 3214307.76 frames. ], batch size: 124, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:47:36,945 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:47:42,809 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 16:48:20,487 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.423e+02 2.781e+02 3.267e+02 5.283e+02, threshold=5.561e+02, percent-clipped=0.0 2023-04-28 16:48:40,524 INFO [train.py:904] (3/8) Epoch 7, batch 5400, loss[loss=0.2444, simple_loss=0.3274, pruned_loss=0.08063, over 16943.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2904, pruned_loss=0.06059, over 3219672.97 frames. ], batch size: 109, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:49:01,621 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4377, 4.3902, 4.3394, 3.6698, 4.3528, 1.6156, 4.1149, 4.2063], device='cuda:3'), covar=tensor([0.0072, 0.0062, 0.0083, 0.0342, 0.0061, 0.2033, 0.0090, 0.0151], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0093, 0.0142, 0.0141, 0.0109, 0.0155, 0.0125, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:49:05,982 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:49:55,998 INFO [train.py:904] (3/8) Epoch 7, batch 5450, loss[loss=0.2542, simple_loss=0.3275, pruned_loss=0.0904, over 17032.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2945, pruned_loss=0.0632, over 3210322.21 frames. ], batch size: 55, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:50:16,130 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.40 vs. limit=5.0 2023-04-28 16:50:50,980 INFO [optim.py:368] (3/8) 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 16:50:58,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4346, 4.7699, 4.4612, 4.5583, 4.1721, 4.1932, 4.2812, 4.8234], device='cuda:3'), covar=tensor([0.0846, 0.0784, 0.1026, 0.0590, 0.0703, 0.1097, 0.0782, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0436, 0.0558, 0.0474, 0.0366, 0.0348, 0.0371, 0.0459, 0.0409], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 16:51:12,547 INFO [train.py:904] (3/8) Epoch 7, batch 5500, loss[loss=0.2737, simple_loss=0.3412, pruned_loss=0.1032, over 15338.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3031, pruned_loss=0.0698, over 3163198.30 frames. ], batch size: 190, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:52:29,866 INFO [train.py:904] (3/8) Epoch 7, batch 5550, loss[loss=0.2413, simple_loss=0.3129, pruned_loss=0.08484, over 16556.00 frames. ], tot_loss[loss=0.2327, simple_loss=0.3117, pruned_loss=0.07691, over 3133283.02 frames. ], batch size: 68, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:53:27,054 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.385e+02 3.933e+02 4.760e+02 6.147e+02 1.111e+03, threshold=9.520e+02, percent-clipped=8.0 2023-04-28 16:53:36,376 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:53:38,466 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0541, 3.2262, 3.4158, 1.5546, 3.6129, 3.6247, 2.6878, 2.6736], device='cuda:3'), covar=tensor([0.0924, 0.0199, 0.0194, 0.1297, 0.0061, 0.0097, 0.0453, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0097, 0.0082, 0.0141, 0.0072, 0.0087, 0.0121, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 16:53:47,092 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:53:49,152 INFO [train.py:904] (3/8) Epoch 7, batch 5600, loss[loss=0.2608, simple_loss=0.3422, pruned_loss=0.08975, over 16690.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3171, pruned_loss=0.08181, over 3102964.39 frames. ], batch size: 89, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:54:11,812 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:54:20,966 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:54:54,203 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:55:08,329 INFO [train.py:904] (3/8) Epoch 7, batch 5650, loss[loss=0.2569, simple_loss=0.3415, pruned_loss=0.08615, over 16976.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3232, pruned_loss=0.08684, over 3087577.33 frames. ], batch size: 41, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:55:53,668 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:56:03,089 INFO [optim.py:368] (3/8) 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 16:56:23,910 INFO [train.py:904] (3/8) Epoch 7, batch 5700, loss[loss=0.2845, simple_loss=0.3529, pruned_loss=0.108, over 15281.00 frames. ], tot_loss[loss=0.2518, simple_loss=0.3254, pruned_loss=0.08912, over 3051586.48 frames. ], batch size: 190, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:56:42,483 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:56:53,317 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 16:57:20,225 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7627, 5.2770, 5.5037, 5.3131, 5.4052, 5.8808, 5.3643, 5.1965], device='cuda:3'), covar=tensor([0.0876, 0.1374, 0.1294, 0.1316, 0.1766, 0.0770, 0.1114, 0.2229], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0436, 0.0438, 0.0370, 0.0497, 0.0471, 0.0350, 0.0509], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:57:25,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7373, 3.6921, 3.8338, 3.6677, 3.8149, 4.1687, 3.8997, 3.6882], device='cuda:3'), covar=tensor([0.1750, 0.1856, 0.1547, 0.2041, 0.2220, 0.1332, 0.1271, 0.2516], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0435, 0.0438, 0.0370, 0.0496, 0.0471, 0.0350, 0.0509], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 16:57:32,650 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6456, 1.3813, 2.0992, 2.5178, 2.5131, 2.8537, 1.6229, 2.8087], device='cuda:3'), covar=tensor([0.0105, 0.0333, 0.0214, 0.0172, 0.0163, 0.0099, 0.0346, 0.0067], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0158, 0.0143, 0.0142, 0.0151, 0.0106, 0.0158, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 16:57:41,028 INFO [train.py:904] (3/8) Epoch 7, batch 5750, loss[loss=0.2541, simple_loss=0.3304, pruned_loss=0.0889, over 16718.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3288, pruned_loss=0.09144, over 3016732.66 frames. ], batch size: 134, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:58:19,762 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 16:58:23,343 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6812, 1.4505, 2.0736, 2.5118, 2.5002, 2.7719, 1.6767, 2.7263], device='cuda:3'), covar=tensor([0.0090, 0.0303, 0.0200, 0.0147, 0.0145, 0.0096, 0.0284, 0.0078], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0157, 0.0143, 0.0141, 0.0150, 0.0106, 0.0157, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 16:58:39,455 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 5800, loss[loss=0.2634, simple_loss=0.3389, pruned_loss=0.09401, over 16391.00 frames. ], tot_loss[loss=0.2537, simple_loss=0.3284, pruned_loss=0.08946, over 3038988.52 frames. ], batch size: 146, lr: 9.70e-03, grad_scale: 16.0 2023-04-28 16:59:48,304 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:59:48,409 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0089, 2.3580, 2.2790, 2.9928, 2.4143, 3.2623, 1.7079, 2.7647], device='cuda:3'), covar=tensor([0.1103, 0.0478, 0.0895, 0.0101, 0.0181, 0.0329, 0.1188, 0.0599], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0148, 0.0171, 0.0104, 0.0198, 0.0196, 0.0168, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 17:00:19,932 INFO [train.py:904] (3/8) Epoch 7, batch 5850, loss[loss=0.2004, simple_loss=0.2933, pruned_loss=0.05371, over 16903.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3258, pruned_loss=0.08706, over 3052430.68 frames. ], batch size: 90, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:00:41,622 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-28 17:00:44,495 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4117, 4.0157, 4.1328, 2.7319, 3.6465, 4.0984, 3.8545, 2.1648], device='cuda:3'), covar=tensor([0.0337, 0.0021, 0.0025, 0.0256, 0.0052, 0.0054, 0.0036, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0059, 0.0060, 0.0120, 0.0068, 0.0078, 0.0068, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 17:00:51,607 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 17:01:20,005 INFO [optim.py:368] (3/8) 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 17:01:25,701 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:01:38,962 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:01:41,100 INFO [train.py:904] (3/8) Epoch 7, batch 5900, loss[loss=0.2253, simple_loss=0.3038, pruned_loss=0.07337, over 15331.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3244, pruned_loss=0.08621, over 3063093.65 frames. ], batch size: 191, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:02:07,198 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:02:56,754 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:03:02,862 INFO [train.py:904] (3/8) Epoch 7, batch 5950, loss[loss=0.2327, simple_loss=0.3183, pruned_loss=0.07352, over 16915.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3249, pruned_loss=0.08445, over 3076089.39 frames. ], batch size: 109, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:03:19,293 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3787, 5.3241, 5.1121, 3.8559, 5.2481, 2.0027, 4.9134, 5.0897], device='cuda:3'), covar=tensor([0.0078, 0.0059, 0.0125, 0.0511, 0.0064, 0.2299, 0.0109, 0.0182], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0092, 0.0139, 0.0137, 0.0106, 0.0154, 0.0122, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:03:23,046 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:03:41,731 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:04:00,335 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 6000, loss[loss=0.2718, simple_loss=0.3281, pruned_loss=0.1077, over 11137.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3238, pruned_loss=0.08419, over 3066815.77 frames. ], batch size: 247, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:04:22,260 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 17:04:32,879 INFO [train.py:938] (3/8) Epoch 7, validation: loss=0.1758, simple_loss=0.2891, pruned_loss=0.03127, over 944034.00 frames. 2023-04-28 17:04:32,880 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 17:04:45,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0578, 4.1738, 2.2589, 4.7265, 2.8988, 4.7073, 2.2426, 2.9609], device='cuda:3'), covar=tensor([0.0148, 0.0239, 0.1499, 0.0042, 0.0677, 0.0207, 0.1448, 0.0640], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0157, 0.0180, 0.0096, 0.0163, 0.0194, 0.0188, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 17:04:51,832 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:05:28,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8066, 4.7535, 4.6398, 3.9314, 4.6225, 1.7635, 4.4637, 4.5896], device='cuda:3'), covar=tensor([0.0074, 0.0067, 0.0105, 0.0356, 0.0068, 0.2027, 0.0096, 0.0150], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0092, 0.0139, 0.0137, 0.0105, 0.0154, 0.0122, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:05:34,313 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 17:05:53,512 INFO [train.py:904] (3/8) Epoch 7, batch 6050, loss[loss=0.278, simple_loss=0.3296, pruned_loss=0.1133, over 11733.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3214, pruned_loss=0.08281, over 3074830.29 frames. ], batch size: 246, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:06:02,547 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:09,559 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:50,893 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.435e+02 4.153e+02 5.087e+02 8.285e+02, threshold=8.306e+02, percent-clipped=0.0 2023-04-28 17:07:12,265 INFO [train.py:904] (3/8) Epoch 7, batch 6100, loss[loss=0.2083, simple_loss=0.2986, pruned_loss=0.05901, over 16716.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3198, pruned_loss=0.0806, over 3106447.51 frames. ], batch size: 89, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:07:39,936 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:07:54,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5025, 4.6679, 4.7506, 4.7442, 4.6695, 5.2149, 4.8172, 4.5495], device='cuda:3'), covar=tensor([0.0988, 0.1436, 0.1432, 0.1554, 0.2349, 0.0896, 0.1089, 0.2260], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0436, 0.0443, 0.0375, 0.0509, 0.0479, 0.0354, 0.0513], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 17:08:32,437 INFO [train.py:904] (3/8) Epoch 7, batch 6150, loss[loss=0.2459, simple_loss=0.3249, pruned_loss=0.08339, over 16300.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3179, pruned_loss=0.07982, over 3115539.48 frames. ], batch size: 165, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:09:23,201 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:09:29,183 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:09:51,778 INFO [train.py:904] (3/8) Epoch 7, batch 6200, loss[loss=0.2446, simple_loss=0.3272, pruned_loss=0.081, over 16735.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3164, pruned_loss=0.07976, over 3104496.53 frames. ], batch size: 83, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:09:53,485 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 17:11:02,464 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:11:14,125 INFO [train.py:904] (3/8) Epoch 7, batch 6250, loss[loss=0.2222, simple_loss=0.3101, pruned_loss=0.06716, over 16290.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3155, pruned_loss=0.07942, over 3093443.89 frames. ], batch size: 35, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:14,653 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9615, 2.6647, 2.6376, 1.7460, 2.7890, 2.8326, 2.4235, 2.3451], device='cuda:3'), covar=tensor([0.0805, 0.0195, 0.0191, 0.1060, 0.0086, 0.0134, 0.0414, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0095, 0.0081, 0.0137, 0.0070, 0.0085, 0.0118, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 17:11:52,528 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:12:07,946 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5551, 4.3543, 4.0622, 1.9793, 3.2543, 2.7353, 3.9845, 4.1013], device='cuda:3'), covar=tensor([0.0242, 0.0435, 0.0499, 0.1863, 0.0701, 0.0853, 0.0609, 0.0708], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0130, 0.0155, 0.0141, 0.0133, 0.0124, 0.0136, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 17:12:11,782 INFO [optim.py:368] (3/8) 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 17:12:13,373 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 17:12:30,898 INFO [train.py:904] (3/8) Epoch 7, batch 6300, loss[loss=0.2395, simple_loss=0.3142, pruned_loss=0.08242, over 15165.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3154, pruned_loss=0.07888, over 3094594.41 frames. ], batch size: 190, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:12:49,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7269, 2.1953, 1.6239, 1.9583, 2.5366, 2.2773, 2.7388, 2.8300], device='cuda:3'), covar=tensor([0.0061, 0.0209, 0.0303, 0.0256, 0.0138, 0.0190, 0.0116, 0.0111], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0172, 0.0169, 0.0168, 0.0165, 0.0171, 0.0164, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:12:53,818 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8499, 4.1553, 3.9582, 3.9954, 3.7058, 3.7575, 3.8654, 4.0829], device='cuda:3'), covar=tensor([0.0770, 0.0706, 0.0817, 0.0513, 0.0587, 0.1328, 0.0701, 0.0886], device='cuda:3'), in_proj_covar=tensor([0.0438, 0.0553, 0.0475, 0.0362, 0.0347, 0.0369, 0.0461, 0.0410], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:13:08,170 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:13:49,867 INFO [train.py:904] (3/8) Epoch 7, batch 6350, loss[loss=0.2315, simple_loss=0.3134, pruned_loss=0.07478, over 16730.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3171, pruned_loss=0.08088, over 3090449.46 frames. ], batch size: 134, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:14:31,067 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 17:14:47,871 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 6400, loss[loss=0.2077, simple_loss=0.2863, pruned_loss=0.06454, over 17223.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3178, pruned_loss=0.08238, over 3075091.31 frames. ], batch size: 44, lr: 9.66e-03, grad_scale: 8.0 2023-04-28 17:15:22,391 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:16:20,298 INFO [train.py:904] (3/8) Epoch 7, batch 6450, loss[loss=0.2157, simple_loss=0.2956, pruned_loss=0.06789, over 16838.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3161, pruned_loss=0.0803, over 3094882.19 frames. ], batch size: 116, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:17:16,664 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:17:22,054 INFO [optim.py:368] (3/8) 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 17:17:38,435 INFO [train.py:904] (3/8) Epoch 7, batch 6500, loss[loss=0.252, simple_loss=0.3124, pruned_loss=0.09586, over 11922.00 frames. ], tot_loss[loss=0.236, simple_loss=0.314, pruned_loss=0.07901, over 3113625.28 frames. ], batch size: 247, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:18:26,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8863, 5.5651, 5.6860, 5.5490, 5.4651, 6.1136, 5.6428, 5.3417], device='cuda:3'), covar=tensor([0.0788, 0.1452, 0.1676, 0.1405, 0.2294, 0.0741, 0.1128, 0.2232], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0425, 0.0436, 0.0365, 0.0493, 0.0466, 0.0349, 0.0501], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 17:18:29,447 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:18:36,061 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:18:57,740 INFO [train.py:904] (3/8) Epoch 7, batch 6550, loss[loss=0.2136, simple_loss=0.3105, pruned_loss=0.05831, over 16190.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3171, pruned_loss=0.08056, over 3101398.79 frames. ], batch size: 165, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:18:58,340 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 17:19:21,402 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-28 17:19:56,014 INFO [optim.py:368] (3/8) 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 17:20:13,102 INFO [train.py:904] (3/8) Epoch 7, batch 6600, loss[loss=0.3107, simple_loss=0.3546, pruned_loss=0.1334, over 11362.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3202, pruned_loss=0.08159, over 3105667.76 frames. ], batch size: 247, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:20:28,341 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 17:20:30,797 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:20:45,930 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8662, 1.9748, 2.2623, 3.1209, 2.0792, 2.3216, 2.2289, 2.0291], device='cuda:3'), covar=tensor([0.0717, 0.2305, 0.1238, 0.0456, 0.2761, 0.1486, 0.2042, 0.2344], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0350, 0.0293, 0.0319, 0.0391, 0.0379, 0.0316, 0.0414], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:20:47,535 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-28 17:21:29,933 INFO [train.py:904] (3/8) Epoch 7, batch 6650, loss[loss=0.2955, simple_loss=0.3462, pruned_loss=0.1224, over 11779.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3203, pruned_loss=0.08235, over 3085877.91 frames. ], batch size: 248, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:21:46,728 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4380, 1.9227, 1.3612, 1.6712, 2.3549, 2.0516, 2.4999, 2.4756], device='cuda:3'), covar=tensor([0.0075, 0.0295, 0.0439, 0.0362, 0.0166, 0.0276, 0.0143, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0174, 0.0173, 0.0171, 0.0168, 0.0173, 0.0165, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:22:02,865 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:22:27,319 INFO [optim.py:368] (3/8) 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 17:22:31,005 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5999, 2.6652, 2.2545, 3.7787, 2.8618, 3.7216, 1.4365, 2.7405], device='cuda:3'), covar=tensor([0.1409, 0.0604, 0.1291, 0.0121, 0.0293, 0.0374, 0.1565, 0.0784], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0148, 0.0171, 0.0104, 0.0198, 0.0198, 0.0167, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 17:22:43,254 INFO [train.py:904] (3/8) Epoch 7, batch 6700, loss[loss=0.2106, simple_loss=0.2907, pruned_loss=0.0652, over 16587.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3191, pruned_loss=0.08232, over 3094811.26 frames. ], batch size: 62, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:23:00,287 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:23:55,788 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8019, 4.7813, 4.6142, 4.4125, 4.2012, 4.6998, 4.5569, 4.3329], device='cuda:3'), covar=tensor([0.0426, 0.0265, 0.0206, 0.0194, 0.0905, 0.0277, 0.0297, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0237, 0.0239, 0.0212, 0.0271, 0.0242, 0.0168, 0.0279], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:23:56,587 INFO [train.py:904] (3/8) Epoch 7, batch 6750, loss[loss=0.213, simple_loss=0.2942, pruned_loss=0.06591, over 16824.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3174, pruned_loss=0.08191, over 3084274.23 frames. ], batch size: 83, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:23:58,545 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-28 17:24:11,055 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:37,913 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:54,245 INFO [optim.py:368] (3/8) 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 17:24:55,940 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 6800, loss[loss=0.2385, simple_loss=0.3113, pruned_loss=0.08289, over 16550.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3168, pruned_loss=0.08134, over 3076716.44 frames. ], batch size: 62, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:06,148 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:10,107 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 6850, loss[loss=0.212, simple_loss=0.3173, pruned_loss=0.05334, over 17245.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3184, pruned_loss=0.08168, over 3070860.02 frames. ], batch size: 52, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:25,803 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:26:36,105 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 17:27:14,416 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 7, batch 6900, loss[loss=0.3419, simple_loss=0.382, pruned_loss=0.1509, over 11923.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3211, pruned_loss=0.08183, over 3063036.67 frames. ], batch size: 246, lr: 9.63e-03, grad_scale: 2.0 2023-04-28 17:27:59,584 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-28 17:28:36,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6498, 4.6893, 4.5022, 3.8621, 4.4729, 1.8000, 4.3147, 4.4658], device='cuda:3'), covar=tensor([0.0055, 0.0043, 0.0096, 0.0273, 0.0059, 0.1837, 0.0089, 0.0113], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0090, 0.0138, 0.0135, 0.0105, 0.0154, 0.0123, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:28:46,545 INFO [train.py:904] (3/8) Epoch 7, batch 6950, loss[loss=0.2158, simple_loss=0.2945, pruned_loss=0.06858, over 16626.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3227, pruned_loss=0.0835, over 3064191.55 frames. ], batch size: 62, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:29:13,202 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:29:46,385 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 7000, loss[loss=0.2087, simple_loss=0.3088, pruned_loss=0.05433, over 16873.00 frames. ], tot_loss[loss=0.2444, simple_loss=0.3228, pruned_loss=0.08298, over 3063259.55 frames. ], batch size: 96, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:30:17,002 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8553, 5.1137, 4.8356, 4.8911, 4.5441, 4.4337, 4.5666, 5.1874], device='cuda:3'), covar=tensor([0.0822, 0.0744, 0.1081, 0.0577, 0.0758, 0.0839, 0.0907, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0449, 0.0559, 0.0484, 0.0372, 0.0355, 0.0381, 0.0468, 0.0418], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:30:27,611 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1720, 1.3873, 1.7791, 2.0463, 2.2256, 2.3831, 1.3602, 2.2484], device='cuda:3'), covar=tensor([0.0119, 0.0296, 0.0175, 0.0204, 0.0165, 0.0102, 0.0327, 0.0072], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0153, 0.0135, 0.0137, 0.0146, 0.0102, 0.0153, 0.0094], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 17:31:13,269 INFO [train.py:904] (3/8) Epoch 7, batch 7050, loss[loss=0.3149, simple_loss=0.3597, pruned_loss=0.1351, over 11230.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.324, pruned_loss=0.08323, over 3063732.47 frames. ], batch size: 247, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:31:57,417 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 7, batch 7100, loss[loss=0.2136, simple_loss=0.3036, pruned_loss=0.06176, over 16826.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3217, pruned_loss=0.08222, over 3074153.45 frames. ], batch size: 96, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:33:21,413 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:30,277 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:39,219 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:33:44,542 INFO [train.py:904] (3/8) Epoch 7, batch 7150, loss[loss=0.2508, simple_loss=0.3283, pruned_loss=0.08663, over 16919.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3208, pruned_loss=0.08267, over 3079076.90 frames. ], batch size: 109, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:34:03,085 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 17:34:37,160 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:34:44,522 INFO [optim.py:368] (3/8) 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 17:34:58,220 INFO [train.py:904] (3/8) Epoch 7, batch 7200, loss[loss=0.2484, simple_loss=0.321, pruned_loss=0.08784, over 11764.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3182, pruned_loss=0.08088, over 3066698.16 frames. ], batch size: 248, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:35:32,821 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0278, 3.4920, 3.5220, 2.1469, 3.0734, 3.4837, 3.3064, 1.9861], device='cuda:3'), covar=tensor([0.0341, 0.0019, 0.0026, 0.0269, 0.0053, 0.0057, 0.0044, 0.0301], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0054, 0.0059, 0.0114, 0.0063, 0.0075, 0.0066, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 17:35:51,856 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3316, 1.5114, 2.6166, 3.2032, 2.7604, 3.4865, 1.7008, 3.3813], device='cuda:3'), covar=tensor([0.0089, 0.0342, 0.0156, 0.0100, 0.0140, 0.0071, 0.0358, 0.0063], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0156, 0.0136, 0.0137, 0.0147, 0.0102, 0.0154, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 17:36:11,179 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:36:16,027 INFO [train.py:904] (3/8) Epoch 7, batch 7250, loss[loss=0.2031, simple_loss=0.2862, pruned_loss=0.06004, over 16785.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3154, pruned_loss=0.07885, over 3073462.78 frames. ], batch size: 83, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:36:41,036 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.214e+02 3.991e+02 5.016e+02 1.051e+03, threshold=7.982e+02, percent-clipped=2.0 2023-04-28 17:37:29,946 INFO [train.py:904] (3/8) Epoch 7, batch 7300, loss[loss=0.2402, simple_loss=0.3219, pruned_loss=0.07923, over 15394.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3148, pruned_loss=0.07918, over 3068000.93 frames. ], batch size: 190, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:37:52,949 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:37:58,759 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2045, 1.9131, 1.5323, 1.6232, 2.1490, 1.8441, 2.2082, 2.3298], device='cuda:3'), covar=tensor([0.0062, 0.0216, 0.0298, 0.0286, 0.0134, 0.0220, 0.0105, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0099, 0.0171, 0.0172, 0.0170, 0.0166, 0.0174, 0.0163, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:38:44,625 INFO [train.py:904] (3/8) Epoch 7, batch 7350, loss[loss=0.2882, simple_loss=0.3384, pruned_loss=0.119, over 11357.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3149, pruned_loss=0.07946, over 3051638.11 frames. ], batch size: 248, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:38:53,808 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 17:39:28,839 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 17:39:48,569 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 3.085e+02 3.772e+02 4.630e+02 1.239e+03, threshold=7.544e+02, percent-clipped=2.0 2023-04-28 17:39:49,492 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 17:39:51,985 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3055, 4.0511, 4.2867, 4.4383, 4.5642, 4.0968, 4.5432, 4.5693], device='cuda:3'), covar=tensor([0.1176, 0.0904, 0.1265, 0.0592, 0.0493, 0.0933, 0.0546, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0432, 0.0530, 0.0662, 0.0536, 0.0407, 0.0408, 0.0425, 0.0459], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:40:01,145 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-28 17:40:02,148 INFO [train.py:904] (3/8) Epoch 7, batch 7400, loss[loss=0.2457, simple_loss=0.3289, pruned_loss=0.08125, over 16573.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3159, pruned_loss=0.0799, over 3062294.05 frames. ], batch size: 57, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:40:53,697 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:40:54,971 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:14,135 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:19,088 INFO [train.py:904] (3/8) Epoch 7, batch 7450, loss[loss=0.2404, simple_loss=0.3249, pruned_loss=0.07789, over 16912.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3179, pruned_loss=0.08194, over 3054124.43 frames. ], batch size: 116, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:42:11,321 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:35,764 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:39,839 INFO [train.py:904] (3/8) Epoch 7, batch 7500, loss[loss=0.2176, simple_loss=0.2968, pruned_loss=0.06925, over 16191.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3178, pruned_loss=0.0805, over 3078266.08 frames. ], batch size: 165, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:43:44,347 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:43:57,923 INFO [train.py:904] (3/8) Epoch 7, batch 7550, loss[loss=0.2223, simple_loss=0.2992, pruned_loss=0.07269, over 16506.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3166, pruned_loss=0.08064, over 3067774.31 frames. ], batch size: 75, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:44:10,830 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:44:21,239 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 17:44:59,735 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.309e+02 3.638e+02 4.521e+02 5.512e+02 1.190e+03, threshold=9.043e+02, percent-clipped=2.0 2023-04-28 17:45:12,772 INFO [train.py:904] (3/8) Epoch 7, batch 7600, loss[loss=0.247, simple_loss=0.3255, pruned_loss=0.0843, over 16885.00 frames. ], tot_loss[loss=0.2391, simple_loss=0.3163, pruned_loss=0.081, over 3069014.31 frames. ], batch size: 42, lr: 9.58e-03, grad_scale: 8.0 2023-04-28 17:45:29,691 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4296, 4.3790, 4.3137, 3.0064, 4.2926, 1.3168, 4.0002, 4.0104], device='cuda:3'), covar=tensor([0.0135, 0.0101, 0.0175, 0.0649, 0.0106, 0.3064, 0.0158, 0.0304], device='cuda:3'), in_proj_covar=tensor([0.0104, 0.0091, 0.0139, 0.0133, 0.0106, 0.0157, 0.0123, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:46:28,276 INFO [train.py:904] (3/8) Epoch 7, batch 7650, loss[loss=0.2698, simple_loss=0.3255, pruned_loss=0.107, over 10985.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.317, pruned_loss=0.08226, over 3061439.86 frames. ], batch size: 248, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:47:08,536 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:47:29,673 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 7700, loss[loss=0.2265, simple_loss=0.3074, pruned_loss=0.07284, over 16749.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3172, pruned_loss=0.08242, over 3073717.09 frames. ], batch size: 76, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:48:34,691 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:38,387 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:43,242 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8928, 4.7833, 4.7316, 4.5286, 4.2802, 4.7004, 4.6631, 4.4590], device='cuda:3'), covar=tensor([0.0477, 0.0472, 0.0236, 0.0235, 0.0874, 0.0432, 0.0297, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0203, 0.0236, 0.0235, 0.0209, 0.0265, 0.0238, 0.0165, 0.0272], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:48:57,123 INFO [train.py:904] (3/8) Epoch 7, batch 7750, loss[loss=0.2728, simple_loss=0.3494, pruned_loss=0.09806, over 16241.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3174, pruned_loss=0.08233, over 3080090.74 frames. ], batch size: 165, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:49:44,494 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:44,707 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:57,426 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 7800, loss[loss=0.3003, simple_loss=0.3512, pruned_loss=0.1247, over 11723.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.319, pruned_loss=0.08333, over 3074556.02 frames. ], batch size: 247, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:12,790 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:51:17,584 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:51:25,175 INFO [train.py:904] (3/8) Epoch 7, batch 7850, loss[loss=0.239, simple_loss=0.3163, pruned_loss=0.08089, over 16751.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3191, pruned_loss=0.0823, over 3092832.99 frames. ], batch size: 124, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:30,905 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:52:24,298 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:52:26,860 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 7900, loss[loss=0.2645, simple_loss=0.3392, pruned_loss=0.09491, over 16199.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3176, pruned_loss=0.08132, over 3090212.93 frames. ], batch size: 165, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:53:58,006 INFO [train.py:904] (3/8) Epoch 7, batch 7950, loss[loss=0.24, simple_loss=0.316, pruned_loss=0.082, over 16774.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3182, pruned_loss=0.08187, over 3086444.04 frames. ], batch size: 124, lr: 9.55e-03, grad_scale: 2.0 2023-04-28 17:54:43,417 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-28 17:55:03,363 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.365e+02 3.868e+02 4.958e+02 9.415e+02, threshold=7.735e+02, percent-clipped=1.0 2023-04-28 17:55:12,758 INFO [train.py:904] (3/8) Epoch 7, batch 8000, loss[loss=0.2883, simple_loss=0.3451, pruned_loss=0.1158, over 11172.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3192, pruned_loss=0.08271, over 3072547.59 frames. ], batch size: 248, lr: 9.55e-03, grad_scale: 4.0 2023-04-28 17:56:00,095 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:56:25,020 INFO [train.py:904] (3/8) Epoch 7, batch 8050, loss[loss=0.2713, simple_loss=0.327, pruned_loss=0.1078, over 11302.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3192, pruned_loss=0.08278, over 3054537.50 frames. ], batch size: 247, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:56:31,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9085, 5.2218, 5.0029, 4.9345, 4.5980, 4.5522, 4.6712, 5.3194], device='cuda:3'), covar=tensor([0.0811, 0.0730, 0.0930, 0.0550, 0.0738, 0.0749, 0.0846, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0445, 0.0565, 0.0482, 0.0374, 0.0355, 0.0384, 0.0470, 0.0419], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:56:41,042 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7617, 4.0544, 4.3555, 2.0275, 4.5867, 4.5818, 2.9758, 3.3703], device='cuda:3'), covar=tensor([0.0714, 0.0130, 0.0123, 0.1036, 0.0030, 0.0070, 0.0359, 0.0355], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0094, 0.0081, 0.0138, 0.0068, 0.0085, 0.0118, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 17:56:59,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6016, 3.6345, 2.8690, 2.2580, 2.6673, 2.2461, 3.8991, 3.6610], device='cuda:3'), covar=tensor([0.2537, 0.0815, 0.1491, 0.1869, 0.2035, 0.1690, 0.0418, 0.0858], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0253, 0.0277, 0.0261, 0.0283, 0.0210, 0.0252, 0.0272], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:57:14,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8077, 3.0624, 2.9040, 4.6490, 3.7274, 4.3072, 1.7067, 3.2514], device='cuda:3'), covar=tensor([0.1241, 0.0598, 0.0905, 0.0108, 0.0357, 0.0317, 0.1340, 0.0678], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0148, 0.0170, 0.0106, 0.0200, 0.0198, 0.0169, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 17:57:30,025 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 8100, loss[loss=0.1942, simple_loss=0.2867, pruned_loss=0.05082, over 16902.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3176, pruned_loss=0.08089, over 3084149.41 frames. ], batch size: 96, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:58:37,848 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 17:58:40,375 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:43,430 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 8150, loss[loss=0.2009, simple_loss=0.281, pruned_loss=0.06036, over 16613.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3149, pruned_loss=0.07966, over 3089199.43 frames. ], batch size: 62, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:59:00,761 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:59:18,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5084, 3.4492, 2.7917, 2.1523, 2.3840, 2.1972, 3.4743, 3.3659], device='cuda:3'), covar=tensor([0.2374, 0.0638, 0.1404, 0.1954, 0.2014, 0.1629, 0.0502, 0.0841], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0252, 0.0276, 0.0261, 0.0282, 0.0210, 0.0252, 0.0271], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 17:59:59,698 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.706e+02 3.772e+02 4.723e+02 6.532e+02 2.181e+03, threshold=9.446e+02, percent-clipped=10.0 2023-04-28 18:00:10,161 INFO [train.py:904] (3/8) Epoch 7, batch 8200, loss[loss=0.2199, simple_loss=0.2944, pruned_loss=0.07269, over 16932.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3119, pruned_loss=0.07812, over 3118781.60 frames. ], batch size: 109, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:00:12,050 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:00:13,561 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:30,736 INFO [train.py:904] (3/8) Epoch 7, batch 8250, loss[loss=0.2171, simple_loss=0.3075, pruned_loss=0.06333, over 16405.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3109, pruned_loss=0.07593, over 3103767.20 frames. ], batch size: 146, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:01:47,250 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:58,247 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.107e+02 2.942e+02 3.716e+02 4.471e+02 8.485e+02, threshold=7.433e+02, percent-clipped=0.0 2023-04-28 18:02:52,449 INFO [train.py:904] (3/8) Epoch 7, batch 8300, loss[loss=0.1991, simple_loss=0.2775, pruned_loss=0.06038, over 12052.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3076, pruned_loss=0.07251, over 3082952.61 frames. ], batch size: 246, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:02:54,898 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9550, 5.3878, 5.5129, 5.4287, 5.2820, 5.8865, 5.4895, 5.2245], device='cuda:3'), covar=tensor([0.0734, 0.1377, 0.1679, 0.1544, 0.2487, 0.0860, 0.1069, 0.2029], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0413, 0.0433, 0.0359, 0.0482, 0.0459, 0.0345, 0.0492], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 18:03:00,755 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9731, 3.7657, 3.9571, 4.1083, 4.2380, 3.7794, 4.1531, 4.2157], device='cuda:3'), covar=tensor([0.1081, 0.0946, 0.1298, 0.0631, 0.0476, 0.1325, 0.0598, 0.0553], device='cuda:3'), in_proj_covar=tensor([0.0439, 0.0539, 0.0667, 0.0539, 0.0407, 0.0413, 0.0431, 0.0467], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:03:21,320 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9693, 1.8973, 2.2829, 3.1413, 2.0327, 2.2298, 2.1412, 1.9327], device='cuda:3'), covar=tensor([0.0681, 0.2996, 0.1500, 0.0442, 0.3544, 0.1764, 0.2662, 0.3088], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0351, 0.0293, 0.0314, 0.0390, 0.0376, 0.0312, 0.0412], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:03:26,035 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:29,187 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:35,008 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:42,610 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:43,205 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-28 18:04:00,648 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:04:11,882 INFO [train.py:904] (3/8) Epoch 7, batch 8350, loss[loss=0.2195, simple_loss=0.307, pruned_loss=0.06601, over 15344.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3057, pruned_loss=0.07023, over 3075111.53 frames. ], batch size: 191, lr: 9.52e-03, grad_scale: 4.0 2023-04-28 18:04:18,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0918, 3.3363, 3.5953, 3.5696, 3.5660, 3.3672, 3.3881, 3.4358], device='cuda:3'), covar=tensor([0.0326, 0.0549, 0.0376, 0.0411, 0.0430, 0.0398, 0.0707, 0.0373], device='cuda:3'), in_proj_covar=tensor([0.0267, 0.0272, 0.0270, 0.0264, 0.0316, 0.0288, 0.0386, 0.0240], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 18:04:28,887 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5435, 3.2928, 3.1583, 1.8815, 2.8219, 2.3356, 3.0893, 3.3081], device='cuda:3'), covar=tensor([0.0312, 0.0572, 0.0499, 0.1671, 0.0684, 0.0915, 0.0702, 0.0589], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0128, 0.0150, 0.0138, 0.0131, 0.0122, 0.0133, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 18:04:30,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8350, 4.7877, 4.6468, 4.4686, 4.2723, 4.7264, 4.6040, 4.4143], device='cuda:3'), covar=tensor([0.0421, 0.0377, 0.0204, 0.0192, 0.0805, 0.0340, 0.0271, 0.0571], device='cuda:3'), in_proj_covar=tensor([0.0203, 0.0237, 0.0235, 0.0209, 0.0261, 0.0236, 0.0167, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:04:46,454 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:00,053 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:01,327 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3835, 3.3208, 3.4029, 3.5083, 3.5411, 3.2600, 3.5008, 3.5953], device='cuda:3'), covar=tensor([0.1032, 0.0698, 0.0928, 0.0497, 0.0591, 0.1938, 0.0705, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0437, 0.0531, 0.0658, 0.0535, 0.0406, 0.0409, 0.0427, 0.0462], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:05:06,532 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:20,697 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 8400, loss[loss=0.2096, simple_loss=0.2862, pruned_loss=0.0665, over 11804.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3027, pruned_loss=0.06757, over 3062570.52 frames. ], batch size: 247, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:05:39,336 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:06:02,565 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:23,776 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:06:35,206 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:52,097 INFO [train.py:904] (3/8) Epoch 7, batch 8450, loss[loss=0.203, simple_loss=0.276, pruned_loss=0.06494, over 12344.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3005, pruned_loss=0.06609, over 3038281.17 frames. ], batch size: 248, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:07:10,154 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7983, 3.8857, 4.2861, 4.2319, 4.2243, 3.9629, 3.9629, 3.9351], device='cuda:3'), covar=tensor([0.0304, 0.0544, 0.0340, 0.0422, 0.0425, 0.0330, 0.0832, 0.0385], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0274, 0.0271, 0.0267, 0.0317, 0.0293, 0.0387, 0.0240], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-28 18:07:18,728 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:23,139 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-28 18:07:38,192 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:41,422 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 18:07:50,679 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:58,610 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:08:10,405 INFO [train.py:904] (3/8) Epoch 7, batch 8500, loss[loss=0.1846, simple_loss=0.2757, pruned_loss=0.04677, over 16731.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2959, pruned_loss=0.06289, over 3038030.92 frames. ], batch size: 124, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:08:37,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8623, 1.3456, 2.3269, 2.7981, 2.5409, 2.9370, 1.6939, 3.0067], device='cuda:3'), covar=tensor([0.0081, 0.0356, 0.0155, 0.0133, 0.0142, 0.0123, 0.0358, 0.0072], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0156, 0.0136, 0.0136, 0.0146, 0.0100, 0.0153, 0.0092], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 18:08:55,022 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 8550, loss[loss=0.2013, simple_loss=0.2792, pruned_loss=0.06169, over 12052.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2931, pruned_loss=0.06154, over 3038817.94 frames. ], batch size: 250, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:09:41,811 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:10:58,766 INFO [optim.py:368] (3/8) 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 18:11:11,617 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 18:11:12,034 INFO [train.py:904] (3/8) Epoch 7, batch 8600, loss[loss=0.1882, simple_loss=0.2721, pruned_loss=0.05211, over 12453.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2931, pruned_loss=0.06055, over 3020574.66 frames. ], batch size: 248, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:11:46,306 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:11:56,250 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:12:05,301 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-28 18:12:17,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1440, 2.2134, 1.7835, 2.0259, 2.6763, 2.3645, 3.0221, 2.9480], device='cuda:3'), covar=tensor([0.0045, 0.0287, 0.0354, 0.0315, 0.0146, 0.0236, 0.0106, 0.0145], device='cuda:3'), in_proj_covar=tensor([0.0097, 0.0172, 0.0167, 0.0168, 0.0164, 0.0170, 0.0156, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:12:24,839 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 18:12:47,741 INFO [train.py:904] (3/8) Epoch 7, batch 8650, loss[loss=0.1834, simple_loss=0.2812, pruned_loss=0.04279, over 15462.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2911, pruned_loss=0.0588, over 3011428.82 frames. ], batch size: 191, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:13:54,362 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:14:32,612 INFO [train.py:904] (3/8) Epoch 7, batch 8700, loss[loss=0.2225, simple_loss=0.3106, pruned_loss=0.06723, over 15287.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2886, pruned_loss=0.05723, over 3020494.91 frames. ], batch size: 191, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:15:05,539 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-04-28 18:15:22,841 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:16:08,632 INFO [train.py:904] (3/8) Epoch 7, batch 8750, loss[loss=0.182, simple_loss=0.2651, pruned_loss=0.04946, over 12241.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2882, pruned_loss=0.05684, over 3029826.71 frames. ], batch size: 248, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:16:46,172 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 18:17:04,333 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:17:07,493 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:17:32,592 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 18:17:45,985 INFO [optim.py:368] (3/8) 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 18:17:55,790 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:18:01,164 INFO [train.py:904] (3/8) Epoch 7, batch 8800, loss[loss=0.2172, simple_loss=0.3001, pruned_loss=0.06714, over 12481.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2863, pruned_loss=0.05545, over 3045953.46 frames. ], batch size: 246, lr: 9.49e-03, grad_scale: 8.0 2023-04-28 18:18:47,923 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:11,475 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:19:26,669 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.36 vs. limit=5.0 2023-04-28 18:19:36,494 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:43,874 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 7, batch 8850, loss[loss=0.1774, simple_loss=0.2663, pruned_loss=0.04429, over 12452.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2886, pruned_loss=0.05489, over 3039195.97 frames. ], batch size: 250, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:21:21,320 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 8900, loss[loss=0.2018, simple_loss=0.2949, pruned_loss=0.05435, over 16327.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2884, pruned_loss=0.05382, over 3035473.88 frames. ], batch size: 165, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:22:06,690 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:22:21,948 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:23:02,966 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9375, 3.9607, 4.3773, 4.3479, 4.3616, 4.0600, 4.0893, 4.0028], device='cuda:3'), covar=tensor([0.0273, 0.0411, 0.0321, 0.0397, 0.0372, 0.0312, 0.0677, 0.0358], device='cuda:3'), in_proj_covar=tensor([0.0259, 0.0263, 0.0263, 0.0255, 0.0302, 0.0283, 0.0370, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 18:23:38,920 INFO [train.py:904] (3/8) Epoch 7, batch 8950, loss[loss=0.1857, simple_loss=0.2756, pruned_loss=0.04784, over 15295.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2881, pruned_loss=0.05349, over 3067256.22 frames. ], batch size: 191, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:24:09,508 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:22,086 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:42,965 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:45,995 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9396, 4.0083, 3.7667, 3.6345, 3.4563, 3.8761, 3.4810, 3.6110], device='cuda:3'), covar=tensor([0.0457, 0.0344, 0.0244, 0.0220, 0.0710, 0.0366, 0.0932, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0228, 0.0227, 0.0202, 0.0249, 0.0229, 0.0159, 0.0265], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:25:14,731 INFO [optim.py:368] (3/8) 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 18:25:27,263 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:25:27,970 INFO [train.py:904] (3/8) Epoch 7, batch 9000, loss[loss=0.1893, simple_loss=0.2843, pruned_loss=0.04714, over 16713.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2853, pruned_loss=0.05216, over 3083756.07 frames. ], batch size: 134, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:25:27,971 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 18:25:37,200 INFO [train.py:938] (3/8) Epoch 7, validation: loss=0.1647, simple_loss=0.2682, pruned_loss=0.03062, over 944034.00 frames. 2023-04-28 18:25:37,201 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 18:26:33,159 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:26:33,288 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:26:34,854 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3362, 1.3542, 1.6778, 2.2386, 2.1702, 2.3319, 1.6183, 2.3801], device='cuda:3'), covar=tensor([0.0102, 0.0316, 0.0186, 0.0162, 0.0165, 0.0119, 0.0270, 0.0079], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0152, 0.0134, 0.0134, 0.0143, 0.0098, 0.0151, 0.0089], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 18:26:57,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7388, 4.7195, 4.5537, 4.0853, 4.6049, 1.8127, 4.3568, 4.5675], device='cuda:3'), covar=tensor([0.0067, 0.0059, 0.0124, 0.0254, 0.0067, 0.1907, 0.0097, 0.0130], device='cuda:3'), in_proj_covar=tensor([0.0102, 0.0089, 0.0137, 0.0126, 0.0103, 0.0157, 0.0121, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 18:27:14,336 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:27:19,719 INFO [train.py:904] (3/8) Epoch 7, batch 9050, loss[loss=0.1937, simple_loss=0.2818, pruned_loss=0.05281, over 15399.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2872, pruned_loss=0.05345, over 3081387.89 frames. ], batch size: 191, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:27:52,771 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6751, 2.7555, 1.7223, 2.8618, 2.1058, 2.8340, 1.9817, 2.4913], device='cuda:3'), covar=tensor([0.0186, 0.0318, 0.1194, 0.0116, 0.0666, 0.0416, 0.1169, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0149, 0.0176, 0.0090, 0.0157, 0.0180, 0.0186, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 18:28:08,845 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 7, batch 9100, loss[loss=0.1883, simple_loss=0.2863, pruned_loss=0.04508, over 16864.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2862, pruned_loss=0.05357, over 3084411.49 frames. ], batch size: 102, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:29:58,209 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:29:58,322 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:30:13,169 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:31:01,027 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:31:03,199 INFO [train.py:904] (3/8) Epoch 7, batch 9150, loss[loss=0.1611, simple_loss=0.257, pruned_loss=0.03257, over 16908.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2868, pruned_loss=0.05338, over 3085022.11 frames. ], batch size: 96, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:31:06,651 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 18:31:47,575 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:13,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0741, 2.7059, 2.7008, 1.7500, 2.9012, 2.9130, 2.5059, 2.4395], device='cuda:3'), covar=tensor([0.0633, 0.0159, 0.0155, 0.0958, 0.0062, 0.0107, 0.0367, 0.0383], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0091, 0.0079, 0.0134, 0.0063, 0.0081, 0.0112, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 18:32:35,334 INFO [optim.py:368] (3/8) 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 18:32:36,697 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7319, 2.7748, 2.3975, 3.9553, 3.0437, 3.9961, 1.2905, 3.1667], device='cuda:3'), covar=tensor([0.1370, 0.0619, 0.1099, 0.0090, 0.0187, 0.0324, 0.1588, 0.0600], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0146, 0.0168, 0.0103, 0.0177, 0.0194, 0.0170, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 18:32:39,006 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:45,142 INFO [train.py:904] (3/8) Epoch 7, batch 9200, loss[loss=0.2222, simple_loss=0.3155, pruned_loss=0.06445, over 16695.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2823, pruned_loss=0.05225, over 3083972.29 frames. ], batch size: 134, lr: 9.47e-03, grad_scale: 8.0 2023-04-28 18:33:01,856 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:34:22,191 INFO [train.py:904] (3/8) Epoch 7, batch 9250, loss[loss=0.194, simple_loss=0.2805, pruned_loss=0.05371, over 16136.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2815, pruned_loss=0.05236, over 3063377.37 frames. ], batch size: 165, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:34:26,934 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1352, 3.2786, 3.5766, 3.5722, 3.5505, 3.3233, 3.3771, 3.4658], device='cuda:3'), covar=tensor([0.0376, 0.0755, 0.0519, 0.0545, 0.0582, 0.0535, 0.0814, 0.0464], device='cuda:3'), in_proj_covar=tensor([0.0261, 0.0267, 0.0266, 0.0256, 0.0305, 0.0287, 0.0373, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 18:34:34,249 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3348, 3.4592, 1.6646, 3.6134, 2.3354, 3.5771, 1.8974, 2.6342], device='cuda:3'), covar=tensor([0.0190, 0.0267, 0.1665, 0.0116, 0.0800, 0.0439, 0.1590, 0.0660], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0148, 0.0176, 0.0091, 0.0156, 0.0180, 0.0185, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 18:35:01,738 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:36:01,187 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.717e+02 3.344e+02 3.929e+02 8.889e+02, threshold=6.688e+02, percent-clipped=4.0 2023-04-28 18:36:13,551 INFO [train.py:904] (3/8) Epoch 7, batch 9300, loss[loss=0.1804, simple_loss=0.2754, pruned_loss=0.04271, over 16477.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2802, pruned_loss=0.05165, over 3066331.51 frames. ], batch size: 147, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:37:59,027 INFO [train.py:904] (3/8) Epoch 7, batch 9350, loss[loss=0.1767, simple_loss=0.2603, pruned_loss=0.0466, over 12112.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2799, pruned_loss=0.05154, over 3054305.03 frames. ], batch size: 248, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:38:18,320 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:39:31,294 INFO [optim.py:368] (3/8) 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 18:39:40,129 INFO [train.py:904] (3/8) Epoch 7, batch 9400, loss[loss=0.2011, simple_loss=0.2988, pruned_loss=0.05174, over 15354.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2792, pruned_loss=0.05147, over 3028579.27 frames. ], batch size: 190, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:40:10,319 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3049, 4.3287, 4.8010, 4.7737, 4.7271, 4.4030, 4.3830, 4.2384], device='cuda:3'), covar=tensor([0.0270, 0.0379, 0.0317, 0.0388, 0.0404, 0.0293, 0.0742, 0.0405], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0257, 0.0255, 0.0246, 0.0290, 0.0276, 0.0358, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 18:40:14,508 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 18:40:18,698 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:40:36,533 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:40:49,793 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 18:41:20,563 INFO [train.py:904] (3/8) Epoch 7, batch 9450, loss[loss=0.1933, simple_loss=0.2801, pruned_loss=0.05324, over 16872.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2811, pruned_loss=0.05181, over 3032392.92 frames. ], batch size: 116, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:42:14,166 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:42:50,912 INFO [optim.py:368] (3/8) 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 18:43:02,097 INFO [train.py:904] (3/8) Epoch 7, batch 9500, loss[loss=0.1876, simple_loss=0.27, pruned_loss=0.05259, over 12650.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2801, pruned_loss=0.05114, over 3032670.96 frames. ], batch size: 246, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:43:35,210 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0997, 2.8838, 2.7528, 1.9545, 2.6539, 2.0606, 2.7758, 2.9160], device='cuda:3'), covar=tensor([0.0241, 0.0526, 0.0439, 0.1405, 0.0597, 0.0852, 0.0498, 0.0590], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0122, 0.0151, 0.0137, 0.0130, 0.0122, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 18:43:45,443 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 18:44:04,280 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 18:44:45,656 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2274, 3.3991, 3.4527, 1.5979, 3.7247, 3.6770, 2.8004, 2.8621], device='cuda:3'), covar=tensor([0.0807, 0.0156, 0.0155, 0.1321, 0.0051, 0.0099, 0.0424, 0.0396], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0093, 0.0079, 0.0138, 0.0065, 0.0082, 0.0116, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 18:44:47,823 INFO [train.py:904] (3/8) Epoch 7, batch 9550, loss[loss=0.1958, simple_loss=0.2847, pruned_loss=0.05347, over 16510.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2804, pruned_loss=0.05138, over 3046212.49 frames. ], batch size: 68, lr: 9.44e-03, grad_scale: 4.0 2023-04-28 18:45:18,574 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:46:04,322 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6854, 2.7327, 1.5578, 2.8250, 2.0838, 2.8088, 1.7975, 2.3995], device='cuda:3'), covar=tensor([0.0186, 0.0335, 0.1521, 0.0149, 0.0724, 0.0469, 0.1361, 0.0563], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0147, 0.0175, 0.0090, 0.0156, 0.0178, 0.0183, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 18:46:18,719 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.839e+02 3.500e+02 4.170e+02 7.519e+02, threshold=6.999e+02, percent-clipped=3.0 2023-04-28 18:46:26,838 INFO [train.py:904] (3/8) Epoch 7, batch 9600, loss[loss=0.2033, simple_loss=0.2805, pruned_loss=0.06304, over 12271.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2814, pruned_loss=0.05195, over 3030427.33 frames. ], batch size: 247, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:48:15,878 INFO [train.py:904] (3/8) Epoch 7, batch 9650, loss[loss=0.2008, simple_loss=0.278, pruned_loss=0.06186, over 12095.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2836, pruned_loss=0.0526, over 3034133.92 frames. ], batch size: 247, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:49:43,136 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 18:49:55,419 INFO [optim.py:368] (3/8) 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 18:50:06,082 INFO [train.py:904] (3/8) Epoch 7, batch 9700, loss[loss=0.2094, simple_loss=0.2934, pruned_loss=0.0627, over 15515.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2828, pruned_loss=0.05212, over 3057228.47 frames. ], batch size: 190, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:50:33,465 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:51:47,796 INFO [train.py:904] (3/8) Epoch 7, batch 9750, loss[loss=0.197, simple_loss=0.2751, pruned_loss=0.05945, over 12097.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2818, pruned_loss=0.05238, over 3053988.52 frames. ], batch size: 246, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:53:01,468 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-28 18:53:18,480 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 7, batch 9800, loss[loss=0.2014, simple_loss=0.3042, pruned_loss=0.04928, over 16163.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.283, pruned_loss=0.05178, over 3079570.13 frames. ], batch size: 165, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:54:22,919 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2593, 3.5000, 3.4979, 2.5710, 3.2569, 3.3522, 3.4496, 1.9722], device='cuda:3'), covar=tensor([0.0322, 0.0021, 0.0026, 0.0202, 0.0060, 0.0065, 0.0038, 0.0333], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0056, 0.0059, 0.0117, 0.0066, 0.0073, 0.0066, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 18:55:12,739 INFO [train.py:904] (3/8) Epoch 7, batch 9850, loss[loss=0.1935, simple_loss=0.2854, pruned_loss=0.05078, over 15436.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2837, pruned_loss=0.05113, over 3082063.93 frames. ], batch size: 190, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:55:43,253 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:56:38,804 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7808, 3.0569, 2.9517, 4.8736, 3.9304, 4.5568, 1.4696, 3.4197], device='cuda:3'), covar=tensor([0.1291, 0.0566, 0.0855, 0.0070, 0.0146, 0.0267, 0.1463, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0145, 0.0167, 0.0102, 0.0170, 0.0192, 0.0169, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 18:56:54,659 INFO [optim.py:368] (3/8) 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 18:57:04,338 INFO [train.py:904] (3/8) Epoch 7, batch 9900, loss[loss=0.22, simple_loss=0.3119, pruned_loss=0.06404, over 16330.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2838, pruned_loss=0.05125, over 3068253.73 frames. ], batch size: 146, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:57:33,006 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:57:35,117 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3018, 3.5180, 1.8899, 3.7130, 2.4907, 3.6612, 1.8388, 2.7039], device='cuda:3'), covar=tensor([0.0190, 0.0287, 0.1619, 0.0090, 0.0797, 0.0414, 0.1681, 0.0644], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0149, 0.0177, 0.0091, 0.0160, 0.0180, 0.0188, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 18:58:35,204 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 18:59:01,223 INFO [train.py:904] (3/8) Epoch 7, batch 9950, loss[loss=0.1727, simple_loss=0.2706, pruned_loss=0.0374, over 16498.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.286, pruned_loss=0.05181, over 3053563.94 frames. ], batch size: 75, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 19:00:47,806 INFO [optim.py:368] (3/8) 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 19:01:00,566 INFO [train.py:904] (3/8) Epoch 7, batch 10000, loss[loss=0.1831, simple_loss=0.2688, pruned_loss=0.04867, over 12577.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2841, pruned_loss=0.05124, over 3057835.07 frames. ], batch size: 248, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:01:29,857 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:02:19,872 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9632, 3.9407, 4.3591, 4.3484, 4.3225, 4.0555, 4.0774, 3.9942], device='cuda:3'), covar=tensor([0.0256, 0.0446, 0.0325, 0.0390, 0.0399, 0.0275, 0.0643, 0.0346], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0261, 0.0259, 0.0249, 0.0293, 0.0277, 0.0358, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 19:02:40,827 INFO [train.py:904] (3/8) Epoch 7, batch 10050, loss[loss=0.1945, simple_loss=0.2907, pruned_loss=0.04919, over 16218.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2836, pruned_loss=0.05095, over 3049976.72 frames. ], batch size: 165, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:03:04,142 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:03:56,367 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 19:04:09,794 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7913, 2.1848, 2.3177, 4.5060, 2.0647, 2.7761, 2.2697, 2.4813], device='cuda:3'), covar=tensor([0.0663, 0.2754, 0.1662, 0.0275, 0.3425, 0.1700, 0.2474, 0.2825], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0335, 0.0289, 0.0303, 0.0377, 0.0361, 0.0303, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:04:12,142 INFO [train.py:904] (3/8) Epoch 7, batch 10100, loss[loss=0.1673, simple_loss=0.2607, pruned_loss=0.03694, over 16899.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2835, pruned_loss=0.05096, over 3063504.98 frames. ], batch size: 96, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:04:55,629 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1440, 3.2842, 3.5797, 3.5538, 3.5661, 3.3089, 3.3609, 3.4137], device='cuda:3'), covar=tensor([0.0445, 0.0946, 0.0471, 0.0550, 0.0488, 0.0507, 0.0802, 0.0471], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0260, 0.0255, 0.0246, 0.0289, 0.0274, 0.0355, 0.0225], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 19:05:06,985 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:05:54,944 INFO [train.py:904] (3/8) Epoch 8, batch 0, loss[loss=0.2303, simple_loss=0.3218, pruned_loss=0.06944, over 17128.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3218, pruned_loss=0.06944, over 17128.00 frames. ], batch size: 49, lr: 8.86e-03, grad_scale: 8.0 2023-04-28 19:05:54,945 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 19:06:02,580 INFO [train.py:938] (3/8) Epoch 8, validation: loss=0.1627, simple_loss=0.2663, pruned_loss=0.02958, over 944034.00 frames. 2023-04-28 19:06:02,580 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 19:06:04,002 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:06:25,212 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 19:06:45,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9814, 4.9037, 5.4903, 5.4497, 5.4555, 5.0451, 4.9903, 4.7450], device='cuda:3'), covar=tensor([0.0239, 0.0433, 0.0326, 0.0408, 0.0408, 0.0263, 0.0753, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0254, 0.0262, 0.0260, 0.0248, 0.0294, 0.0278, 0.0361, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-28 19:06:55,582 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 3.480e+02 4.237e+02 5.161e+02 1.332e+03, threshold=8.475e+02, percent-clipped=15.0 2023-04-28 19:07:09,983 INFO [train.py:904] (3/8) Epoch 8, batch 50, loss[loss=0.2409, simple_loss=0.2971, pruned_loss=0.09233, over 16716.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3043, pruned_loss=0.0771, over 755256.36 frames. ], batch size: 83, lr: 8.86e-03, grad_scale: 1.0 2023-04-28 19:07:38,964 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 19:07:56,649 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3231, 4.3174, 4.2676, 3.5886, 4.1792, 1.6399, 3.9627, 3.8817], device='cuda:3'), covar=tensor([0.0097, 0.0083, 0.0138, 0.0358, 0.0100, 0.2338, 0.0137, 0.0250], device='cuda:3'), in_proj_covar=tensor([0.0103, 0.0091, 0.0139, 0.0126, 0.0105, 0.0159, 0.0123, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:08:17,891 INFO [train.py:904] (3/8) Epoch 8, batch 100, loss[loss=0.2151, simple_loss=0.3008, pruned_loss=0.06469, over 17072.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2959, pruned_loss=0.07131, over 1321344.39 frames. ], batch size: 53, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:08:21,301 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0827, 5.6437, 5.7038, 5.5684, 5.5476, 6.1250, 5.7185, 5.5204], device='cuda:3'), covar=tensor([0.0647, 0.1576, 0.1796, 0.1649, 0.2519, 0.0921, 0.1179, 0.2109], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0424, 0.0441, 0.0372, 0.0488, 0.0463, 0.0352, 0.0493], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 19:09:06,432 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 19:09:10,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0135, 4.1396, 4.5223, 3.2787, 4.0932, 4.3279, 4.1156, 2.6484], device='cuda:3'), covar=tensor([0.0299, 0.0033, 0.0019, 0.0206, 0.0048, 0.0038, 0.0035, 0.0275], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0062, 0.0062, 0.0120, 0.0067, 0.0075, 0.0067, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 19:09:20,972 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 19:09:23,827 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 150, loss[loss=0.2325, simple_loss=0.312, pruned_loss=0.07647, over 16591.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2933, pruned_loss=0.06935, over 1759743.41 frames. ], batch size: 62, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:10:33,447 INFO [train.py:904] (3/8) Epoch 8, batch 200, loss[loss=0.2053, simple_loss=0.2953, pruned_loss=0.05761, over 17117.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2898, pruned_loss=0.06617, over 2120089.24 frames. ], batch size: 47, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:11:03,043 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-28 19:11:36,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0661, 4.4196, 2.5368, 4.8084, 3.2200, 4.6988, 2.3954, 3.4087], device='cuda:3'), covar=tensor([0.0187, 0.0224, 0.1347, 0.0085, 0.0706, 0.0324, 0.1450, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0154, 0.0180, 0.0097, 0.0161, 0.0187, 0.0190, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:11:40,015 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 250, loss[loss=0.2294, simple_loss=0.32, pruned_loss=0.06934, over 16621.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2861, pruned_loss=0.06355, over 2399526.32 frames. ], batch size: 62, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:11:50,321 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 19:11:53,293 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2734, 3.4603, 2.0465, 3.4715, 2.5676, 3.5265, 1.9490, 2.7953], device='cuda:3'), covar=tensor([0.0174, 0.0265, 0.1179, 0.0166, 0.0611, 0.0461, 0.1244, 0.0485], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0154, 0.0179, 0.0097, 0.0161, 0.0187, 0.0190, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:12:31,238 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2912, 3.4402, 1.8830, 3.3967, 2.5867, 3.4839, 1.9943, 2.7743], device='cuda:3'), covar=tensor([0.0183, 0.0351, 0.1423, 0.0267, 0.0697, 0.0551, 0.1276, 0.0543], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0154, 0.0179, 0.0097, 0.0161, 0.0188, 0.0189, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:12:47,162 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:12:52,134 INFO [train.py:904] (3/8) Epoch 8, batch 300, loss[loss=0.1606, simple_loss=0.2434, pruned_loss=0.03894, over 17003.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2824, pruned_loss=0.06208, over 2598947.82 frames. ], batch size: 41, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:13:38,974 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 8, batch 350, loss[loss=0.2132, simple_loss=0.2859, pruned_loss=0.07021, over 16445.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2811, pruned_loss=0.06047, over 2765604.83 frames. ], batch size: 68, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:14:16,252 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3010, 3.8954, 3.6938, 1.8803, 3.0578, 2.4203, 3.6705, 3.8328], device='cuda:3'), covar=tensor([0.0256, 0.0620, 0.0487, 0.1685, 0.0703, 0.0896, 0.0623, 0.0903], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0130, 0.0153, 0.0139, 0.0132, 0.0123, 0.0133, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:15:10,461 INFO [train.py:904] (3/8) Epoch 8, batch 400, loss[loss=0.1851, simple_loss=0.2705, pruned_loss=0.04981, over 16833.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2796, pruned_loss=0.06041, over 2892887.87 frames. ], batch size: 42, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:16:17,821 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.026e+02 2.889e+02 3.429e+02 4.074e+02 1.363e+03, threshold=6.859e+02, percent-clipped=4.0 2023-04-28 19:16:20,152 INFO [train.py:904] (3/8) Epoch 8, batch 450, loss[loss=0.1891, simple_loss=0.2673, pruned_loss=0.05545, over 16243.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2776, pruned_loss=0.05943, over 2974488.79 frames. ], batch size: 164, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:25,763 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8199, 5.2202, 4.9250, 4.9204, 4.6717, 4.5918, 4.6408, 5.2974], device='cuda:3'), covar=tensor([0.0952, 0.0772, 0.0996, 0.0640, 0.0767, 0.0960, 0.0943, 0.0717], device='cuda:3'), in_proj_covar=tensor([0.0470, 0.0598, 0.0501, 0.0404, 0.0383, 0.0397, 0.0498, 0.0445], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:17:27,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9127, 4.6124, 4.8830, 5.1189, 5.3004, 4.5895, 5.2569, 5.2380], device='cuda:3'), covar=tensor([0.1194, 0.0967, 0.1372, 0.0577, 0.0445, 0.0746, 0.0443, 0.0477], device='cuda:3'), in_proj_covar=tensor([0.0474, 0.0589, 0.0725, 0.0587, 0.0445, 0.0446, 0.0458, 0.0511], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:17:28,362 INFO [train.py:904] (3/8) Epoch 8, batch 500, loss[loss=0.2009, simple_loss=0.2915, pruned_loss=0.05512, over 17055.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2772, pruned_loss=0.0597, over 3044848.57 frames. ], batch size: 53, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:54,860 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0975, 5.6785, 5.7830, 5.6445, 5.6611, 6.1942, 5.7869, 5.5488], device='cuda:3'), covar=tensor([0.0667, 0.1622, 0.1428, 0.1726, 0.2548, 0.0874, 0.1084, 0.1937], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0448, 0.0464, 0.0392, 0.0517, 0.0487, 0.0370, 0.0525], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 19:18:33,694 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.499e+02 3.026e+02 3.865e+02 9.130e+02, threshold=6.053e+02, percent-clipped=1.0 2023-04-28 19:18:37,347 INFO [train.py:904] (3/8) Epoch 8, batch 550, loss[loss=0.1783, simple_loss=0.2704, pruned_loss=0.04305, over 17182.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2766, pruned_loss=0.05999, over 3097738.17 frames. ], batch size: 46, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:18:56,783 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:35,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5810, 3.4112, 4.1818, 2.9518, 3.7300, 4.1059, 3.8966, 2.3853], device='cuda:3'), covar=tensor([0.0336, 0.0154, 0.0028, 0.0215, 0.0051, 0.0059, 0.0045, 0.0308], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0063, 0.0063, 0.0117, 0.0067, 0.0075, 0.0068, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 19:19:41,707 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:45,779 INFO [train.py:904] (3/8) Epoch 8, batch 600, loss[loss=0.2258, simple_loss=0.2898, pruned_loss=0.0809, over 16322.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2763, pruned_loss=0.05997, over 3142526.15 frames. ], batch size: 165, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:20:19,021 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:34,092 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:45,947 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 8, batch 650, loss[loss=0.2208, simple_loss=0.28, pruned_loss=0.08081, over 16736.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2749, pruned_loss=0.05874, over 3192060.65 frames. ], batch size: 89, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:21:37,502 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:21:56,355 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 19:22:01,894 INFO [train.py:904] (3/8) Epoch 8, batch 700, loss[loss=0.197, simple_loss=0.2695, pruned_loss=0.06221, over 16470.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2745, pruned_loss=0.05853, over 3223359.97 frames. ], batch size: 68, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:23:05,954 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 2.585e+02 3.182e+02 3.748e+02 9.363e+02, threshold=6.364e+02, percent-clipped=4.0 2023-04-28 19:23:08,640 INFO [train.py:904] (3/8) Epoch 8, batch 750, loss[loss=0.1735, simple_loss=0.2562, pruned_loss=0.04537, over 16828.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2747, pruned_loss=0.0587, over 3243363.62 frames. ], batch size: 42, lr: 8.81e-03, grad_scale: 2.0 2023-04-28 19:24:10,638 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0513, 4.4605, 3.4637, 2.5302, 3.1328, 2.4981, 4.7527, 4.1636], device='cuda:3'), covar=tensor([0.2097, 0.0564, 0.1182, 0.1910, 0.2168, 0.1651, 0.0316, 0.0884], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0250, 0.0276, 0.0262, 0.0270, 0.0213, 0.0255, 0.0279], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:24:17,945 INFO [train.py:904] (3/8) Epoch 8, batch 800, loss[loss=0.1854, simple_loss=0.2548, pruned_loss=0.05804, over 16713.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2753, pruned_loss=0.05922, over 3252870.05 frames. ], batch size: 124, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:24:58,193 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2152, 4.4130, 4.6516, 2.1638, 4.8629, 4.9496, 3.4250, 3.8962], device='cuda:3'), covar=tensor([0.0584, 0.0117, 0.0130, 0.1026, 0.0043, 0.0064, 0.0307, 0.0267], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0094, 0.0082, 0.0140, 0.0068, 0.0089, 0.0120, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 19:25:23,720 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 850, loss[loss=0.1815, simple_loss=0.2557, pruned_loss=0.05368, over 16812.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2739, pruned_loss=0.05832, over 3264543.01 frames. ], batch size: 102, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:49,770 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:26:13,342 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2843, 4.6217, 4.3779, 4.4578, 4.0700, 4.0687, 4.1937, 4.6345], device='cuda:3'), covar=tensor([0.0906, 0.0841, 0.0983, 0.0566, 0.0758, 0.1360, 0.0837, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0478, 0.0611, 0.0510, 0.0409, 0.0389, 0.0402, 0.0506, 0.0451], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:26:18,902 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 19:26:32,698 INFO [train.py:904] (3/8) Epoch 8, batch 900, loss[loss=0.2004, simple_loss=0.2724, pruned_loss=0.06416, over 16709.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2715, pruned_loss=0.05712, over 3284926.22 frames. ], batch size: 89, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:26:39,976 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8611, 4.0567, 2.1599, 4.5080, 2.7931, 4.4568, 2.1931, 3.1610], device='cuda:3'), covar=tensor([0.0219, 0.0317, 0.1545, 0.0123, 0.0853, 0.0384, 0.1493, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0158, 0.0179, 0.0100, 0.0162, 0.0194, 0.0190, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:26:47,061 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0942, 3.4181, 3.4060, 1.6639, 3.6389, 3.6365, 2.9428, 2.6938], device='cuda:3'), covar=tensor([0.0899, 0.0138, 0.0178, 0.1168, 0.0060, 0.0113, 0.0389, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0093, 0.0082, 0.0139, 0.0067, 0.0089, 0.0118, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 19:27:00,518 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:13,331 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 8, batch 950, loss[loss=0.1817, simple_loss=0.2577, pruned_loss=0.05278, over 15574.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2712, pruned_loss=0.05671, over 3301797.54 frames. ], batch size: 190, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:28:15,865 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3221, 5.6820, 5.3846, 5.4808, 5.0542, 5.0122, 5.1552, 5.7509], device='cuda:3'), covar=tensor([0.0967, 0.0794, 0.1135, 0.0574, 0.0777, 0.0721, 0.0805, 0.0754], device='cuda:3'), in_proj_covar=tensor([0.0480, 0.0611, 0.0512, 0.0411, 0.0392, 0.0404, 0.0511, 0.0453], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:28:52,161 INFO [train.py:904] (3/8) Epoch 8, batch 1000, loss[loss=0.1988, simple_loss=0.2878, pruned_loss=0.05488, over 17261.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2708, pruned_loss=0.05689, over 3304490.38 frames. ], batch size: 52, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:29:09,305 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6303, 2.5627, 2.3790, 3.7913, 3.1139, 3.8982, 1.4884, 2.7487], device='cuda:3'), covar=tensor([0.1346, 0.0609, 0.1062, 0.0114, 0.0230, 0.0334, 0.1381, 0.0773], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0152, 0.0173, 0.0113, 0.0191, 0.0204, 0.0171, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 19:29:13,319 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7417, 3.9808, 4.3835, 3.0284, 3.8855, 4.3034, 3.9733, 2.5641], device='cuda:3'), covar=tensor([0.0338, 0.0031, 0.0025, 0.0222, 0.0045, 0.0047, 0.0041, 0.0302], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0064, 0.0063, 0.0119, 0.0068, 0.0077, 0.0070, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 19:29:58,311 INFO [optim.py:368] (3/8) 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 19:30:01,626 INFO [train.py:904] (3/8) Epoch 8, batch 1050, loss[loss=0.2069, simple_loss=0.2746, pruned_loss=0.06962, over 16867.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2705, pruned_loss=0.05705, over 3311671.38 frames. ], batch size: 96, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:30:40,231 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:31:10,580 INFO [train.py:904] (3/8) Epoch 8, batch 1100, loss[loss=0.1651, simple_loss=0.2487, pruned_loss=0.04078, over 16986.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2702, pruned_loss=0.05676, over 3316374.53 frames. ], batch size: 41, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:03,002 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 8, batch 1150, loss[loss=0.1945, simple_loss=0.2669, pruned_loss=0.0611, over 16514.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2692, pruned_loss=0.05622, over 3314064.82 frames. ], batch size: 68, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:48,874 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6520, 2.2128, 2.2943, 4.3720, 1.9504, 2.7352, 2.3120, 2.4042], device='cuda:3'), covar=tensor([0.0808, 0.3028, 0.1868, 0.0366, 0.3583, 0.1898, 0.2607, 0.3018], device='cuda:3'), in_proj_covar=tensor([0.0348, 0.0361, 0.0305, 0.0325, 0.0398, 0.0402, 0.0326, 0.0431], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:33:26,639 INFO [train.py:904] (3/8) Epoch 8, batch 1200, loss[loss=0.1914, simple_loss=0.2644, pruned_loss=0.05914, over 16465.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2683, pruned_loss=0.05568, over 3308792.60 frames. ], batch size: 146, lr: 8.79e-03, grad_scale: 8.0 2023-04-28 19:33:33,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3975, 5.3478, 5.0465, 4.5007, 5.1090, 2.1351, 4.8732, 5.1623], device='cuda:3'), covar=tensor([0.0050, 0.0047, 0.0127, 0.0334, 0.0066, 0.1821, 0.0102, 0.0127], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0101, 0.0152, 0.0145, 0.0117, 0.0165, 0.0138, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:33:50,175 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5219, 3.0285, 2.8655, 1.8960, 2.5690, 2.1916, 3.0869, 3.0705], device='cuda:3'), covar=tensor([0.0287, 0.0714, 0.0574, 0.1650, 0.0851, 0.0902, 0.0637, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0136, 0.0155, 0.0140, 0.0135, 0.0124, 0.0136, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:33:53,177 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:33:56,990 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 8, batch 1250, loss[loss=0.2065, simple_loss=0.2819, pruned_loss=0.06558, over 15583.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2693, pruned_loss=0.05594, over 3302170.78 frames. ], batch size: 191, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:34:42,234 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6381, 2.6940, 2.3020, 2.5423, 3.1168, 2.9241, 3.5151, 3.3145], device='cuda:3'), covar=tensor([0.0055, 0.0214, 0.0280, 0.0249, 0.0123, 0.0194, 0.0164, 0.0123], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0184, 0.0180, 0.0180, 0.0179, 0.0187, 0.0183, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:34:58,451 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:06,129 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:41,551 INFO [train.py:904] (3/8) Epoch 8, batch 1300, loss[loss=0.1912, simple_loss=0.2681, pruned_loss=0.05717, over 16373.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2685, pruned_loss=0.05566, over 3300965.33 frames. ], batch size: 146, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:36:30,343 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 19:36:52,118 INFO [train.py:904] (3/8) Epoch 8, batch 1350, loss[loss=0.2121, simple_loss=0.2785, pruned_loss=0.07287, over 16864.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2684, pruned_loss=0.05538, over 3300116.09 frames. ], batch size: 102, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:36:52,663 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1921, 3.4475, 3.2782, 1.9293, 2.9085, 2.4192, 3.6116, 3.4520], device='cuda:3'), covar=tensor([0.0199, 0.0581, 0.0615, 0.1576, 0.0684, 0.0818, 0.0449, 0.0714], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0135, 0.0154, 0.0139, 0.0133, 0.0123, 0.0135, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:38:01,922 INFO [train.py:904] (3/8) Epoch 8, batch 1400, loss[loss=0.2069, simple_loss=0.2721, pruned_loss=0.07088, over 16864.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2688, pruned_loss=0.0556, over 3302048.32 frames. ], batch size: 90, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:38:09,938 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 19:38:47,781 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 19:39:11,104 INFO [train.py:904] (3/8) Epoch 8, batch 1450, loss[loss=0.2099, simple_loss=0.2674, pruned_loss=0.07618, over 16750.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2686, pruned_loss=0.05572, over 3315870.06 frames. ], batch size: 89, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:39:47,284 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6796, 3.0039, 2.5312, 3.9823, 3.5179, 4.0594, 1.5615, 2.8647], device='cuda:3'), covar=tensor([0.1287, 0.0473, 0.0957, 0.0134, 0.0218, 0.0362, 0.1265, 0.0727], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0153, 0.0173, 0.0116, 0.0193, 0.0205, 0.0171, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 19:40:20,581 INFO [train.py:904] (3/8) Epoch 8, batch 1500, loss[loss=0.1789, simple_loss=0.2623, pruned_loss=0.04771, over 17268.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2688, pruned_loss=0.05624, over 3314199.73 frames. ], batch size: 45, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:24,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3754, 4.3703, 4.3389, 3.8231, 4.3358, 1.6762, 4.0981, 4.0294], device='cuda:3'), covar=tensor([0.0087, 0.0069, 0.0123, 0.0285, 0.0075, 0.2155, 0.0124, 0.0166], device='cuda:3'), in_proj_covar=tensor([0.0115, 0.0103, 0.0156, 0.0149, 0.0119, 0.0168, 0.0141, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:40:52,057 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:40:52,131 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8685, 4.1071, 2.1191, 4.4642, 2.8379, 4.4480, 2.1581, 3.2038], device='cuda:3'), covar=tensor([0.0175, 0.0261, 0.1490, 0.0120, 0.0793, 0.0395, 0.1479, 0.0571], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0159, 0.0180, 0.0104, 0.0162, 0.0198, 0.0189, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:41:25,445 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 1550, loss[loss=0.2204, simple_loss=0.2915, pruned_loss=0.07469, over 16171.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.27, pruned_loss=0.05724, over 3315561.62 frames. ], batch size: 164, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:41:44,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0074, 4.3301, 2.4296, 4.6230, 3.0748, 4.7473, 2.3273, 3.2807], device='cuda:3'), covar=tensor([0.0195, 0.0227, 0.1315, 0.0160, 0.0681, 0.0293, 0.1409, 0.0545], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0160, 0.0179, 0.0104, 0.0162, 0.0198, 0.0189, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 19:41:58,083 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:42:38,757 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5765, 2.5489, 2.0699, 2.2312, 2.9748, 2.7555, 3.4582, 3.1460], device='cuda:3'), covar=tensor([0.0051, 0.0246, 0.0280, 0.0264, 0.0142, 0.0210, 0.0117, 0.0129], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0189, 0.0184, 0.0183, 0.0184, 0.0191, 0.0188, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:42:39,324 INFO [train.py:904] (3/8) Epoch 8, batch 1600, loss[loss=0.2178, simple_loss=0.3025, pruned_loss=0.06658, over 17034.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2722, pruned_loss=0.058, over 3310112.73 frames. ], batch size: 50, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:43:20,416 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:43:44,410 INFO [optim.py:368] (3/8) 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 19:43:47,306 INFO [train.py:904] (3/8) Epoch 8, batch 1650, loss[loss=0.1837, simple_loss=0.2777, pruned_loss=0.04484, over 17084.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2726, pruned_loss=0.05762, over 3304953.98 frames. ], batch size: 49, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:44:58,249 INFO [train.py:904] (3/8) Epoch 8, batch 1700, loss[loss=0.2025, simple_loss=0.2905, pruned_loss=0.05725, over 16532.00 frames. ], tot_loss[loss=0.195, simple_loss=0.274, pruned_loss=0.05797, over 3302884.43 frames. ], batch size: 68, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:45:44,092 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:45:47,622 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 19:46:05,115 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 1750, loss[loss=0.2037, simple_loss=0.2806, pruned_loss=0.06341, over 16865.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2745, pruned_loss=0.05762, over 3310507.22 frames. ], batch size: 102, lr: 8.75e-03, grad_scale: 8.0 2023-04-28 19:46:12,521 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 19:46:34,816 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3486, 4.3368, 4.2664, 3.8177, 4.2942, 1.7721, 4.0721, 4.0026], device='cuda:3'), covar=tensor([0.0077, 0.0061, 0.0114, 0.0240, 0.0064, 0.1948, 0.0107, 0.0159], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0102, 0.0153, 0.0147, 0.0118, 0.0164, 0.0139, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:46:38,035 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8411, 2.7615, 2.3870, 2.5350, 3.1086, 2.8561, 3.6968, 3.2816], device='cuda:3'), covar=tensor([0.0044, 0.0217, 0.0255, 0.0255, 0.0130, 0.0210, 0.0101, 0.0131], device='cuda:3'), in_proj_covar=tensor([0.0118, 0.0185, 0.0181, 0.0181, 0.0182, 0.0188, 0.0186, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:46:43,797 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0699, 5.0382, 5.5765, 5.5449, 5.5134, 5.1735, 5.1200, 4.8227], device='cuda:3'), covar=tensor([0.0232, 0.0371, 0.0261, 0.0379, 0.0402, 0.0270, 0.0778, 0.0372], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0312, 0.0310, 0.0296, 0.0353, 0.0330, 0.0433, 0.0267], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 19:46:49,372 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5306, 2.1010, 2.3583, 4.2769, 2.1318, 2.7601, 2.2626, 2.3537], device='cuda:3'), covar=tensor([0.0809, 0.2905, 0.1645, 0.0332, 0.3135, 0.1792, 0.2549, 0.2559], device='cuda:3'), in_proj_covar=tensor([0.0346, 0.0362, 0.0303, 0.0324, 0.0392, 0.0402, 0.0325, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:46:50,155 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:46:51,803 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 19:47:16,117 INFO [train.py:904] (3/8) Epoch 8, batch 1800, loss[loss=0.1867, simple_loss=0.2739, pruned_loss=0.04972, over 16826.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2762, pruned_loss=0.05773, over 3315757.08 frames. ], batch size: 42, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:47:45,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0883, 1.9295, 2.3533, 3.0551, 2.7965, 3.5082, 2.2041, 3.3701], device='cuda:3'), covar=tensor([0.0123, 0.0259, 0.0200, 0.0154, 0.0165, 0.0080, 0.0258, 0.0071], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0158, 0.0145, 0.0144, 0.0150, 0.0106, 0.0157, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 19:48:15,323 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:48:24,990 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 1850, loss[loss=0.1601, simple_loss=0.247, pruned_loss=0.0366, over 17213.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2774, pruned_loss=0.05819, over 3311986.34 frames. ], batch size: 44, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:48:38,282 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2018, 4.1687, 4.6360, 4.6477, 4.6376, 4.2737, 4.3231, 4.1048], device='cuda:3'), covar=tensor([0.0316, 0.0429, 0.0396, 0.0402, 0.0442, 0.0330, 0.0789, 0.0589], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0307, 0.0308, 0.0292, 0.0349, 0.0327, 0.0427, 0.0264], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 19:49:35,219 INFO [train.py:904] (3/8) Epoch 8, batch 1900, loss[loss=0.168, simple_loss=0.2501, pruned_loss=0.04292, over 16857.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2768, pruned_loss=0.05765, over 3314177.39 frames. ], batch size: 42, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:49:39,729 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:49:51,975 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 19:50:16,599 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:50:41,095 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 1950, loss[loss=0.1747, simple_loss=0.2644, pruned_loss=0.04248, over 17272.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2763, pruned_loss=0.05697, over 3318331.11 frames. ], batch size: 45, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:51:14,225 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 19:51:21,211 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:51:50,195 INFO [train.py:904] (3/8) Epoch 8, batch 2000, loss[loss=0.1813, simple_loss=0.2577, pruned_loss=0.05239, over 15841.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2763, pruned_loss=0.05761, over 3316104.37 frames. ], batch size: 35, lr: 8.74e-03, grad_scale: 8.0 2023-04-28 19:51:55,331 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 19:52:58,786 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 2050, loss[loss=0.2182, simple_loss=0.2843, pruned_loss=0.07607, over 16718.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2765, pruned_loss=0.05826, over 3308022.67 frames. ], batch size: 89, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:53:19,030 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 19:53:36,660 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 19:54:09,330 INFO [train.py:904] (3/8) Epoch 8, batch 2100, loss[loss=0.2024, simple_loss=0.2741, pruned_loss=0.06532, over 16784.00 frames. ], tot_loss[loss=0.197, simple_loss=0.277, pruned_loss=0.05848, over 3310878.28 frames. ], batch size: 83, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:54:33,033 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0727, 1.7166, 2.3802, 2.7937, 2.6052, 3.2396, 1.7984, 3.1656], device='cuda:3'), covar=tensor([0.0119, 0.0293, 0.0197, 0.0178, 0.0179, 0.0108, 0.0306, 0.0086], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0158, 0.0145, 0.0146, 0.0151, 0.0107, 0.0158, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 19:54:38,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0794, 4.1228, 4.4813, 2.0294, 4.7969, 4.7220, 3.3615, 3.6176], device='cuda:3'), covar=tensor([0.0647, 0.0148, 0.0163, 0.1137, 0.0047, 0.0087, 0.0366, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0094, 0.0083, 0.0137, 0.0069, 0.0092, 0.0119, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 19:55:14,761 INFO [optim.py:368] (3/8) 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 19:55:16,489 INFO [train.py:904] (3/8) Epoch 8, batch 2150, loss[loss=0.2436, simple_loss=0.3097, pruned_loss=0.08875, over 16416.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2785, pruned_loss=0.05891, over 3309635.82 frames. ], batch size: 146, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:56:21,714 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:56:23,751 INFO [train.py:904] (3/8) Epoch 8, batch 2200, loss[loss=0.206, simple_loss=0.2808, pruned_loss=0.06559, over 16825.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2799, pruned_loss=0.05988, over 3318381.22 frames. ], batch size: 90, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:57:13,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7800, 1.6337, 2.1495, 2.5529, 2.6231, 2.5795, 1.7944, 2.7540], device='cuda:3'), covar=tensor([0.0082, 0.0270, 0.0186, 0.0160, 0.0126, 0.0110, 0.0273, 0.0069], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0159, 0.0145, 0.0147, 0.0152, 0.0106, 0.0159, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 19:57:20,377 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.689e+02 3.130e+02 3.780e+02 7.244e+02, threshold=6.259e+02, percent-clipped=1.0 2023-04-28 19:57:31,975 INFO [train.py:904] (3/8) Epoch 8, batch 2250, loss[loss=0.1998, simple_loss=0.2927, pruned_loss=0.05342, over 17142.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2804, pruned_loss=0.05977, over 3317352.27 frames. ], batch size: 49, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:57:46,705 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0340, 4.7653, 4.9871, 5.2672, 5.4311, 4.7051, 5.3429, 5.3363], device='cuda:3'), covar=tensor([0.1303, 0.0963, 0.1467, 0.0568, 0.0460, 0.0789, 0.0501, 0.0440], device='cuda:3'), in_proj_covar=tensor([0.0495, 0.0619, 0.0770, 0.0619, 0.0469, 0.0473, 0.0485, 0.0530], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:57:56,939 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:58:40,778 INFO [train.py:904] (3/8) Epoch 8, batch 2300, loss[loss=0.188, simple_loss=0.2723, pruned_loss=0.05182, over 16117.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2789, pruned_loss=0.05945, over 3320567.80 frames. ], batch size: 35, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:58:44,174 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:59:18,723 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1061, 5.5170, 5.7190, 5.4903, 5.5437, 6.0754, 5.6566, 5.3808], device='cuda:3'), covar=tensor([0.0808, 0.1799, 0.1707, 0.1871, 0.2711, 0.0974, 0.1092, 0.2098], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0466, 0.0481, 0.0410, 0.0536, 0.0512, 0.0386, 0.0550], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 19:59:24,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4892, 2.0792, 2.3254, 4.1404, 2.0719, 2.6910, 2.1375, 2.3465], device='cuda:3'), covar=tensor([0.0787, 0.2602, 0.1642, 0.0358, 0.3018, 0.1623, 0.2683, 0.2326], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0366, 0.0308, 0.0327, 0.0395, 0.0408, 0.0330, 0.0435], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 19:59:48,860 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.976e+02 2.910e+02 3.420e+02 4.127e+02 1.105e+03, threshold=6.841e+02, percent-clipped=4.0 2023-04-28 19:59:49,988 INFO [train.py:904] (3/8) Epoch 8, batch 2350, loss[loss=0.2187, simple_loss=0.3006, pruned_loss=0.0684, over 17028.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2803, pruned_loss=0.06039, over 3313423.47 frames. ], batch size: 55, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 20:00:25,372 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 20:00:58,115 INFO [train.py:904] (3/8) Epoch 8, batch 2400, loss[loss=0.2335, simple_loss=0.304, pruned_loss=0.08153, over 16771.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2826, pruned_loss=0.06114, over 3296962.55 frames. ], batch size: 134, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:01:20,743 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 20:02:04,599 INFO [optim.py:368] (3/8) 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 20:02:05,789 INFO [train.py:904] (3/8) Epoch 8, batch 2450, loss[loss=0.2171, simple_loss=0.2905, pruned_loss=0.07186, over 16891.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2837, pruned_loss=0.06151, over 3292965.50 frames. ], batch size: 96, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:02:52,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9020, 4.1146, 4.4165, 1.9881, 4.7954, 4.6990, 3.2993, 3.3303], device='cuda:3'), covar=tensor([0.0688, 0.0137, 0.0182, 0.1126, 0.0045, 0.0087, 0.0365, 0.0402], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0096, 0.0085, 0.0139, 0.0069, 0.0093, 0.0121, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 20:03:12,589 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:03:14,147 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 20:03:14,440 INFO [train.py:904] (3/8) Epoch 8, batch 2500, loss[loss=0.274, simple_loss=0.3345, pruned_loss=0.1068, over 11967.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2838, pruned_loss=0.0612, over 3284248.25 frames. ], batch size: 246, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:04:07,441 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:11,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8903, 4.2131, 4.4680, 4.4450, 4.4999, 4.1907, 3.8640, 4.0550], device='cuda:3'), covar=tensor([0.0593, 0.0612, 0.0545, 0.0667, 0.0633, 0.0545, 0.1438, 0.0604], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0306, 0.0306, 0.0292, 0.0347, 0.0324, 0.0426, 0.0262], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 20:04:18,852 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:22,384 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.774e+02 3.514e+02 4.177e+02 6.912e+02, threshold=7.028e+02, percent-clipped=4.0 2023-04-28 20:04:23,439 INFO [train.py:904] (3/8) Epoch 8, batch 2550, loss[loss=0.2087, simple_loss=0.2828, pruned_loss=0.06727, over 16874.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2827, pruned_loss=0.06051, over 3296228.82 frames. ], batch size: 109, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:04:49,476 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:04:59,685 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-28 20:05:29,132 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,310 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,987 INFO [train.py:904] (3/8) Epoch 8, batch 2600, loss[loss=0.1702, simple_loss=0.2678, pruned_loss=0.03626, over 17120.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2828, pruned_loss=0.06002, over 3296559.69 frames. ], batch size: 47, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:05:55,318 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:06:43,000 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 2650, loss[loss=0.198, simple_loss=0.2742, pruned_loss=0.06084, over 16472.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2827, pruned_loss=0.05969, over 3309544.85 frames. ], batch size: 146, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:23,271 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:07:50,958 INFO [train.py:904] (3/8) Epoch 8, batch 2700, loss[loss=0.1668, simple_loss=0.2573, pruned_loss=0.03817, over 16796.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2827, pruned_loss=0.05946, over 3317158.37 frames. ], batch size: 39, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:08:38,956 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7527, 4.8827, 5.1199, 4.9193, 4.8509, 5.5665, 5.1407, 4.8111], device='cuda:3'), covar=tensor([0.1175, 0.1757, 0.1852, 0.1560, 0.2568, 0.1004, 0.1334, 0.2335], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0464, 0.0480, 0.0406, 0.0536, 0.0513, 0.0387, 0.0545], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 20:08:46,236 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:08:57,113 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.547e+02 3.040e+02 3.898e+02 5.528e+02, threshold=6.080e+02, percent-clipped=0.0 2023-04-28 20:08:57,128 INFO [train.py:904] (3/8) Epoch 8, batch 2750, loss[loss=0.2203, simple_loss=0.2969, pruned_loss=0.07181, over 16881.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.283, pruned_loss=0.05918, over 3317635.11 frames. ], batch size: 116, lr: 8.69e-03, grad_scale: 4.0 2023-04-28 20:09:58,589 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 20:10:05,127 INFO [train.py:904] (3/8) Epoch 8, batch 2800, loss[loss=0.2203, simple_loss=0.3072, pruned_loss=0.06673, over 16613.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2821, pruned_loss=0.05854, over 3321220.78 frames. ], batch size: 62, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:10:19,787 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-28 20:10:48,047 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 20:11:12,916 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 20:11:14,609 INFO [optim.py:368] (3/8) 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 20:11:14,625 INFO [train.py:904] (3/8) Epoch 8, batch 2850, loss[loss=0.1656, simple_loss=0.2498, pruned_loss=0.04073, over 16817.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2813, pruned_loss=0.0579, over 3331876.25 frames. ], batch size: 39, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:11:34,886 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3894, 5.7649, 5.1600, 5.6552, 5.1102, 4.8374, 5.3456, 5.8062], device='cuda:3'), covar=tensor([0.2180, 0.1393, 0.2728, 0.0973, 0.1753, 0.1444, 0.1898, 0.1721], device='cuda:3'), in_proj_covar=tensor([0.0499, 0.0638, 0.0522, 0.0426, 0.0400, 0.0412, 0.0525, 0.0473], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:12:16,871 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:20,721 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:22,123 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3868, 1.4567, 2.0096, 2.2722, 2.4187, 2.2851, 1.5815, 2.3801], device='cuda:3'), covar=tensor([0.0106, 0.0264, 0.0152, 0.0183, 0.0136, 0.0128, 0.0257, 0.0068], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0162, 0.0146, 0.0150, 0.0157, 0.0111, 0.0160, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 20:12:23,980 INFO [train.py:904] (3/8) Epoch 8, batch 2900, loss[loss=0.1597, simple_loss=0.2419, pruned_loss=0.03879, over 17215.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2791, pruned_loss=0.05801, over 3334374.74 frames. ], batch size: 44, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:12:53,855 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0468, 4.9801, 4.9333, 4.5686, 4.4189, 4.9369, 4.9735, 4.5488], device='cuda:3'), covar=tensor([0.0548, 0.0354, 0.0240, 0.0255, 0.1108, 0.0339, 0.0310, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0284, 0.0280, 0.0255, 0.0317, 0.0285, 0.0194, 0.0326], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 20:12:57,291 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 20:13:28,888 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:13:31,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5230, 4.3020, 4.5493, 4.7468, 4.8542, 4.3664, 4.6735, 4.7912], device='cuda:3'), covar=tensor([0.1180, 0.0944, 0.1262, 0.0568, 0.0581, 0.0915, 0.1088, 0.0510], device='cuda:3'), in_proj_covar=tensor([0.0501, 0.0616, 0.0766, 0.0624, 0.0471, 0.0474, 0.0488, 0.0537], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:13:38,231 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 2950, loss[loss=0.1737, simple_loss=0.2657, pruned_loss=0.04083, over 17119.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2789, pruned_loss=0.05856, over 3337420.96 frames. ], batch size: 49, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:29,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8171, 4.1323, 4.3755, 1.8884, 4.6471, 4.5981, 3.1108, 3.5317], device='cuda:3'), covar=tensor([0.0696, 0.0127, 0.0211, 0.1092, 0.0053, 0.0100, 0.0355, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0094, 0.0084, 0.0136, 0.0069, 0.0094, 0.0118, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 20:14:46,251 INFO [train.py:904] (3/8) Epoch 8, batch 3000, loss[loss=0.1983, simple_loss=0.2883, pruned_loss=0.05412, over 17021.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.279, pruned_loss=0.05865, over 3338713.54 frames. ], batch size: 50, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:46,251 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 20:14:55,853 INFO [train.py:938] (3/8) Epoch 8, validation: loss=0.1462, simple_loss=0.2525, pruned_loss=0.01995, over 944034.00 frames. 2023-04-28 20:14:55,854 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 20:15:13,452 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:15:41,405 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1965, 5.4328, 5.1209, 4.9255, 4.2136, 5.2964, 5.3761, 4.7872], device='cuda:3'), covar=tensor([0.0774, 0.0409, 0.0385, 0.0311, 0.1818, 0.0362, 0.0232, 0.0645], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0284, 0.0280, 0.0253, 0.0316, 0.0284, 0.0193, 0.0324], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 20:15:45,961 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 8, batch 3050, loss[loss=0.2102, simple_loss=0.2799, pruned_loss=0.07023, over 16279.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2796, pruned_loss=0.05966, over 3325101.14 frames. ], batch size: 145, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:16:38,623 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:17:05,337 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7379, 4.5125, 4.7496, 4.9350, 5.0650, 4.5586, 4.9811, 4.9880], device='cuda:3'), covar=tensor([0.1228, 0.0914, 0.1250, 0.0565, 0.0553, 0.0831, 0.0809, 0.0586], device='cuda:3'), in_proj_covar=tensor([0.0501, 0.0617, 0.0769, 0.0625, 0.0472, 0.0473, 0.0488, 0.0536], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:17:07,565 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6778, 5.0280, 4.7499, 4.7966, 4.4763, 4.5131, 4.5354, 5.1346], device='cuda:3'), covar=tensor([0.0961, 0.0784, 0.0998, 0.0583, 0.0714, 0.0871, 0.0950, 0.0743], device='cuda:3'), in_proj_covar=tensor([0.0494, 0.0632, 0.0518, 0.0422, 0.0395, 0.0407, 0.0526, 0.0467], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:17:13,362 INFO [train.py:904] (3/8) Epoch 8, batch 3100, loss[loss=0.2028, simple_loss=0.2953, pruned_loss=0.05521, over 17061.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2793, pruned_loss=0.0599, over 3321985.91 frames. ], batch size: 53, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:17:45,306 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 20:18:21,290 INFO [optim.py:368] (3/8) 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 20:18:21,305 INFO [train.py:904] (3/8) Epoch 8, batch 3150, loss[loss=0.181, simple_loss=0.267, pruned_loss=0.04751, over 17168.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2781, pruned_loss=0.05919, over 3320943.95 frames. ], batch size: 46, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:19:15,103 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:23,396 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:32,372 INFO [train.py:904] (3/8) Epoch 8, batch 3200, loss[loss=0.1976, simple_loss=0.2713, pruned_loss=0.06197, over 15926.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2768, pruned_loss=0.05853, over 3318003.96 frames. ], batch size: 35, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:20:30,670 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:33,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2460, 3.7260, 3.6252, 2.0824, 2.9498, 2.5169, 3.7473, 3.6533], device='cuda:3'), covar=tensor([0.0211, 0.0601, 0.0503, 0.1478, 0.0721, 0.0832, 0.0517, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0142, 0.0154, 0.0139, 0.0135, 0.0123, 0.0137, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 20:20:40,579 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:41,301 INFO [train.py:904] (3/8) Epoch 8, batch 3250, loss[loss=0.152, simple_loss=0.2336, pruned_loss=0.0352, over 16821.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2766, pruned_loss=0.05816, over 3328125.39 frames. ], batch size: 39, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:20:42,357 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.591e+02 3.147e+02 3.802e+02 9.622e+02, threshold=6.293e+02, percent-clipped=5.0 2023-04-28 20:21:39,996 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5816, 2.4312, 1.9511, 2.2316, 2.9174, 2.7532, 3.4173, 3.1892], device='cuda:3'), covar=tensor([0.0053, 0.0263, 0.0358, 0.0296, 0.0149, 0.0227, 0.0141, 0.0133], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0185, 0.0182, 0.0183, 0.0183, 0.0187, 0.0190, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:21:52,434 INFO [train.py:904] (3/8) Epoch 8, batch 3300, loss[loss=0.208, simple_loss=0.2801, pruned_loss=0.068, over 16418.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2766, pruned_loss=0.05804, over 3327734.91 frames. ], batch size: 146, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:22:41,285 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:22:52,651 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1025, 5.0894, 4.8368, 3.5655, 4.9693, 1.6359, 4.5496, 4.7933], device='cuda:3'), covar=tensor([0.0116, 0.0101, 0.0204, 0.0737, 0.0104, 0.2803, 0.0182, 0.0289], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0107, 0.0158, 0.0154, 0.0124, 0.0168, 0.0144, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:23:01,501 INFO [train.py:904] (3/8) Epoch 8, batch 3350, loss[loss=0.1926, simple_loss=0.2819, pruned_loss=0.05161, over 17047.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2784, pruned_loss=0.05865, over 3329212.59 frames. ], batch size: 53, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:23:02,761 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.621e+02 3.247e+02 4.157e+02 8.305e+02, threshold=6.494e+02, percent-clipped=1.0 2023-04-28 20:23:27,942 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:33,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7543, 3.8975, 4.1606, 1.8956, 4.3929, 4.3137, 3.1851, 3.2732], device='cuda:3'), covar=tensor([0.0670, 0.0136, 0.0148, 0.1124, 0.0055, 0.0128, 0.0333, 0.0400], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0095, 0.0083, 0.0135, 0.0069, 0.0093, 0.0117, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 20:23:49,803 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:24:05,693 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 20:24:11,178 INFO [train.py:904] (3/8) Epoch 8, batch 3400, loss[loss=0.2073, simple_loss=0.2742, pruned_loss=0.07023, over 15789.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2792, pruned_loss=0.05909, over 3314631.14 frames. ], batch size: 35, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:24:36,638 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5790, 3.6878, 3.9223, 1.9083, 4.0552, 4.0112, 3.1074, 2.9970], device='cuda:3'), covar=tensor([0.0703, 0.0129, 0.0129, 0.1143, 0.0056, 0.0130, 0.0313, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0096, 0.0085, 0.0138, 0.0071, 0.0095, 0.0119, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 20:25:19,954 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5161, 2.5358, 2.1277, 2.2360, 2.8822, 2.7461, 3.4863, 3.1670], device='cuda:3'), covar=tensor([0.0068, 0.0232, 0.0317, 0.0285, 0.0164, 0.0239, 0.0170, 0.0135], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0185, 0.0181, 0.0182, 0.0182, 0.0187, 0.0190, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:25:21,710 INFO [train.py:904] (3/8) Epoch 8, batch 3450, loss[loss=0.1903, simple_loss=0.2587, pruned_loss=0.06097, over 16853.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2777, pruned_loss=0.05896, over 3301544.67 frames. ], batch size: 96, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:25:22,261 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1226, 4.5844, 3.3099, 2.5332, 3.1025, 2.4196, 4.7179, 4.1798], device='cuda:3'), covar=tensor([0.2038, 0.0433, 0.1223, 0.1885, 0.2285, 0.1639, 0.0297, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0257, 0.0277, 0.0263, 0.0283, 0.0212, 0.0256, 0.0287], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:25:22,841 INFO [optim.py:368] (3/8) 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 20:26:14,059 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6484, 3.8312, 1.7693, 3.9416, 2.8075, 3.9307, 1.9633, 2.9312], device='cuda:3'), covar=tensor([0.0167, 0.0238, 0.1426, 0.0145, 0.0545, 0.0412, 0.1260, 0.0461], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0163, 0.0178, 0.0110, 0.0164, 0.0205, 0.0189, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 20:26:30,765 INFO [train.py:904] (3/8) Epoch 8, batch 3500, loss[loss=0.1866, simple_loss=0.2786, pruned_loss=0.04724, over 17129.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.276, pruned_loss=0.05792, over 3304380.92 frames. ], batch size: 49, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:32,327 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:42,855 INFO [train.py:904] (3/8) Epoch 8, batch 3550, loss[loss=0.2097, simple_loss=0.2788, pruned_loss=0.07036, over 16287.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2746, pruned_loss=0.05744, over 3304280.33 frames. ], batch size: 165, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:43,958 INFO [optim.py:368] (3/8) 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 20:28:10,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0532, 4.6938, 4.4975, 5.0716, 5.3139, 4.6471, 5.1616, 5.2654], device='cuda:3'), covar=tensor([0.1158, 0.0885, 0.2353, 0.0945, 0.0698, 0.0920, 0.0814, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0508, 0.0627, 0.0783, 0.0639, 0.0481, 0.0484, 0.0495, 0.0550], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:28:51,887 INFO [train.py:904] (3/8) Epoch 8, batch 3600, loss[loss=0.197, simple_loss=0.2762, pruned_loss=0.05893, over 16498.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.273, pruned_loss=0.05662, over 3302379.32 frames. ], batch size: 68, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:28:53,508 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1058, 4.3096, 4.4854, 2.1697, 4.8103, 4.7318, 3.3565, 3.7654], device='cuda:3'), covar=tensor([0.0634, 0.0116, 0.0201, 0.0989, 0.0044, 0.0084, 0.0320, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0096, 0.0086, 0.0137, 0.0071, 0.0094, 0.0120, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 20:28:56,992 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 3650, loss[loss=0.1979, simple_loss=0.2637, pruned_loss=0.06607, over 16726.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2706, pruned_loss=0.05619, over 3313681.29 frames. ], batch size: 134, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:30:02,111 INFO [optim.py:368] (3/8) 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 20:30:27,845 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 3700, loss[loss=0.1823, simple_loss=0.2535, pruned_loss=0.05556, over 10948.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.269, pruned_loss=0.05744, over 3289721.52 frames. ], batch size: 246, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:31:38,734 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:32:12,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7496, 4.8836, 5.2573, 5.2266, 5.2484, 4.8495, 4.8075, 4.5506], device='cuda:3'), covar=tensor([0.0273, 0.0353, 0.0306, 0.0362, 0.0374, 0.0281, 0.0743, 0.0374], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0306, 0.0305, 0.0295, 0.0350, 0.0324, 0.0429, 0.0262], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 20:32:29,721 INFO [train.py:904] (3/8) Epoch 8, batch 3750, loss[loss=0.2384, simple_loss=0.2952, pruned_loss=0.09085, over 16698.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2703, pruned_loss=0.05915, over 3273438.11 frames. ], batch size: 124, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:32:30,686 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.639e+02 3.130e+02 3.703e+02 1.006e+03, threshold=6.260e+02, percent-clipped=3.0 2023-04-28 20:33:41,191 INFO [train.py:904] (3/8) Epoch 8, batch 3800, loss[loss=0.2243, simple_loss=0.3107, pruned_loss=0.06893, over 16989.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2726, pruned_loss=0.06136, over 3260976.56 frames. ], batch size: 55, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:33:46,991 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0299, 2.5455, 2.0066, 2.3537, 2.9616, 2.7635, 3.2287, 3.1724], device='cuda:3'), covar=tensor([0.0098, 0.0231, 0.0309, 0.0275, 0.0134, 0.0191, 0.0117, 0.0118], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0185, 0.0181, 0.0183, 0.0182, 0.0186, 0.0188, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:34:45,331 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:34:51,949 INFO [train.py:904] (3/8) Epoch 8, batch 3850, loss[loss=0.2089, simple_loss=0.2783, pruned_loss=0.06969, over 16629.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2727, pruned_loss=0.06192, over 3273412.18 frames. ], batch size: 134, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:53,141 INFO [optim.py:368] (3/8) 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 20:34:56,990 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-28 20:35:03,404 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:35:04,802 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4953, 2.0135, 2.2799, 4.1812, 2.0312, 2.5672, 2.1325, 2.1478], device='cuda:3'), covar=tensor([0.0802, 0.3227, 0.1735, 0.0363, 0.3432, 0.2082, 0.3109, 0.2869], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0373, 0.0311, 0.0330, 0.0399, 0.0421, 0.0334, 0.0443], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:35:20,348 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 20:35:52,590 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:01,444 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:02,285 INFO [train.py:904] (3/8) Epoch 8, batch 3900, loss[loss=0.1797, simple_loss=0.2565, pruned_loss=0.05146, over 16907.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2721, pruned_loss=0.06208, over 3273461.51 frames. ], batch size: 109, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:36:20,341 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:28,128 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:37:12,342 INFO [train.py:904] (3/8) Epoch 8, batch 3950, loss[loss=0.2209, simple_loss=0.2891, pruned_loss=0.07636, over 16273.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2722, pruned_loss=0.06291, over 3275283.37 frames. ], batch size: 165, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:37:14,091 INFO [optim.py:368] (3/8) 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 20:37:46,161 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 4000, loss[loss=0.2047, simple_loss=0.2721, pruned_loss=0.06861, over 16878.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2725, pruned_loss=0.0638, over 3273769.74 frames. ], batch size: 116, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:36,988 INFO [train.py:904] (3/8) Epoch 8, batch 4050, loss[loss=0.2027, simple_loss=0.2861, pruned_loss=0.05969, over 15561.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2728, pruned_loss=0.06227, over 3269475.26 frames. ], batch size: 190, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:38,164 INFO [optim.py:368] (3/8) 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 20:40:02,015 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 20:40:10,423 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.44 vs. limit=5.0 2023-04-28 20:40:19,671 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7930, 1.6764, 2.1485, 2.7349, 2.6922, 2.9709, 1.8058, 2.9872], device='cuda:3'), covar=tensor([0.0105, 0.0295, 0.0191, 0.0144, 0.0136, 0.0077, 0.0283, 0.0041], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0159, 0.0144, 0.0148, 0.0154, 0.0110, 0.0159, 0.0100], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 20:40:49,053 INFO [train.py:904] (3/8) Epoch 8, batch 4100, loss[loss=0.2728, simple_loss=0.3312, pruned_loss=0.1072, over 11677.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2741, pruned_loss=0.06121, over 3266352.28 frames. ], batch size: 246, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:42:02,416 INFO [train.py:904] (3/8) Epoch 8, batch 4150, loss[loss=0.2281, simple_loss=0.3132, pruned_loss=0.07146, over 15306.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2817, pruned_loss=0.06422, over 3240013.62 frames. ], batch size: 190, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:42:04,252 INFO [optim.py:368] (3/8) 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 20:42:39,002 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7977, 4.5787, 4.7435, 4.9437, 5.0833, 4.5194, 5.0640, 5.0626], device='cuda:3'), covar=tensor([0.1079, 0.0896, 0.1401, 0.0524, 0.0409, 0.0703, 0.0429, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0474, 0.0580, 0.0721, 0.0589, 0.0445, 0.0448, 0.0453, 0.0505], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:42:39,045 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0697, 5.0184, 4.7673, 4.2348, 5.0056, 1.7245, 4.7308, 4.8275], device='cuda:3'), covar=tensor([0.0055, 0.0042, 0.0112, 0.0290, 0.0044, 0.2139, 0.0080, 0.0121], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0105, 0.0155, 0.0151, 0.0121, 0.0166, 0.0141, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:42:43,271 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:42:59,936 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5002, 2.4981, 1.9260, 2.2339, 2.9100, 2.4291, 3.3259, 3.1951], device='cuda:3'), covar=tensor([0.0039, 0.0221, 0.0322, 0.0286, 0.0140, 0.0232, 0.0105, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0185, 0.0181, 0.0183, 0.0183, 0.0186, 0.0185, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:43:19,351 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 4200, loss[loss=0.2461, simple_loss=0.3204, pruned_loss=0.08588, over 11485.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2888, pruned_loss=0.066, over 3218054.03 frames. ], batch size: 248, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:43:23,697 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5518, 4.6434, 4.7688, 4.6747, 4.5943, 5.1951, 4.8093, 4.4865], device='cuda:3'), covar=tensor([0.0966, 0.1559, 0.1285, 0.1605, 0.2295, 0.0872, 0.1114, 0.2310], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0455, 0.0470, 0.0397, 0.0524, 0.0500, 0.0381, 0.0539], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 20:43:31,600 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 20:43:40,419 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:47,955 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:16,495 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:30,801 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 4250, loss[loss=0.1675, simple_loss=0.2605, pruned_loss=0.0372, over 15433.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2915, pruned_loss=0.0659, over 3185187.32 frames. ], batch size: 190, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:44:36,196 INFO [optim.py:368] (3/8) 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 20:45:01,425 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:45:18,386 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 20:45:48,478 INFO [train.py:904] (3/8) Epoch 8, batch 4300, loss[loss=0.2057, simple_loss=0.3006, pruned_loss=0.05539, over 16699.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2926, pruned_loss=0.06499, over 3194798.58 frames. ], batch size: 89, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:46:04,946 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 20:46:37,773 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 20:46:40,261 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 20:47:01,161 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9710, 3.1164, 2.4897, 5.0404, 3.9884, 4.3788, 2.0031, 3.2756], device='cuda:3'), covar=tensor([0.1185, 0.0651, 0.1230, 0.0096, 0.0392, 0.0339, 0.1260, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0152, 0.0172, 0.0119, 0.0203, 0.0205, 0.0172, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 20:47:02,655 INFO [train.py:904] (3/8) Epoch 8, batch 4350, loss[loss=0.2003, simple_loss=0.2847, pruned_loss=0.05799, over 16794.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.296, pruned_loss=0.06573, over 3207317.67 frames. ], batch size: 39, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:47:03,853 INFO [optim.py:368] (3/8) 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 20:48:01,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9336, 2.1504, 1.7344, 2.0200, 2.5656, 2.2616, 2.9644, 2.9356], device='cuda:3'), covar=tensor([0.0062, 0.0229, 0.0316, 0.0266, 0.0133, 0.0241, 0.0091, 0.0108], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0182, 0.0179, 0.0180, 0.0181, 0.0184, 0.0180, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:48:17,344 INFO [train.py:904] (3/8) Epoch 8, batch 4400, loss[loss=0.2358, simple_loss=0.3076, pruned_loss=0.082, over 11976.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2978, pruned_loss=0.06623, over 3219174.28 frames. ], batch size: 246, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:48:37,110 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-04-28 20:48:38,561 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 20:49:26,864 INFO [train.py:904] (3/8) Epoch 8, batch 4450, loss[loss=0.2185, simple_loss=0.3065, pruned_loss=0.06524, over 16866.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.301, pruned_loss=0.06731, over 3218529.62 frames. ], batch size: 102, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:28,913 INFO [optim.py:368] (3/8) 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 20:50:07,440 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7834, 4.4903, 4.7210, 4.9284, 5.0524, 4.5694, 5.0416, 5.0516], device='cuda:3'), covar=tensor([0.1031, 0.0931, 0.1161, 0.0425, 0.0352, 0.0665, 0.0354, 0.0351], device='cuda:3'), in_proj_covar=tensor([0.0462, 0.0573, 0.0708, 0.0579, 0.0436, 0.0440, 0.0446, 0.0491], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:50:38,184 INFO [train.py:904] (3/8) Epoch 8, batch 4500, loss[loss=0.2188, simple_loss=0.2997, pruned_loss=0.06889, over 15362.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.301, pruned_loss=0.06753, over 3213459.25 frames. ], batch size: 190, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:50:57,708 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:24,919 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:27,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4396, 5.3437, 5.0914, 4.5417, 5.3251, 1.7426, 4.9959, 5.0748], device='cuda:3'), covar=tensor([0.0031, 0.0030, 0.0075, 0.0240, 0.0031, 0.2060, 0.0061, 0.0089], device='cuda:3'), in_proj_covar=tensor([0.0113, 0.0100, 0.0148, 0.0145, 0.0116, 0.0160, 0.0135, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:51:41,026 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 20:51:51,099 INFO [train.py:904] (3/8) Epoch 8, batch 4550, loss[loss=0.2065, simple_loss=0.2914, pruned_loss=0.06076, over 16389.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3022, pruned_loss=0.06828, over 3218352.90 frames. ], batch size: 35, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:51:52,276 INFO [optim.py:368] (3/8) 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 20:52:06,456 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:17,119 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:24,999 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 20:52:52,912 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4402, 4.5538, 4.2185, 4.0772, 3.8203, 4.3699, 4.0667, 3.9278], device='cuda:3'), covar=tensor([0.0428, 0.0195, 0.0262, 0.0228, 0.0929, 0.0246, 0.0639, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0214, 0.0251, 0.0252, 0.0225, 0.0280, 0.0253, 0.0173, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 20:53:02,628 INFO [train.py:904] (3/8) Epoch 8, batch 4600, loss[loss=0.1803, simple_loss=0.2673, pruned_loss=0.04662, over 17045.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3028, pruned_loss=0.06842, over 3210766.00 frames. ], batch size: 50, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:53:25,888 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:53:31,531 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-28 20:54:09,534 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 20:54:12,063 INFO [train.py:904] (3/8) Epoch 8, batch 4650, loss[loss=0.2146, simple_loss=0.2961, pruned_loss=0.06652, over 16539.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3023, pruned_loss=0.06878, over 3193502.76 frames. ], batch size: 75, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:54:13,263 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.294e+02 2.810e+02 3.370e+02 9.861e+02, threshold=5.619e+02, percent-clipped=3.0 2023-04-28 20:55:23,492 INFO [train.py:904] (3/8) Epoch 8, batch 4700, loss[loss=0.2157, simple_loss=0.2929, pruned_loss=0.06926, over 16878.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2988, pruned_loss=0.06686, over 3205131.24 frames. ], batch size: 116, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:55:33,634 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3700, 4.1812, 4.3265, 2.7634, 3.7423, 4.1576, 3.8230, 2.3010], device='cuda:3'), covar=tensor([0.0355, 0.0019, 0.0014, 0.0245, 0.0041, 0.0047, 0.0036, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0063, 0.0062, 0.0119, 0.0068, 0.0078, 0.0070, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 20:56:31,926 INFO [train.py:904] (3/8) Epoch 8, batch 4750, loss[loss=0.1927, simple_loss=0.272, pruned_loss=0.05676, over 16666.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2947, pruned_loss=0.06483, over 3204762.91 frames. ], batch size: 62, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:56:33,066 INFO [optim.py:368] (3/8) 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 20:56:35,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-28 20:57:44,145 INFO [train.py:904] (3/8) Epoch 8, batch 4800, loss[loss=0.191, simple_loss=0.2831, pruned_loss=0.04942, over 16778.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2914, pruned_loss=0.0629, over 3215847.19 frames. ], batch size: 116, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:58:31,983 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:58:58,201 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 20:58:58,545 INFO [train.py:904] (3/8) Epoch 8, batch 4850, loss[loss=0.234, simple_loss=0.3142, pruned_loss=0.07684, over 11856.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2922, pruned_loss=0.06232, over 3190896.17 frames. ], batch size: 248, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 20:59:01,503 INFO [optim.py:368] (3/8) 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 20:59:36,898 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 20:59:46,581 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:00:17,705 INFO [train.py:904] (3/8) Epoch 8, batch 4900, loss[loss=0.1808, simple_loss=0.2711, pruned_loss=0.04526, over 16502.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2922, pruned_loss=0.06173, over 3172434.55 frames. ], batch size: 68, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:00:49,945 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:01:23,235 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0207, 2.6690, 2.6843, 1.8393, 2.8587, 2.8743, 2.4341, 2.3310], device='cuda:3'), covar=tensor([0.0660, 0.0162, 0.0135, 0.0851, 0.0067, 0.0106, 0.0309, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0096, 0.0082, 0.0137, 0.0067, 0.0089, 0.0116, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 21:01:34,529 INFO [train.py:904] (3/8) Epoch 8, batch 4950, loss[loss=0.2222, simple_loss=0.314, pruned_loss=0.06521, over 15488.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2924, pruned_loss=0.06165, over 3171932.34 frames. ], batch size: 190, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:01:36,816 INFO [optim.py:368] (3/8) 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 21:02:45,331 INFO [train.py:904] (3/8) Epoch 8, batch 5000, loss[loss=0.1974, simple_loss=0.2818, pruned_loss=0.05648, over 16593.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2938, pruned_loss=0.06202, over 3176693.46 frames. ], batch size: 57, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:55,755 INFO [train.py:904] (3/8) Epoch 8, batch 5050, loss[loss=0.2664, simple_loss=0.34, pruned_loss=0.09646, over 12209.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2946, pruned_loss=0.06183, over 3189806.46 frames. ], batch size: 247, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:57,925 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.576e+02 2.969e+02 3.606e+02 8.836e+02, threshold=5.938e+02, percent-clipped=5.0 2023-04-28 21:04:04,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3195, 4.1116, 4.3299, 4.4869, 4.6513, 4.2472, 4.6081, 4.6036], device='cuda:3'), covar=tensor([0.1173, 0.0909, 0.1245, 0.0526, 0.0401, 0.0775, 0.0409, 0.0422], device='cuda:3'), in_proj_covar=tensor([0.0468, 0.0571, 0.0713, 0.0583, 0.0439, 0.0443, 0.0447, 0.0497], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:05:07,316 INFO [train.py:904] (3/8) Epoch 8, batch 5100, loss[loss=0.1771, simple_loss=0.2663, pruned_loss=0.04392, over 16869.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.292, pruned_loss=0.06054, over 3197114.66 frames. ], batch size: 96, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:20,870 INFO [train.py:904] (3/8) Epoch 8, batch 5150, loss[loss=0.2339, simple_loss=0.3179, pruned_loss=0.07501, over 11849.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2921, pruned_loss=0.05967, over 3195908.76 frames. ], batch size: 246, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:24,096 INFO [optim.py:368] (3/8) 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 21:07:16,109 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6408, 3.1980, 3.0003, 1.7001, 2.6121, 2.1380, 3.1357, 3.2148], device='cuda:3'), covar=tensor([0.0260, 0.0511, 0.0571, 0.1683, 0.0751, 0.0873, 0.0633, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0133, 0.0153, 0.0140, 0.0132, 0.0122, 0.0134, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 21:07:33,713 INFO [train.py:904] (3/8) Epoch 8, batch 5200, loss[loss=0.2032, simple_loss=0.286, pruned_loss=0.06015, over 16840.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2909, pruned_loss=0.05935, over 3181028.05 frames. ], batch size: 116, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:07:59,732 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:08:03,267 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-04-28 21:08:36,425 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:08:48,737 INFO [train.py:904] (3/8) Epoch 8, batch 5250, loss[loss=0.2028, simple_loss=0.2949, pruned_loss=0.05533, over 16437.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2886, pruned_loss=0.05921, over 3183282.28 frames. ], batch size: 146, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:08:51,152 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.354e+02 2.757e+02 3.436e+02 6.643e+02, threshold=5.515e+02, percent-clipped=1.0 2023-04-28 21:09:16,869 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3843, 4.3442, 4.8018, 4.7960, 4.7603, 4.4627, 4.4225, 4.2632], device='cuda:3'), covar=tensor([0.0250, 0.0459, 0.0272, 0.0311, 0.0344, 0.0262, 0.0745, 0.0375], device='cuda:3'), in_proj_covar=tensor([0.0289, 0.0291, 0.0291, 0.0282, 0.0333, 0.0309, 0.0414, 0.0255], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 21:09:19,382 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:09:30,932 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:09:35,366 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 21:09:41,172 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8253, 2.7777, 2.6684, 1.9083, 2.4744, 2.6544, 2.6580, 1.7849], device='cuda:3'), covar=tensor([0.0341, 0.0041, 0.0039, 0.0271, 0.0072, 0.0069, 0.0055, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0064, 0.0063, 0.0121, 0.0069, 0.0079, 0.0070, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 21:10:01,828 INFO [train.py:904] (3/8) Epoch 8, batch 5300, loss[loss=0.1783, simple_loss=0.2611, pruned_loss=0.04774, over 16634.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2842, pruned_loss=0.05745, over 3193072.60 frames. ], batch size: 134, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:10:06,433 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:10:48,018 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:11:13,477 INFO [train.py:904] (3/8) Epoch 8, batch 5350, loss[loss=0.2091, simple_loss=0.299, pruned_loss=0.05962, over 16693.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2826, pruned_loss=0.05676, over 3200327.78 frames. ], batch size: 134, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:11:15,924 INFO [optim.py:368] (3/8) 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 21:11:56,087 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3968, 3.4611, 2.7279, 2.0355, 2.5241, 2.1767, 3.5946, 3.3660], device='cuda:3'), covar=tensor([0.2386, 0.0679, 0.1355, 0.2018, 0.1870, 0.1578, 0.0493, 0.0846], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0249, 0.0276, 0.0263, 0.0278, 0.0210, 0.0259, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:12:26,519 INFO [train.py:904] (3/8) Epoch 8, batch 5400, loss[loss=0.198, simple_loss=0.2887, pruned_loss=0.05365, over 16883.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2856, pruned_loss=0.05762, over 3208301.87 frames. ], batch size: 96, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:33,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5703, 3.9803, 3.9170, 1.8891, 3.1373, 2.7250, 3.7736, 3.8336], device='cuda:3'), covar=tensor([0.0197, 0.0457, 0.0416, 0.1611, 0.0651, 0.0733, 0.0537, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0132, 0.0154, 0.0139, 0.0132, 0.0121, 0.0134, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 21:13:43,648 INFO [train.py:904] (3/8) Epoch 8, batch 5450, loss[loss=0.2473, simple_loss=0.3124, pruned_loss=0.0911, over 11837.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2888, pruned_loss=0.05963, over 3213401.27 frames. ], batch size: 247, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:46,706 INFO [optim.py:368] (3/8) 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 21:13:49,180 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3269, 3.8599, 3.2201, 3.7093, 3.3313, 3.5276, 3.5342, 3.7943], device='cuda:3'), covar=tensor([0.4186, 0.2061, 0.3802, 0.1725, 0.2256, 0.2703, 0.2143, 0.2257], device='cuda:3'), in_proj_covar=tensor([0.0463, 0.0584, 0.0493, 0.0402, 0.0373, 0.0382, 0.0488, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:13:53,338 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4767, 5.8411, 5.5031, 5.5897, 5.1458, 5.0331, 5.2900, 5.9440], device='cuda:3'), covar=tensor([0.0834, 0.0647, 0.0881, 0.0559, 0.0670, 0.0531, 0.0726, 0.0677], device='cuda:3'), in_proj_covar=tensor([0.0463, 0.0584, 0.0492, 0.0402, 0.0373, 0.0382, 0.0488, 0.0428], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:14:08,665 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5931, 2.6546, 2.3050, 3.8553, 2.9370, 3.9947, 1.3017, 2.8678], device='cuda:3'), covar=tensor([0.1358, 0.0703, 0.1209, 0.0127, 0.0265, 0.0319, 0.1601, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0153, 0.0175, 0.0118, 0.0201, 0.0204, 0.0172, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 21:14:50,951 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 21:15:01,740 INFO [train.py:904] (3/8) Epoch 8, batch 5500, loss[loss=0.238, simple_loss=0.3157, pruned_loss=0.08015, over 16359.00 frames. ], tot_loss[loss=0.215, simple_loss=0.298, pruned_loss=0.066, over 3176960.52 frames. ], batch size: 146, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:16:16,796 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4742, 3.5149, 3.2386, 3.0244, 3.0688, 3.3456, 3.2514, 3.1428], device='cuda:3'), covar=tensor([0.0541, 0.0439, 0.0233, 0.0203, 0.0524, 0.0345, 0.0945, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0220, 0.0259, 0.0255, 0.0226, 0.0285, 0.0263, 0.0173, 0.0295], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:16:22,240 INFO [train.py:904] (3/8) Epoch 8, batch 5550, loss[loss=0.2285, simple_loss=0.3047, pruned_loss=0.0762, over 16548.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3059, pruned_loss=0.07251, over 3128381.31 frames. ], batch size: 62, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:16:26,052 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:17:00,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1009, 3.8445, 3.7324, 4.2501, 4.3065, 3.9784, 4.2427, 4.3490], device='cuda:3'), covar=tensor([0.1125, 0.0994, 0.2175, 0.0807, 0.0820, 0.1397, 0.1062, 0.0835], device='cuda:3'), in_proj_covar=tensor([0.0468, 0.0569, 0.0707, 0.0579, 0.0437, 0.0438, 0.0451, 0.0501], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:17:40,833 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 5600, loss[loss=0.2329, simple_loss=0.3216, pruned_loss=0.07208, over 16715.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3121, pruned_loss=0.0783, over 3067767.56 frames. ], batch size: 89, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:18:12,571 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 21:18:28,907 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:19:06,822 INFO [train.py:904] (3/8) Epoch 8, batch 5650, loss[loss=0.3016, simple_loss=0.3586, pruned_loss=0.1223, over 11338.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3179, pruned_loss=0.08322, over 3038622.87 frames. ], batch size: 248, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:19:10,222 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 5700, loss[loss=0.3008, simple_loss=0.3446, pruned_loss=0.1285, over 11503.00 frames. ], tot_loss[loss=0.245, simple_loss=0.32, pruned_loss=0.08495, over 3050908.19 frames. ], batch size: 248, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:21:12,206 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3258, 4.6208, 4.3577, 4.3764, 4.0801, 4.1003, 4.2051, 4.6620], device='cuda:3'), covar=tensor([0.0892, 0.0762, 0.1090, 0.0685, 0.0736, 0.1274, 0.0824, 0.0773], device='cuda:3'), in_proj_covar=tensor([0.0467, 0.0586, 0.0499, 0.0401, 0.0374, 0.0385, 0.0489, 0.0432], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:21:49,368 INFO [train.py:904] (3/8) Epoch 8, batch 5750, loss[loss=0.2563, simple_loss=0.3366, pruned_loss=0.08805, over 16718.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.323, pruned_loss=0.08659, over 3045557.54 frames. ], batch size: 39, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:21:54,072 INFO [optim.py:368] (3/8) 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 21:22:40,015 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 21:22:52,050 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:23:12,544 INFO [train.py:904] (3/8) Epoch 8, batch 5800, loss[loss=0.2462, simple_loss=0.3258, pruned_loss=0.08329, over 15287.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3225, pruned_loss=0.08502, over 3042336.67 frames. ], batch size: 190, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:23:40,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5326, 5.4950, 5.2542, 4.7792, 5.4721, 2.2323, 5.1802, 5.3233], device='cuda:3'), covar=tensor([0.0051, 0.0051, 0.0119, 0.0270, 0.0051, 0.1816, 0.0084, 0.0100], device='cuda:3'), in_proj_covar=tensor([0.0111, 0.0098, 0.0146, 0.0142, 0.0114, 0.0161, 0.0130, 0.0136], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:24:13,457 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4773, 1.4821, 2.1273, 2.3702, 2.3210, 2.5326, 1.4943, 2.5970], device='cuda:3'), covar=tensor([0.0100, 0.0311, 0.0176, 0.0166, 0.0170, 0.0115, 0.0339, 0.0072], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0157, 0.0143, 0.0142, 0.0150, 0.0106, 0.0159, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 21:24:21,619 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2882, 1.8746, 1.3623, 1.6299, 2.1675, 1.9830, 2.3460, 2.4466], device='cuda:3'), covar=tensor([0.0094, 0.0287, 0.0415, 0.0346, 0.0148, 0.0276, 0.0126, 0.0167], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0180, 0.0179, 0.0179, 0.0177, 0.0181, 0.0178, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:24:30,519 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 5850, loss[loss=0.2325, simple_loss=0.315, pruned_loss=0.07499, over 16236.00 frames. ], tot_loss[loss=0.2428, simple_loss=0.3199, pruned_loss=0.08283, over 3042538.85 frames. ], batch size: 165, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:24:37,987 INFO [optim.py:368] (3/8) 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 21:25:02,283 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6347, 2.6644, 1.7643, 2.7800, 2.1564, 2.7722, 2.0742, 2.4039], device='cuda:3'), covar=tensor([0.0220, 0.0322, 0.1246, 0.0148, 0.0595, 0.0466, 0.1074, 0.0534], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0160, 0.0183, 0.0102, 0.0165, 0.0197, 0.0190, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 21:25:09,372 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:25:52,751 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:25:56,471 INFO [train.py:904] (3/8) Epoch 8, batch 5900, loss[loss=0.2029, simple_loss=0.2918, pruned_loss=0.05704, over 16618.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3187, pruned_loss=0.08216, over 3043899.60 frames. ], batch size: 57, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:26:33,733 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:26:42,726 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:27:11,066 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:27:17,804 INFO [train.py:904] (3/8) Epoch 8, batch 5950, loss[loss=0.2243, simple_loss=0.3074, pruned_loss=0.07063, over 16891.00 frames. ], tot_loss[loss=0.2402, simple_loss=0.3193, pruned_loss=0.08056, over 3041890.70 frames. ], batch size: 109, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:27:21,560 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:28:33,565 INFO [train.py:904] (3/8) Epoch 8, batch 6000, loss[loss=0.2586, simple_loss=0.3239, pruned_loss=0.09661, over 11826.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3187, pruned_loss=0.0807, over 3032336.09 frames. ], batch size: 248, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:28:33,565 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 21:28:44,114 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 21:29:48,569 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:30:00,317 INFO [train.py:904] (3/8) Epoch 8, batch 6050, loss[loss=0.2192, simple_loss=0.3066, pruned_loss=0.06588, over 15424.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3166, pruned_loss=0.0789, over 3067764.36 frames. ], batch size: 190, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:30:04,223 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.264e+02 3.621e+02 4.460e+02 5.601e+02 1.783e+03, threshold=8.919e+02, percent-clipped=7.0 2023-04-28 21:30:48,177 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4908, 3.6237, 3.8513, 1.7046, 4.0854, 4.0572, 2.9311, 2.9171], device='cuda:3'), covar=tensor([0.0764, 0.0152, 0.0141, 0.1241, 0.0049, 0.0076, 0.0366, 0.0427], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0098, 0.0083, 0.0142, 0.0070, 0.0092, 0.0119, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 21:31:11,090 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-28 21:31:19,254 INFO [train.py:904] (3/8) Epoch 8, batch 6100, loss[loss=0.2075, simple_loss=0.2985, pruned_loss=0.05822, over 16725.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3153, pruned_loss=0.07688, over 3096477.00 frames. ], batch size: 62, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:31:26,157 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:31:38,639 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2515, 1.4001, 1.8790, 2.1775, 2.2309, 2.4373, 1.4482, 2.3130], device='cuda:3'), covar=tensor([0.0132, 0.0326, 0.0191, 0.0208, 0.0187, 0.0100, 0.0332, 0.0073], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0159, 0.0144, 0.0143, 0.0152, 0.0107, 0.0161, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 21:31:48,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0198, 4.0202, 3.9170, 3.3145, 3.9600, 1.7560, 3.7332, 3.6259], device='cuda:3'), covar=tensor([0.0078, 0.0061, 0.0115, 0.0258, 0.0065, 0.1994, 0.0090, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0110, 0.0097, 0.0145, 0.0142, 0.0114, 0.0160, 0.0130, 0.0136], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:32:26,659 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:37,673 INFO [train.py:904] (3/8) Epoch 8, batch 6150, loss[loss=0.213, simple_loss=0.2926, pruned_loss=0.06666, over 16695.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3132, pruned_loss=0.07577, over 3117330.15 frames. ], batch size: 62, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:32:42,687 INFO [optim.py:368] (3/8) 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 21:32:52,974 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:58,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 21:33:51,732 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 21:33:59,282 INFO [train.py:904] (3/8) Epoch 8, batch 6200, loss[loss=0.2236, simple_loss=0.3011, pruned_loss=0.073, over 16277.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3108, pruned_loss=0.07526, over 3122447.99 frames. ], batch size: 165, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:34:31,309 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:31,961 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 21:34:33,174 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:39,075 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 6250, loss[loss=0.2388, simple_loss=0.3169, pruned_loss=0.08039, over 16467.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3103, pruned_loss=0.07504, over 3127138.71 frames. ], batch size: 146, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:35:22,795 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 3.064e+02 3.768e+02 4.806e+02 9.942e+02, threshold=7.536e+02, percent-clipped=2.0 2023-04-28 21:35:51,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4760, 2.0694, 1.6792, 1.9024, 2.4031, 2.1737, 2.5605, 2.6451], device='cuda:3'), covar=tensor([0.0078, 0.0273, 0.0335, 0.0318, 0.0131, 0.0238, 0.0136, 0.0164], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0179, 0.0177, 0.0178, 0.0175, 0.0179, 0.0176, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:36:07,283 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:12,931 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 6300, loss[loss=0.2259, simple_loss=0.3117, pruned_loss=0.07006, over 16761.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3105, pruned_loss=0.07488, over 3122363.05 frames. ], batch size: 124, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:37:04,609 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6782, 4.5192, 4.5705, 4.9039, 4.9711, 4.5499, 4.9688, 4.9588], device='cuda:3'), covar=tensor([0.1264, 0.1060, 0.1842, 0.0646, 0.0617, 0.0832, 0.0640, 0.0641], device='cuda:3'), in_proj_covar=tensor([0.0466, 0.0572, 0.0706, 0.0584, 0.0441, 0.0435, 0.0455, 0.0503], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:37:54,082 INFO [train.py:904] (3/8) Epoch 8, batch 6350, loss[loss=0.2972, simple_loss=0.3397, pruned_loss=0.1273, over 11214.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3117, pruned_loss=0.07665, over 3099031.27 frames. ], batch size: 248, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:38:00,461 INFO [optim.py:368] (3/8) 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 21:39:10,451 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:39:11,215 INFO [train.py:904] (3/8) Epoch 8, batch 6400, loss[loss=0.2494, simple_loss=0.3239, pruned_loss=0.08743, over 15457.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3114, pruned_loss=0.07736, over 3096926.65 frames. ], batch size: 191, lr: 8.49e-03, grad_scale: 8.0 2023-04-28 21:39:14,522 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6299, 3.7610, 3.0067, 2.2867, 2.7454, 2.2965, 4.1371, 3.6845], device='cuda:3'), covar=tensor([0.2362, 0.0685, 0.1338, 0.1809, 0.1874, 0.1578, 0.0334, 0.0734], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0250, 0.0273, 0.0262, 0.0278, 0.0210, 0.0256, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:39:33,294 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-28 21:39:52,905 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0778, 3.4369, 3.5173, 1.7978, 2.7825, 2.3176, 3.5259, 3.6190], device='cuda:3'), covar=tensor([0.0244, 0.0635, 0.0518, 0.1770, 0.0794, 0.0972, 0.0587, 0.0736], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0134, 0.0156, 0.0140, 0.0135, 0.0125, 0.0136, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 21:40:17,148 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:40:26,236 INFO [train.py:904] (3/8) Epoch 8, batch 6450, loss[loss=0.2206, simple_loss=0.3082, pruned_loss=0.06649, over 16498.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3105, pruned_loss=0.07568, over 3125911.98 frames. ], batch size: 75, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:40:33,077 INFO [optim.py:368] (3/8) 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 21:41:06,592 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1130, 2.6355, 2.6392, 1.8629, 2.7828, 2.8378, 2.5077, 2.4246], device='cuda:3'), covar=tensor([0.0716, 0.0208, 0.0207, 0.1016, 0.0096, 0.0180, 0.0381, 0.0423], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0098, 0.0084, 0.0140, 0.0069, 0.0091, 0.0119, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 21:41:31,679 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:41:45,631 INFO [train.py:904] (3/8) Epoch 8, batch 6500, loss[loss=0.2234, simple_loss=0.3036, pruned_loss=0.07163, over 17046.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3089, pruned_loss=0.0749, over 3122631.90 frames. ], batch size: 50, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:42:09,110 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 6550, loss[loss=0.2332, simple_loss=0.3248, pruned_loss=0.07087, over 17101.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3113, pruned_loss=0.07519, over 3131295.40 frames. ], batch size: 49, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:43:11,237 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.133e+02 3.759e+02 4.463e+02 1.371e+03, threshold=7.519e+02, percent-clipped=1.0 2023-04-28 21:43:28,276 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:47,922 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:53,323 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:44:22,370 INFO [train.py:904] (3/8) Epoch 8, batch 6600, loss[loss=0.2257, simple_loss=0.3172, pruned_loss=0.06704, over 16872.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.314, pruned_loss=0.07594, over 3126117.12 frames. ], batch size: 96, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:45:00,782 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:38,844 INFO [train.py:904] (3/8) Epoch 8, batch 6650, loss[loss=0.214, simple_loss=0.2914, pruned_loss=0.06835, over 17074.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3136, pruned_loss=0.07614, over 3138164.45 frames. ], batch size: 53, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:45:45,543 INFO [optim.py:368] (3/8) 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 21:46:18,965 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5941, 4.6309, 5.1158, 5.0402, 5.0540, 4.6671, 4.7176, 4.4700], device='cuda:3'), covar=tensor([0.0245, 0.0397, 0.0271, 0.0366, 0.0367, 0.0264, 0.0749, 0.0390], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0297, 0.0297, 0.0287, 0.0334, 0.0316, 0.0419, 0.0256], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 21:46:35,423 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:46:40,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7715, 2.6390, 2.6474, 1.8673, 2.4298, 2.6075, 2.5425, 1.8506], device='cuda:3'), covar=tensor([0.0339, 0.0043, 0.0053, 0.0289, 0.0077, 0.0071, 0.0058, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0061, 0.0064, 0.0120, 0.0068, 0.0078, 0.0069, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 21:46:53,704 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:46:54,546 INFO [train.py:904] (3/8) Epoch 8, batch 6700, loss[loss=0.2361, simple_loss=0.3132, pruned_loss=0.07946, over 16339.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3118, pruned_loss=0.07604, over 3136100.51 frames. ], batch size: 35, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:06,139 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:48:10,947 INFO [train.py:904] (3/8) Epoch 8, batch 6750, loss[loss=0.2149, simple_loss=0.2982, pruned_loss=0.06584, over 16356.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3122, pruned_loss=0.0777, over 3094617.69 frames. ], batch size: 146, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:18,410 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.344e+02 4.001e+02 5.088e+02 1.053e+03, threshold=8.003e+02, percent-clipped=2.0 2023-04-28 21:48:32,830 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 21:49:25,314 INFO [train.py:904] (3/8) Epoch 8, batch 6800, loss[loss=0.245, simple_loss=0.3204, pruned_loss=0.08483, over 16465.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3129, pruned_loss=0.0781, over 3098535.61 frames. ], batch size: 62, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:49:49,023 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:50:44,166 INFO [train.py:904] (3/8) Epoch 8, batch 6850, loss[loss=0.2289, simple_loss=0.3223, pruned_loss=0.06773, over 16239.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3149, pruned_loss=0.07943, over 3088903.81 frames. ], batch size: 165, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:50:53,217 INFO [optim.py:368] (3/8) 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 21:51:03,699 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:24,264 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9729, 3.9970, 3.8568, 3.7099, 3.5302, 3.9147, 3.6479, 3.6894], device='cuda:3'), covar=tensor([0.0558, 0.0565, 0.0247, 0.0190, 0.0765, 0.0418, 0.0767, 0.0555], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0255, 0.0250, 0.0220, 0.0277, 0.0256, 0.0173, 0.0290], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:51:24,275 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:30,563 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:59,518 INFO [train.py:904] (3/8) Epoch 8, batch 6900, loss[loss=0.2817, simple_loss=0.347, pruned_loss=0.1082, over 15246.00 frames. ], tot_loss[loss=0.2371, simple_loss=0.3169, pruned_loss=0.07859, over 3094904.09 frames. ], batch size: 190, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:52:31,695 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:45,249 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 6950, loss[loss=0.2962, simple_loss=0.3444, pruned_loss=0.124, over 10922.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3199, pruned_loss=0.08193, over 3055216.96 frames. ], batch size: 247, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:53:29,769 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 3.295e+02 4.352e+02 5.795e+02 9.816e+02, threshold=8.703e+02, percent-clipped=9.0 2023-04-28 21:54:10,657 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 7000, loss[loss=0.2263, simple_loss=0.3134, pruned_loss=0.06955, over 16705.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3194, pruned_loss=0.08045, over 3068914.27 frames. ], batch size: 124, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:54:56,014 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4718, 4.4992, 4.3776, 3.6768, 4.4021, 1.5053, 4.1344, 4.1263], device='cuda:3'), covar=tensor([0.0086, 0.0068, 0.0121, 0.0346, 0.0069, 0.2158, 0.0125, 0.0175], device='cuda:3'), in_proj_covar=tensor([0.0108, 0.0096, 0.0143, 0.0140, 0.0114, 0.0160, 0.0128, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 21:55:53,774 INFO [train.py:904] (3/8) Epoch 8, batch 7050, loss[loss=0.2521, simple_loss=0.3272, pruned_loss=0.0885, over 15299.00 frames. ], tot_loss[loss=0.2401, simple_loss=0.3197, pruned_loss=0.08024, over 3062988.86 frames. ], batch size: 190, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:56:03,766 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 21:56:03,902 INFO [optim.py:368] (3/8) 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 21:56:32,460 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6130, 2.6125, 1.7683, 2.7802, 2.1289, 2.7731, 1.9807, 2.3536], device='cuda:3'), covar=tensor([0.0211, 0.0312, 0.1131, 0.0117, 0.0582, 0.0437, 0.1093, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0158, 0.0181, 0.0101, 0.0165, 0.0197, 0.0193, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 21:56:44,100 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6622, 2.6507, 2.0956, 4.1531, 2.9824, 3.9753, 1.4044, 2.7646], device='cuda:3'), covar=tensor([0.1356, 0.0766, 0.1497, 0.0140, 0.0411, 0.0396, 0.1622, 0.0968], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0151, 0.0172, 0.0118, 0.0200, 0.0203, 0.0173, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 21:57:11,208 INFO [train.py:904] (3/8) Epoch 8, batch 7100, loss[loss=0.2218, simple_loss=0.3089, pruned_loss=0.06736, over 16776.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3179, pruned_loss=0.07937, over 3075431.71 frames. ], batch size: 124, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:26,614 INFO [train.py:904] (3/8) Epoch 8, batch 7150, loss[loss=0.2191, simple_loss=0.2966, pruned_loss=0.07081, over 16619.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3154, pruned_loss=0.07923, over 3065423.09 frames. ], batch size: 57, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:36,169 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.673e+02 3.732e+02 4.325e+02 5.377e+02 1.184e+03, threshold=8.651e+02, percent-clipped=7.0 2023-04-28 21:58:38,124 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 21:58:54,029 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 21:59:09,612 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:59:39,845 INFO [train.py:904] (3/8) Epoch 8, batch 7200, loss[loss=0.2087, simple_loss=0.2902, pruned_loss=0.06361, over 11407.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3129, pruned_loss=0.07748, over 3044320.90 frames. ], batch size: 246, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:00:09,965 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:00:23,325 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-28 22:00:44,964 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:00:47,943 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8576, 5.3516, 5.4819, 5.3969, 5.3439, 5.9166, 5.4501, 5.2137], device='cuda:3'), covar=tensor([0.0863, 0.1584, 0.1558, 0.1462, 0.2125, 0.0876, 0.1283, 0.2238], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0435, 0.0461, 0.0388, 0.0513, 0.0492, 0.0374, 0.0525], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 22:00:59,551 INFO [train.py:904] (3/8) Epoch 8, batch 7250, loss[loss=0.1902, simple_loss=0.2849, pruned_loss=0.04776, over 16692.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3108, pruned_loss=0.07612, over 3036767.53 frames. ], batch size: 76, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:01:10,047 INFO [optim.py:368] (3/8) 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 22:01:14,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1981, 2.0274, 1.6344, 1.7449, 2.1903, 1.9771, 2.1741, 2.3293], device='cuda:3'), covar=tensor([0.0073, 0.0201, 0.0286, 0.0250, 0.0120, 0.0193, 0.0103, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0179, 0.0177, 0.0180, 0.0177, 0.0179, 0.0174, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:01:26,986 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:35,464 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6181, 3.2150, 3.1222, 1.7448, 2.7475, 2.3352, 3.1280, 3.2782], device='cuda:3'), covar=tensor([0.0326, 0.0584, 0.0551, 0.1870, 0.0795, 0.0843, 0.0689, 0.0801], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0132, 0.0155, 0.0141, 0.0133, 0.0124, 0.0136, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 22:01:47,633 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:02:16,550 INFO [train.py:904] (3/8) Epoch 8, batch 7300, loss[loss=0.2405, simple_loss=0.326, pruned_loss=0.07749, over 15210.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3101, pruned_loss=0.0757, over 3047919.37 frames. ], batch size: 190, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:02,792 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:03:34,471 INFO [train.py:904] (3/8) Epoch 8, batch 7350, loss[loss=0.2158, simple_loss=0.2964, pruned_loss=0.06759, over 17065.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3101, pruned_loss=0.0758, over 3035344.03 frames. ], batch size: 53, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:45,277 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 8, batch 7400, loss[loss=0.2344, simple_loss=0.3167, pruned_loss=0.07607, over 16737.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3122, pruned_loss=0.07654, over 3052347.97 frames. ], batch size: 83, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:05:08,409 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:05:58,653 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 22:06:13,356 INFO [train.py:904] (3/8) Epoch 8, batch 7450, loss[loss=0.2395, simple_loss=0.319, pruned_loss=0.07994, over 16969.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3133, pruned_loss=0.07743, over 3076841.22 frames. ], batch size: 109, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:06:26,555 INFO [optim.py:368] (3/8) 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 22:06:47,460 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:07:34,683 INFO [train.py:904] (3/8) Epoch 8, batch 7500, loss[loss=0.2074, simple_loss=0.2949, pruned_loss=0.05992, over 16743.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3134, pruned_loss=0.07642, over 3100866.40 frames. ], batch size: 89, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:07:42,043 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:08:33,285 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:08:55,093 INFO [train.py:904] (3/8) Epoch 8, batch 7550, loss[loss=0.2328, simple_loss=0.3145, pruned_loss=0.07558, over 16763.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3138, pruned_loss=0.07782, over 3062217.67 frames. ], batch size: 124, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:09:05,666 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.406e+02 3.301e+02 4.162e+02 5.431e+02 1.258e+03, threshold=8.325e+02, percent-clipped=7.0 2023-04-28 22:09:20,019 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:09:21,630 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 22:10:11,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0385, 3.0382, 3.1676, 1.6219, 3.3309, 3.3747, 2.5957, 2.4712], device='cuda:3'), covar=tensor([0.0785, 0.0181, 0.0144, 0.1174, 0.0059, 0.0108, 0.0421, 0.0478], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0094, 0.0080, 0.0134, 0.0065, 0.0086, 0.0115, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 22:10:12,495 INFO [train.py:904] (3/8) Epoch 8, batch 7600, loss[loss=0.24, simple_loss=0.3125, pruned_loss=0.08378, over 15371.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.312, pruned_loss=0.07708, over 3084863.73 frames. ], batch size: 190, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:10:26,607 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-28 22:10:41,261 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-28 22:10:45,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1038, 3.3117, 3.5358, 3.5074, 3.4755, 3.2917, 3.3087, 3.3846], device='cuda:3'), covar=tensor([0.0357, 0.0575, 0.0392, 0.0423, 0.0485, 0.0477, 0.0852, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0296, 0.0299, 0.0289, 0.0341, 0.0317, 0.0418, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 22:11:30,196 INFO [train.py:904] (3/8) Epoch 8, batch 7650, loss[loss=0.26, simple_loss=0.3323, pruned_loss=0.0939, over 16473.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3125, pruned_loss=0.07732, over 3091891.78 frames. ], batch size: 75, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:11:40,445 INFO [optim.py:368] (3/8) 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 22:12:21,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3145, 2.5291, 1.9457, 2.1060, 2.9204, 2.5264, 3.2090, 3.2027], device='cuda:3'), covar=tensor([0.0047, 0.0236, 0.0355, 0.0338, 0.0145, 0.0224, 0.0142, 0.0135], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0180, 0.0179, 0.0181, 0.0178, 0.0180, 0.0175, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:12:45,890 INFO [train.py:904] (3/8) Epoch 8, batch 7700, loss[loss=0.222, simple_loss=0.3054, pruned_loss=0.06928, over 16749.00 frames. ], tot_loss[loss=0.235, simple_loss=0.3128, pruned_loss=0.07862, over 3063204.26 frames. ], batch size: 83, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:13:34,001 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4218, 5.3569, 5.0635, 4.0415, 5.1416, 1.8062, 4.9578, 4.8973], device='cuda:3'), covar=tensor([0.0076, 0.0051, 0.0128, 0.0459, 0.0077, 0.2418, 0.0106, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0096, 0.0142, 0.0139, 0.0113, 0.0161, 0.0127, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:14:03,977 INFO [train.py:904] (3/8) Epoch 8, batch 7750, loss[loss=0.2154, simple_loss=0.3069, pruned_loss=0.06197, over 16397.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3129, pruned_loss=0.07878, over 3054904.24 frames. ], batch size: 75, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:14:17,807 INFO [optim.py:368] (3/8) 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 22:14:26,085 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:15:04,466 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3977, 2.0688, 1.5726, 1.8808, 2.4254, 2.1169, 2.4774, 2.6397], device='cuda:3'), covar=tensor([0.0081, 0.0248, 0.0338, 0.0291, 0.0140, 0.0236, 0.0144, 0.0143], device='cuda:3'), in_proj_covar=tensor([0.0109, 0.0181, 0.0180, 0.0181, 0.0179, 0.0181, 0.0176, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:15:19,810 INFO [train.py:904] (3/8) Epoch 8, batch 7800, loss[loss=0.2288, simple_loss=0.3175, pruned_loss=0.07009, over 17188.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3136, pruned_loss=0.07927, over 3048065.01 frames. ], batch size: 44, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:15:58,406 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 22:16:02,799 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8308, 2.6197, 2.7006, 1.9119, 2.4242, 2.6771, 2.5335, 1.8258], device='cuda:3'), covar=tensor([0.0301, 0.0042, 0.0046, 0.0253, 0.0074, 0.0065, 0.0056, 0.0281], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0062, 0.0064, 0.0121, 0.0069, 0.0079, 0.0070, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 22:16:12,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7620, 4.7664, 5.3121, 5.2224, 5.2271, 4.8620, 4.8466, 4.5296], device='cuda:3'), covar=tensor([0.0283, 0.0451, 0.0320, 0.0405, 0.0378, 0.0304, 0.0964, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0304, 0.0306, 0.0295, 0.0345, 0.0322, 0.0426, 0.0265], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 22:16:16,580 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:16:38,225 INFO [train.py:904] (3/8) Epoch 8, batch 7850, loss[loss=0.2093, simple_loss=0.3007, pruned_loss=0.05897, over 16814.00 frames. ], tot_loss[loss=0.2364, simple_loss=0.3147, pruned_loss=0.07902, over 3063975.04 frames. ], batch size: 83, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:49,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4461, 3.2711, 2.6507, 2.1223, 2.2790, 2.0714, 3.3068, 3.1738], device='cuda:3'), covar=tensor([0.2282, 0.0658, 0.1377, 0.1921, 0.1968, 0.1731, 0.0545, 0.0911], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0252, 0.0273, 0.0262, 0.0278, 0.0210, 0.0258, 0.0275], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:16:50,959 INFO [optim.py:368] (3/8) 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 22:16:53,958 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:20,756 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8136, 5.3524, 5.5215, 5.3569, 5.3421, 5.9129, 5.4449, 5.2218], device='cuda:3'), covar=tensor([0.0799, 0.1419, 0.1321, 0.1600, 0.2251, 0.0741, 0.1220, 0.2287], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0437, 0.0469, 0.0395, 0.0517, 0.0489, 0.0378, 0.0531], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 22:17:29,112 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 7900, loss[loss=0.2737, simple_loss=0.3287, pruned_loss=0.1093, over 11246.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3135, pruned_loss=0.07798, over 3080596.11 frames. ], batch size: 250, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:18:15,516 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:19:13,442 INFO [train.py:904] (3/8) Epoch 8, batch 7950, loss[loss=0.205, simple_loss=0.2887, pruned_loss=0.06065, over 16770.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3132, pruned_loss=0.07806, over 3076092.83 frames. ], batch size: 76, lr: 8.40e-03, grad_scale: 2.0 2023-04-28 22:19:25,534 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2048, 3.8844, 3.8258, 1.6299, 4.1293, 4.2247, 3.0068, 2.9014], device='cuda:3'), covar=tensor([0.1119, 0.0120, 0.0211, 0.1396, 0.0057, 0.0068, 0.0413, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0097, 0.0083, 0.0139, 0.0067, 0.0090, 0.0118, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 22:19:27,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5403, 5.9085, 5.6286, 5.7278, 5.2097, 5.1061, 5.3037, 5.9826], device='cuda:3'), covar=tensor([0.0861, 0.0683, 0.0897, 0.0536, 0.0759, 0.0526, 0.0785, 0.0683], device='cuda:3'), in_proj_covar=tensor([0.0465, 0.0577, 0.0496, 0.0398, 0.0363, 0.0389, 0.0487, 0.0431], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:19:28,062 INFO [optim.py:368] (3/8) 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 22:19:49,253 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4452, 4.1720, 4.1413, 2.8067, 3.7367, 4.1562, 3.7878, 2.2611], device='cuda:3'), covar=tensor([0.0376, 0.0025, 0.0033, 0.0270, 0.0051, 0.0062, 0.0042, 0.0372], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0061, 0.0064, 0.0120, 0.0068, 0.0080, 0.0069, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 22:19:52,434 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:20:32,280 INFO [train.py:904] (3/8) Epoch 8, batch 8000, loss[loss=0.2105, simple_loss=0.3045, pruned_loss=0.05824, over 16743.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3136, pruned_loss=0.07849, over 3077144.58 frames. ], batch size: 89, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:21:40,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3107, 1.4320, 1.8174, 2.1862, 2.3067, 2.4910, 1.5301, 2.3842], device='cuda:3'), covar=tensor([0.0127, 0.0326, 0.0205, 0.0210, 0.0175, 0.0113, 0.0335, 0.0071], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0160, 0.0143, 0.0142, 0.0150, 0.0108, 0.0158, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 22:21:48,606 INFO [train.py:904] (3/8) Epoch 8, batch 8050, loss[loss=0.2786, simple_loss=0.3421, pruned_loss=0.1075, over 11879.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3136, pruned_loss=0.07786, over 3085268.04 frames. ], batch size: 248, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:22:02,034 INFO [optim.py:368] (3/8) 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 22:22:11,160 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:23:05,315 INFO [train.py:904] (3/8) Epoch 8, batch 8100, loss[loss=0.2074, simple_loss=0.2956, pruned_loss=0.0596, over 17002.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3128, pruned_loss=0.07677, over 3088930.63 frames. ], batch size: 41, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:23:23,753 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:23:46,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3605, 1.9972, 2.1501, 4.0377, 2.0066, 2.4930, 2.1203, 2.1467], device='cuda:3'), covar=tensor([0.0884, 0.3059, 0.1973, 0.0344, 0.3452, 0.2096, 0.2687, 0.3056], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0371, 0.0308, 0.0319, 0.0407, 0.0409, 0.0328, 0.0436], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:24:22,973 INFO [train.py:904] (3/8) Epoch 8, batch 8150, loss[loss=0.2147, simple_loss=0.294, pruned_loss=0.06774, over 16200.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3106, pruned_loss=0.07633, over 3090750.55 frames. ], batch size: 165, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:24:36,887 INFO [optim.py:368] (3/8) 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 22:24:38,562 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:25:01,488 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8319, 3.0493, 3.1145, 2.1365, 2.9122, 3.0866, 3.0442, 1.8344], device='cuda:3'), covar=tensor([0.0405, 0.0036, 0.0045, 0.0290, 0.0072, 0.0084, 0.0050, 0.0339], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0061, 0.0064, 0.0120, 0.0069, 0.0080, 0.0069, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 22:25:23,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0240, 4.2460, 2.5368, 4.8758, 3.0406, 4.7655, 2.5117, 3.1939], device='cuda:3'), covar=tensor([0.0172, 0.0286, 0.1448, 0.0069, 0.0700, 0.0438, 0.1397, 0.0619], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0157, 0.0182, 0.0104, 0.0166, 0.0199, 0.0192, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 22:25:42,515 INFO [train.py:904] (3/8) Epoch 8, batch 8200, loss[loss=0.2166, simple_loss=0.2934, pruned_loss=0.06988, over 16518.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3082, pruned_loss=0.07572, over 3080630.64 frames. ], batch size: 62, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:25:55,509 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:26:28,022 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8870, 2.2795, 2.3337, 2.9902, 1.9030, 3.2926, 1.6186, 2.6833], device='cuda:3'), covar=tensor([0.1240, 0.0599, 0.0942, 0.0155, 0.0106, 0.0414, 0.1405, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0151, 0.0170, 0.0118, 0.0199, 0.0199, 0.0171, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 22:26:36,811 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:27:04,307 INFO [train.py:904] (3/8) Epoch 8, batch 8250, loss[loss=0.2281, simple_loss=0.3128, pruned_loss=0.07169, over 15429.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3076, pruned_loss=0.07379, over 3059665.36 frames. ], batch size: 191, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:27:19,441 INFO [optim.py:368] (3/8) 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 22:27:37,064 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:27:56,959 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:17,617 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:28:26,463 INFO [train.py:904] (3/8) Epoch 8, batch 8300, loss[loss=0.1961, simple_loss=0.2726, pruned_loss=0.0598, over 11926.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3037, pruned_loss=0.06992, over 3062926.92 frames. ], batch size: 246, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:28:38,143 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0995, 2.8501, 2.7624, 2.0873, 2.5995, 2.1961, 2.7323, 2.9521], device='cuda:3'), covar=tensor([0.0309, 0.0686, 0.0431, 0.1511, 0.0698, 0.0994, 0.0583, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0129, 0.0151, 0.0137, 0.0129, 0.0122, 0.0132, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 22:28:46,892 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:57,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5736, 1.9847, 1.7169, 1.6888, 2.2804, 2.0264, 2.4042, 2.4814], device='cuda:3'), covar=tensor([0.0062, 0.0267, 0.0314, 0.0325, 0.0159, 0.0233, 0.0126, 0.0152], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0179, 0.0178, 0.0177, 0.0175, 0.0179, 0.0173, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:29:05,008 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:29:36,189 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 8350, loss[loss=0.2272, simple_loss=0.3198, pruned_loss=0.06733, over 16795.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3028, pruned_loss=0.06818, over 3042244.08 frames. ], batch size: 116, lr: 8.38e-03, grad_scale: 4.0 2023-04-28 22:30:02,871 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.645e+02 3.311e+02 4.098e+02 6.833e+02, threshold=6.622e+02, percent-clipped=0.0 2023-04-28 22:30:26,196 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:30:43,733 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:31:09,035 INFO [train.py:904] (3/8) Epoch 8, batch 8400, loss[loss=0.2069, simple_loss=0.2804, pruned_loss=0.06672, over 12167.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2997, pruned_loss=0.06596, over 3028559.93 frames. ], batch size: 247, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:27,061 INFO [train.py:904] (3/8) Epoch 8, batch 8450, loss[loss=0.1962, simple_loss=0.2828, pruned_loss=0.05481, over 15371.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2981, pruned_loss=0.06412, over 3052580.30 frames. ], batch size: 191, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:42,128 INFO [optim.py:368] (3/8) 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 22:33:02,035 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-28 22:33:47,271 INFO [train.py:904] (3/8) Epoch 8, batch 8500, loss[loss=0.1727, simple_loss=0.2665, pruned_loss=0.03942, over 16912.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2932, pruned_loss=0.06093, over 3043439.10 frames. ], batch size: 109, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:34:40,432 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:35:09,510 INFO [train.py:904] (3/8) Epoch 8, batch 8550, loss[loss=0.1922, simple_loss=0.2808, pruned_loss=0.05178, over 17209.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2902, pruned_loss=0.05934, over 3035009.40 frames. ], batch size: 44, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:35:26,481 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.652e+02 3.316e+02 4.199e+02 1.038e+03, threshold=6.632e+02, percent-clipped=3.0 2023-04-28 22:35:47,390 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:28,543 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:36:30,476 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3484, 4.0061, 3.8278, 2.1290, 3.1280, 2.6849, 3.6490, 4.1105], device='cuda:3'), covar=tensor([0.0241, 0.0491, 0.0421, 0.1524, 0.0662, 0.0801, 0.0570, 0.0653], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0127, 0.0148, 0.0136, 0.0128, 0.0121, 0.0131, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 22:36:37,260 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 8600, loss[loss=0.1963, simple_loss=0.2752, pruned_loss=0.05871, over 12585.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2904, pruned_loss=0.05822, over 3027209.17 frames. ], batch size: 250, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:36:59,398 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:37:24,926 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:02,629 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:11,299 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 22:38:26,202 INFO [train.py:904] (3/8) Epoch 8, batch 8650, loss[loss=0.1871, simple_loss=0.277, pruned_loss=0.0486, over 16742.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2883, pruned_loss=0.05639, over 3034941.28 frames. ], batch size: 134, lr: 8.37e-03, grad_scale: 4.0 2023-04-28 22:38:42,788 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2896, 4.0810, 4.0735, 4.4165, 4.6388, 4.1093, 4.5339, 4.5654], device='cuda:3'), covar=tensor([0.1329, 0.1082, 0.1959, 0.1034, 0.0693, 0.1223, 0.0901, 0.0891], device='cuda:3'), in_proj_covar=tensor([0.0451, 0.0551, 0.0673, 0.0565, 0.0428, 0.0421, 0.0449, 0.0492], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:38:50,263 INFO [optim.py:368] (3/8) 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 22:39:04,755 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:09,486 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:22,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0089, 3.1685, 1.7030, 3.3277, 2.2954, 3.2752, 1.9530, 2.6499], device='cuda:3'), covar=tensor([0.0202, 0.0310, 0.1489, 0.0126, 0.0793, 0.0558, 0.1512, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0151, 0.0176, 0.0099, 0.0160, 0.0188, 0.0188, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 22:39:32,715 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:39:48,757 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8113, 3.7758, 3.9316, 3.8467, 3.8889, 4.3010, 4.0080, 3.7334], device='cuda:3'), covar=tensor([0.1814, 0.1931, 0.1651, 0.2037, 0.2760, 0.1318, 0.1255, 0.2353], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0412, 0.0442, 0.0368, 0.0486, 0.0465, 0.0362, 0.0490], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:40:00,112 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3951, 3.2133, 2.6582, 2.1342, 2.1327, 2.1648, 3.4061, 3.0379], device='cuda:3'), covar=tensor([0.2430, 0.0722, 0.1485, 0.2216, 0.2339, 0.1782, 0.0460, 0.1010], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0244, 0.0268, 0.0255, 0.0259, 0.0206, 0.0247, 0.0263], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:40:12,066 INFO [train.py:904] (3/8) Epoch 8, batch 8700, loss[loss=0.1709, simple_loss=0.2689, pruned_loss=0.03648, over 16183.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2852, pruned_loss=0.05484, over 3033488.21 frames. ], batch size: 165, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:40:32,984 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:40:56,116 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4622, 1.9129, 1.5201, 1.5881, 2.1341, 1.8257, 2.2273, 2.3757], device='cuda:3'), covar=tensor([0.0085, 0.0286, 0.0344, 0.0313, 0.0161, 0.0253, 0.0119, 0.0144], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0179, 0.0175, 0.0176, 0.0174, 0.0177, 0.0167, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:41:44,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5934, 2.6645, 1.6408, 2.7842, 2.1104, 2.8010, 1.8153, 2.3600], device='cuda:3'), covar=tensor([0.0288, 0.0381, 0.1602, 0.0204, 0.0788, 0.0524, 0.1606, 0.0653], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0150, 0.0175, 0.0098, 0.0159, 0.0186, 0.0186, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 22:41:50,485 INFO [train.py:904] (3/8) Epoch 8, batch 8750, loss[loss=0.1962, simple_loss=0.2955, pruned_loss=0.04846, over 16777.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2854, pruned_loss=0.05447, over 3051590.28 frames. ], batch size: 124, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:42:15,364 INFO [optim.py:368] (3/8) 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 22:42:43,945 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 8800, loss[loss=0.1952, simple_loss=0.2774, pruned_loss=0.0565, over 12599.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2835, pruned_loss=0.05325, over 3041192.25 frames. ], batch size: 248, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:31,356 INFO [train.py:904] (3/8) Epoch 8, batch 8850, loss[loss=0.2113, simple_loss=0.3106, pruned_loss=0.05604, over 16265.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2863, pruned_loss=0.05258, over 3047103.65 frames. ], batch size: 165, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:51,484 INFO [optim.py:368] (3/8) 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 22:46:57,776 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:46:57,897 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:47:20,936 INFO [train.py:904] (3/8) Epoch 8, batch 8900, loss[loss=0.1727, simple_loss=0.275, pruned_loss=0.03521, over 16865.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2857, pruned_loss=0.05135, over 3048017.29 frames. ], batch size: 96, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:48:54,505 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:48:54,533 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:49:29,485 INFO [train.py:904] (3/8) Epoch 8, batch 8950, loss[loss=0.172, simple_loss=0.2633, pruned_loss=0.04028, over 16588.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2852, pruned_loss=0.05158, over 3071545.46 frames. ], batch size: 68, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:49:50,484 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:00,374 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4322, 3.0258, 3.0580, 1.8306, 2.5622, 2.1549, 2.8720, 3.1210], device='cuda:3'), covar=tensor([0.0276, 0.0619, 0.0482, 0.1674, 0.0779, 0.0991, 0.0726, 0.0716], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0127, 0.0151, 0.0137, 0.0129, 0.0122, 0.0132, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-28 22:50:08,327 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:32,390 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:50:46,184 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:51:17,195 INFO [train.py:904] (3/8) Epoch 8, batch 9000, loss[loss=0.1969, simple_loss=0.286, pruned_loss=0.05393, over 16785.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2821, pruned_loss=0.05045, over 3072228.67 frames. ], batch size: 124, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:51:17,196 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 22:51:25,078 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6121, 4.9026, 3.4278, 5.2572, 3.9888, 5.1500, 3.7293, 4.2621], device='cuda:3'), covar=tensor([0.0117, 0.0162, 0.0933, 0.0080, 0.0457, 0.0384, 0.0846, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0149, 0.0177, 0.0098, 0.0159, 0.0186, 0.0187, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-28 22:51:27,534 INFO [train.py:938] (3/8) Epoch 8, validation: loss=0.1608, simple_loss=0.265, pruned_loss=0.02828, over 944034.00 frames. 2023-04-28 22:51:27,534 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 22:52:04,336 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:52:26,507 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:53:14,309 INFO [train.py:904] (3/8) Epoch 8, batch 9050, loss[loss=0.1882, simple_loss=0.2732, pruned_loss=0.05158, over 16193.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2827, pruned_loss=0.05088, over 3080764.35 frames. ], batch size: 165, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:53:35,355 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.717e+02 2.777e+02 3.475e+02 4.225e+02 7.695e+02, threshold=6.949e+02, percent-clipped=5.0 2023-04-28 22:53:48,901 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:54:59,313 INFO [train.py:904] (3/8) Epoch 8, batch 9100, loss[loss=0.2033, simple_loss=0.2985, pruned_loss=0.05408, over 16310.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2834, pruned_loss=0.0518, over 3078708.78 frames. ], batch size: 146, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:55:42,669 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4167, 1.8835, 1.9843, 3.9681, 1.8020, 2.3828, 1.9906, 2.0310], device='cuda:3'), covar=tensor([0.0787, 0.3275, 0.2048, 0.0366, 0.3986, 0.2151, 0.2911, 0.3266], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0352, 0.0301, 0.0305, 0.0388, 0.0387, 0.0315, 0.0413], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 22:56:46,150 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8017, 1.3212, 1.6304, 1.6900, 1.8134, 1.8216, 1.5589, 1.7524], device='cuda:3'), covar=tensor([0.0125, 0.0263, 0.0132, 0.0175, 0.0167, 0.0115, 0.0269, 0.0070], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0156, 0.0141, 0.0137, 0.0144, 0.0102, 0.0154, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 22:56:59,467 INFO [train.py:904] (3/8) Epoch 8, batch 9150, loss[loss=0.2057, simple_loss=0.3034, pruned_loss=0.05402, over 16938.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2836, pruned_loss=0.05142, over 3059201.58 frames. ], batch size: 109, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:57:20,186 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.002e+02 2.760e+02 3.111e+02 3.901e+02 6.426e+02, threshold=6.222e+02, percent-clipped=0.0 2023-04-28 22:58:26,815 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:58:44,956 INFO [train.py:904] (3/8) Epoch 8, batch 9200, loss[loss=0.149, simple_loss=0.2386, pruned_loss=0.02974, over 17071.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2782, pruned_loss=0.04985, over 3058276.87 frames. ], batch size: 50, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:59:43,171 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:22,120 INFO [train.py:904] (3/8) Epoch 8, batch 9250, loss[loss=0.1656, simple_loss=0.2617, pruned_loss=0.03472, over 16875.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.278, pruned_loss=0.0499, over 3051022.33 frames. ], batch size: 90, lr: 8.34e-03, grad_scale: 4.0 2023-04-28 23:00:42,872 INFO [optim.py:368] (3/8) 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 23:00:44,614 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:57,266 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 23:01:56,408 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:02:04,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5167, 3.4952, 3.5002, 3.0780, 3.4427, 2.0991, 3.2303, 2.9328], device='cuda:3'), covar=tensor([0.0099, 0.0089, 0.0103, 0.0220, 0.0075, 0.1726, 0.0114, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0107, 0.0095, 0.0139, 0.0130, 0.0111, 0.0163, 0.0126, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:02:12,075 INFO [train.py:904] (3/8) Epoch 8, batch 9300, loss[loss=0.1684, simple_loss=0.2496, pruned_loss=0.04357, over 12012.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2758, pruned_loss=0.04914, over 3019064.92 frames. ], batch size: 246, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:02:31,767 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:42,302 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 9350, loss[loss=0.1971, simple_loss=0.2843, pruned_loss=0.05496, over 15305.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2755, pruned_loss=0.04882, over 3040323.08 frames. ], batch size: 191, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:04:22,274 INFO [optim.py:368] (3/8) 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 23:04:30,118 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0595, 3.4015, 3.3579, 2.1748, 3.2368, 3.3666, 3.2568, 1.7916], device='cuda:3'), covar=tensor([0.0387, 0.0030, 0.0035, 0.0325, 0.0054, 0.0065, 0.0053, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0060, 0.0062, 0.0118, 0.0068, 0.0077, 0.0068, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 23:04:34,137 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:05:40,898 INFO [train.py:904] (3/8) Epoch 8, batch 9400, loss[loss=0.1939, simple_loss=0.298, pruned_loss=0.0449, over 16639.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2749, pruned_loss=0.04844, over 3020127.54 frames. ], batch size: 83, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:05:46,243 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:05,191 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:10,060 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:07:07,264 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 9450, loss[loss=0.1755, simple_loss=0.2737, pruned_loss=0.03867, over 16440.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2782, pruned_loss=0.04916, over 3048085.63 frames. ], batch size: 68, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:07:38,811 INFO [optim.py:368] (3/8) 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 23:07:53,861 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 23:08:04,893 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:08:58,863 INFO [train.py:904] (3/8) Epoch 8, batch 9500, loss[loss=0.1775, simple_loss=0.272, pruned_loss=0.04146, over 12827.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2771, pruned_loss=0.04857, over 3043637.88 frames. ], batch size: 250, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:09:08,286 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:09:53,705 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4781, 3.3438, 2.6442, 2.1238, 2.2878, 2.1574, 3.3881, 3.1036], device='cuda:3'), covar=tensor([0.2319, 0.0618, 0.1386, 0.2030, 0.2065, 0.1674, 0.0417, 0.0867], device='cuda:3'), in_proj_covar=tensor([0.0278, 0.0240, 0.0264, 0.0252, 0.0243, 0.0202, 0.0244, 0.0256], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:10:36,541 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9746, 3.9988, 3.8444, 3.6701, 3.5271, 3.9612, 3.6085, 3.7225], device='cuda:3'), covar=tensor([0.0523, 0.0525, 0.0282, 0.0246, 0.0751, 0.0367, 0.0788, 0.0500], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0239, 0.0239, 0.0213, 0.0256, 0.0242, 0.0163, 0.0272], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:10:46,570 INFO [train.py:904] (3/8) Epoch 8, batch 9550, loss[loss=0.1847, simple_loss=0.265, pruned_loss=0.05217, over 12380.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2768, pruned_loss=0.04889, over 3044553.51 frames. ], batch size: 246, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:11:10,120 INFO [optim.py:368] (3/8) 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 23:11:57,096 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 23:12:03,959 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 9600, loss[loss=0.2142, simple_loss=0.309, pruned_loss=0.05969, over 16363.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2786, pruned_loss=0.04955, over 3068362.37 frames. ], batch size: 146, lr: 8.32e-03, grad_scale: 8.0 2023-04-28 23:13:40,982 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5822, 4.8276, 4.9477, 4.8523, 4.7742, 5.3702, 4.8851, 4.6015], device='cuda:3'), covar=tensor([0.0903, 0.1657, 0.1380, 0.1628, 0.2430, 0.0867, 0.1433, 0.2369], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0417, 0.0446, 0.0369, 0.0484, 0.0463, 0.0361, 0.0484], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:13:41,126 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3642, 3.6120, 3.6788, 2.6450, 3.4488, 3.6471, 3.5435, 2.0483], device='cuda:3'), covar=tensor([0.0346, 0.0023, 0.0029, 0.0231, 0.0048, 0.0055, 0.0041, 0.0364], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0059, 0.0061, 0.0115, 0.0066, 0.0075, 0.0066, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 23:14:15,056 INFO [train.py:904] (3/8) Epoch 8, batch 9650, loss[loss=0.2134, simple_loss=0.3012, pruned_loss=0.06275, over 16332.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.28, pruned_loss=0.04976, over 3057395.79 frames. ], batch size: 146, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:14:42,862 INFO [optim.py:368] (3/8) 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 23:15:29,960 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:15:59,694 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:16:03,270 INFO [train.py:904] (3/8) Epoch 8, batch 9700, loss[loss=0.1889, simple_loss=0.281, pruned_loss=0.04837, over 16528.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2788, pruned_loss=0.04934, over 3075024.21 frames. ], batch size: 68, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:16:43,463 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1909, 4.2315, 4.3910, 4.3262, 4.3262, 4.7831, 4.4504, 4.1425], device='cuda:3'), covar=tensor([0.1323, 0.1875, 0.1649, 0.1724, 0.2452, 0.1055, 0.1233, 0.2200], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0418, 0.0447, 0.0370, 0.0484, 0.0466, 0.0361, 0.0484], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:16:46,602 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:16:57,348 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7676, 1.2844, 1.6379, 1.6916, 1.7952, 1.8240, 1.5097, 1.7216], device='cuda:3'), covar=tensor([0.0162, 0.0231, 0.0133, 0.0157, 0.0160, 0.0132, 0.0245, 0.0065], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0155, 0.0140, 0.0138, 0.0144, 0.0102, 0.0154, 0.0093], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 23:17:38,592 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:41,986 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9846, 4.0463, 3.8553, 3.6945, 3.5968, 3.9565, 3.6736, 3.7736], device='cuda:3'), covar=tensor([0.0473, 0.0356, 0.0254, 0.0210, 0.0661, 0.0308, 0.0784, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0205, 0.0237, 0.0239, 0.0211, 0.0254, 0.0242, 0.0164, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:17:46,348 INFO [train.py:904] (3/8) Epoch 8, batch 9750, loss[loss=0.1864, simple_loss=0.2812, pruned_loss=0.0458, over 16366.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2782, pruned_loss=0.04961, over 3085171.49 frames. ], batch size: 146, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:18:08,278 INFO [optim.py:368] (3/8) 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 23:18:20,487 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:18:55,692 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:24,253 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 9800, loss[loss=0.1743, simple_loss=0.2594, pruned_loss=0.04464, over 12276.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2787, pruned_loss=0.04897, over 3088745.99 frames. ], batch size: 248, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:20:09,668 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 23:21:11,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3637, 3.0044, 2.5963, 2.1811, 2.1497, 2.1188, 2.8863, 2.8371], device='cuda:3'), covar=tensor([0.2159, 0.0724, 0.1315, 0.1867, 0.2079, 0.1598, 0.0408, 0.1043], device='cuda:3'), in_proj_covar=tensor([0.0281, 0.0240, 0.0266, 0.0253, 0.0243, 0.0202, 0.0244, 0.0256], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:21:11,974 INFO [train.py:904] (3/8) Epoch 8, batch 9850, loss[loss=0.1804, simple_loss=0.2643, pruned_loss=0.04826, over 12469.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2795, pruned_loss=0.0484, over 3085281.40 frames. ], batch size: 250, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:21:33,316 INFO [optim.py:368] (3/8) 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 23:21:48,504 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1497, 4.1330, 3.9305, 3.3897, 4.0699, 1.5120, 3.8489, 3.8010], device='cuda:3'), covar=tensor([0.0082, 0.0074, 0.0142, 0.0299, 0.0081, 0.2311, 0.0115, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0106, 0.0094, 0.0136, 0.0126, 0.0109, 0.0162, 0.0123, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-28 23:22:37,322 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:22:42,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9530, 4.2112, 4.0034, 4.0576, 3.7250, 3.8054, 3.9020, 4.1818], device='cuda:3'), covar=tensor([0.0889, 0.0847, 0.0896, 0.0568, 0.0733, 0.1536, 0.0908, 0.0917], device='cuda:3'), in_proj_covar=tensor([0.0452, 0.0576, 0.0472, 0.0393, 0.0356, 0.0382, 0.0472, 0.0427], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:23:04,041 INFO [train.py:904] (3/8) Epoch 8, batch 9900, loss[loss=0.2011, simple_loss=0.2978, pruned_loss=0.05224, over 16388.00 frames. ], tot_loss[loss=0.188, simple_loss=0.28, pruned_loss=0.04806, over 3089202.80 frames. ], batch size: 146, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:23:30,358 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 23:24:29,642 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:25:03,327 INFO [train.py:904] (3/8) Epoch 8, batch 9950, loss[loss=0.1845, simple_loss=0.2857, pruned_loss=0.04165, over 16403.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.281, pruned_loss=0.04799, over 3078886.12 frames. ], batch size: 146, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:25:14,735 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-28 23:25:29,444 INFO [optim.py:368] (3/8) 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 23:27:01,978 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:27:07,187 INFO [train.py:904] (3/8) Epoch 8, batch 10000, loss[loss=0.1647, simple_loss=0.2529, pruned_loss=0.03819, over 17049.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2796, pruned_loss=0.04748, over 3092239.82 frames. ], batch size: 53, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:28:30,672 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:28:40,811 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:28:47,993 INFO [train.py:904] (3/8) Epoch 8, batch 10050, loss[loss=0.2134, simple_loss=0.3008, pruned_loss=0.06297, over 16648.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2795, pruned_loss=0.04738, over 3089183.51 frames. ], batch size: 134, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:29:08,274 INFO [optim.py:368] (3/8) 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 23:29:15,154 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:21,182 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:38,333 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1215, 5.4433, 5.1589, 5.1725, 4.7870, 4.7795, 4.8102, 5.4547], device='cuda:3'), covar=tensor([0.0843, 0.0742, 0.0829, 0.0498, 0.0694, 0.0730, 0.0882, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0448, 0.0572, 0.0471, 0.0390, 0.0353, 0.0380, 0.0471, 0.0428], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:29:40,278 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:10,758 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8566, 2.5855, 2.4834, 1.8796, 2.4664, 2.6145, 2.4141, 1.7838], device='cuda:3'), covar=tensor([0.0294, 0.0046, 0.0040, 0.0254, 0.0079, 0.0066, 0.0063, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0060, 0.0062, 0.0117, 0.0067, 0.0076, 0.0067, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-28 23:30:18,763 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 8, batch 10100, loss[loss=0.1784, simple_loss=0.2665, pruned_loss=0.04521, over 16163.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2801, pruned_loss=0.04778, over 3092831.39 frames. ], batch size: 165, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:30:23,208 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 23:30:35,793 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6519, 2.0522, 1.6493, 1.8442, 2.4521, 2.1185, 2.5129, 2.6250], device='cuda:3'), covar=tensor([0.0059, 0.0294, 0.0373, 0.0316, 0.0176, 0.0245, 0.0102, 0.0138], device='cuda:3'), in_proj_covar=tensor([0.0105, 0.0183, 0.0178, 0.0179, 0.0176, 0.0181, 0.0168, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:30:51,850 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:14,350 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:37,454 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:32:08,879 INFO [train.py:904] (3/8) Epoch 9, batch 0, loss[loss=0.2834, simple_loss=0.3404, pruned_loss=0.1132, over 16629.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3404, pruned_loss=0.1132, over 16629.00 frames. ], batch size: 134, lr: 7.85e-03, grad_scale: 8.0 2023-04-28 23:32:08,879 INFO [train.py:929] (3/8) Computing validation loss 2023-04-28 23:32:16,264 INFO [train.py:938] (3/8) Epoch 9, validation: loss=0.1602, simple_loss=0.2637, pruned_loss=0.02837, over 944034.00 frames. 2023-04-28 23:32:16,265 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-28 23:32:36,784 INFO [optim.py:368] (3/8) 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 23:33:25,081 INFO [train.py:904] (3/8) Epoch 9, batch 50, loss[loss=0.1829, simple_loss=0.269, pruned_loss=0.0484, over 17127.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2909, pruned_loss=0.06604, over 759807.39 frames. ], batch size: 47, lr: 7.85e-03, grad_scale: 1.0 2023-04-28 23:33:37,269 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9253, 3.1626, 2.8041, 5.0578, 4.4097, 4.6618, 1.5823, 3.5328], device='cuda:3'), covar=tensor([0.1339, 0.0618, 0.1127, 0.0131, 0.0288, 0.0351, 0.1510, 0.0662], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0153, 0.0174, 0.0117, 0.0180, 0.0202, 0.0173, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 23:33:57,877 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9417, 1.7411, 2.3552, 2.8046, 2.8050, 2.8020, 1.8104, 3.0235], device='cuda:3'), covar=tensor([0.0097, 0.0288, 0.0188, 0.0149, 0.0141, 0.0126, 0.0307, 0.0065], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0159, 0.0145, 0.0142, 0.0149, 0.0106, 0.0158, 0.0096], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 23:34:31,265 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:34,836 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 100, loss[loss=0.2001, simple_loss=0.2918, pruned_loss=0.05421, over 17259.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2873, pruned_loss=0.06478, over 1333942.97 frames. ], batch size: 52, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:34:54,536 INFO [optim.py:368] (3/8) 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 23:35:38,086 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1611, 5.0071, 4.9419, 4.5770, 4.5200, 4.9791, 5.0737, 4.5640], device='cuda:3'), covar=tensor([0.0511, 0.0443, 0.0252, 0.0244, 0.0949, 0.0363, 0.0227, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0213, 0.0249, 0.0248, 0.0221, 0.0269, 0.0253, 0.0170, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:35:42,947 INFO [train.py:904] (3/8) Epoch 9, batch 150, loss[loss=0.227, simple_loss=0.3089, pruned_loss=0.07249, over 16692.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2839, pruned_loss=0.06164, over 1778571.94 frames. ], batch size: 62, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:35:55,837 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:35:58,255 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:40,006 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 200, loss[loss=0.255, simple_loss=0.3225, pruned_loss=0.0938, over 12485.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2842, pruned_loss=0.06201, over 2112826.48 frames. ], batch size: 246, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:37:13,043 INFO [optim.py:368] (3/8) 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 23:37:25,955 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8152, 1.6616, 2.1617, 2.5476, 2.5853, 2.4784, 1.8410, 2.8225], device='cuda:3'), covar=tensor([0.0099, 0.0294, 0.0196, 0.0192, 0.0157, 0.0158, 0.0270, 0.0056], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0160, 0.0146, 0.0145, 0.0151, 0.0108, 0.0159, 0.0098], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 23:37:31,967 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:37:45,243 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:01,974 INFO [train.py:904] (3/8) Epoch 9, batch 250, loss[loss=0.193, simple_loss=0.2659, pruned_loss=0.06007, over 16398.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2818, pruned_loss=0.06189, over 2384997.08 frames. ], batch size: 146, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:38:08,899 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7234, 2.8296, 2.3258, 2.5708, 3.1689, 2.9690, 3.6089, 3.3803], device='cuda:3'), covar=tensor([0.0049, 0.0234, 0.0299, 0.0252, 0.0146, 0.0214, 0.0149, 0.0133], device='cuda:3'), in_proj_covar=tensor([0.0114, 0.0189, 0.0184, 0.0184, 0.0182, 0.0187, 0.0181, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:38:29,683 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:35,235 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1345, 1.5660, 2.4223, 2.8903, 2.6299, 3.2541, 1.9363, 3.1708], device='cuda:3'), covar=tensor([0.0107, 0.0339, 0.0200, 0.0176, 0.0191, 0.0123, 0.0319, 0.0102], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0161, 0.0147, 0.0145, 0.0151, 0.0108, 0.0159, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 23:38:36,871 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:52,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3169, 1.5631, 2.6761, 3.1139, 2.8188, 3.2598, 1.8665, 3.3708], device='cuda:3'), covar=tensor([0.0105, 0.0384, 0.0192, 0.0155, 0.0186, 0.0151, 0.0398, 0.0083], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0161, 0.0148, 0.0146, 0.0152, 0.0109, 0.0160, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-28 23:39:10,399 INFO [train.py:904] (3/8) Epoch 9, batch 300, loss[loss=0.1758, simple_loss=0.2625, pruned_loss=0.04457, over 17110.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2791, pruned_loss=0.06013, over 2583145.42 frames. ], batch size: 47, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:39:29,677 INFO [optim.py:368] (3/8) 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 23:39:34,457 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7276, 4.5791, 4.5613, 4.3305, 4.1912, 4.5935, 4.4503, 4.2995], device='cuda:3'), covar=tensor([0.0570, 0.0458, 0.0255, 0.0232, 0.0855, 0.0400, 0.0443, 0.0605], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0261, 0.0258, 0.0230, 0.0283, 0.0264, 0.0178, 0.0299], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:39:38,467 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0558, 3.3061, 3.5428, 3.5029, 3.5000, 3.3199, 3.1378, 3.3631], device='cuda:3'), covar=tensor([0.0631, 0.0645, 0.0541, 0.0614, 0.0693, 0.0543, 0.1202, 0.0604], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0296, 0.0299, 0.0293, 0.0334, 0.0318, 0.0407, 0.0258], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-28 23:40:04,079 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5910, 2.0806, 2.2226, 4.2761, 2.0025, 2.5569, 2.2441, 2.2494], device='cuda:3'), covar=tensor([0.0859, 0.3159, 0.1989, 0.0404, 0.3713, 0.2069, 0.2895, 0.3232], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0368, 0.0311, 0.0321, 0.0399, 0.0403, 0.0328, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:40:17,833 INFO [train.py:904] (3/8) Epoch 9, batch 350, loss[loss=0.2049, simple_loss=0.2687, pruned_loss=0.07057, over 16902.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2774, pruned_loss=0.05925, over 2736502.45 frames. ], batch size: 109, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:40:26,545 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5643, 3.7144, 2.1167, 3.8659, 2.7221, 3.7845, 2.0698, 2.8597], device='cuda:3'), covar=tensor([0.0218, 0.0307, 0.1457, 0.0174, 0.0784, 0.0621, 0.1416, 0.0657], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0159, 0.0183, 0.0108, 0.0164, 0.0195, 0.0192, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 23:41:16,895 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9747, 5.4506, 5.6202, 5.4351, 5.4570, 6.0767, 5.5413, 5.2560], device='cuda:3'), covar=tensor([0.0777, 0.1666, 0.1754, 0.1815, 0.2448, 0.0865, 0.1343, 0.2201], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0457, 0.0487, 0.0406, 0.0536, 0.0508, 0.0388, 0.0541], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-28 23:41:23,824 INFO [train.py:904] (3/8) Epoch 9, batch 400, loss[loss=0.2135, simple_loss=0.2819, pruned_loss=0.07252, over 16451.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2763, pruned_loss=0.05892, over 2868314.76 frames. ], batch size: 75, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:41:43,586 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.386e+02 2.843e+02 3.463e+02 6.249e+02, threshold=5.687e+02, percent-clipped=1.0 2023-04-28 23:41:52,432 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1467, 5.1321, 4.9243, 4.3446, 4.9853, 1.7120, 4.7123, 4.9390], device='cuda:3'), covar=tensor([0.0067, 0.0063, 0.0148, 0.0371, 0.0080, 0.2489, 0.0122, 0.0155], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0104, 0.0152, 0.0143, 0.0122, 0.0175, 0.0137, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:42:33,320 INFO [train.py:904] (3/8) Epoch 9, batch 450, loss[loss=0.2016, simple_loss=0.2892, pruned_loss=0.05702, over 17021.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2743, pruned_loss=0.05724, over 2974346.62 frames. ], batch size: 55, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:42:37,318 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:40,764 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:55,112 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 23:42:58,057 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:43:42,339 INFO [train.py:904] (3/8) Epoch 9, batch 500, loss[loss=0.1887, simple_loss=0.2602, pruned_loss=0.05859, over 16837.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2727, pruned_loss=0.0565, over 3046138.87 frames. ], batch size: 96, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:43:45,771 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5208, 5.9845, 5.6808, 5.7670, 5.2737, 5.1491, 5.4451, 6.0220], device='cuda:3'), covar=tensor([0.1153, 0.0758, 0.0997, 0.0523, 0.0859, 0.0694, 0.0828, 0.0803], device='cuda:3'), in_proj_covar=tensor([0.0498, 0.0638, 0.0531, 0.0435, 0.0394, 0.0414, 0.0528, 0.0471], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:44:01,211 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5189, 3.7225, 3.8794, 2.0864, 4.0223, 4.0728, 3.1082, 2.9636], device='cuda:3'), covar=tensor([0.0728, 0.0129, 0.0143, 0.1066, 0.0054, 0.0103, 0.0343, 0.0391], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0096, 0.0082, 0.0142, 0.0068, 0.0092, 0.0118, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 23:44:01,900 INFO [optim.py:368] (3/8) 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 23:44:22,643 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 23:44:39,597 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7653, 3.6598, 3.8143, 3.9484, 4.0537, 3.5828, 3.8885, 4.0389], device='cuda:3'), covar=tensor([0.1121, 0.0937, 0.1103, 0.0621, 0.0516, 0.1621, 0.1170, 0.0611], device='cuda:3'), in_proj_covar=tensor([0.0494, 0.0605, 0.0755, 0.0617, 0.0460, 0.0456, 0.0484, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:44:50,826 INFO [train.py:904] (3/8) Epoch 9, batch 550, loss[loss=0.2416, simple_loss=0.3038, pruned_loss=0.08971, over 16341.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2721, pruned_loss=0.05647, over 3096360.46 frames. ], batch size: 165, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:44:57,849 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7009, 3.8248, 3.9653, 2.2854, 4.1458, 4.1819, 3.2496, 3.0441], device='cuda:3'), covar=tensor([0.0681, 0.0128, 0.0139, 0.0950, 0.0055, 0.0094, 0.0299, 0.0390], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0097, 0.0083, 0.0143, 0.0068, 0.0093, 0.0119, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-28 23:45:19,860 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:45:29,114 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 23:46:01,931 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8960, 4.0175, 3.7605, 3.5778, 3.3011, 3.9035, 3.5492, 3.5767], device='cuda:3'), covar=tensor([0.0693, 0.0571, 0.0341, 0.0301, 0.1013, 0.0470, 0.1141, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0229, 0.0269, 0.0265, 0.0237, 0.0292, 0.0273, 0.0182, 0.0306], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:46:02,621 INFO [train.py:904] (3/8) Epoch 9, batch 600, loss[loss=0.2045, simple_loss=0.2703, pruned_loss=0.06937, over 15535.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2725, pruned_loss=0.05694, over 3149632.28 frames. ], batch size: 190, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:46:21,575 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.670e+02 3.296e+02 4.052e+02 8.860e+02, threshold=6.592e+02, percent-clipped=6.0 2023-04-28 23:46:26,574 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:46:32,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7047, 4.5592, 4.7189, 4.9361, 5.0742, 4.4903, 4.9931, 5.0309], device='cuda:3'), covar=tensor([0.1397, 0.1032, 0.1320, 0.0660, 0.0555, 0.0999, 0.0787, 0.0682], device='cuda:3'), in_proj_covar=tensor([0.0499, 0.0607, 0.0757, 0.0619, 0.0464, 0.0458, 0.0489, 0.0537], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:47:09,564 INFO [train.py:904] (3/8) Epoch 9, batch 650, loss[loss=0.2033, simple_loss=0.271, pruned_loss=0.06784, over 16481.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2714, pruned_loss=0.05626, over 3188275.48 frames. ], batch size: 146, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:47:18,702 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 23:48:18,160 INFO [train.py:904] (3/8) Epoch 9, batch 700, loss[loss=0.1874, simple_loss=0.279, pruned_loss=0.04792, over 17021.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2719, pruned_loss=0.05631, over 3210587.52 frames. ], batch size: 55, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:48:37,201 INFO [optim.py:368] (3/8) 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 23:48:51,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8084, 3.9209, 1.9690, 4.3388, 2.6085, 4.3606, 2.2784, 3.1620], device='cuda:3'), covar=tensor([0.0189, 0.0310, 0.1784, 0.0149, 0.0898, 0.0391, 0.1399, 0.0609], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0161, 0.0184, 0.0112, 0.0164, 0.0196, 0.0192, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 23:49:25,153 INFO [train.py:904] (3/8) Epoch 9, batch 750, loss[loss=0.2077, simple_loss=0.2757, pruned_loss=0.06982, over 12345.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2719, pruned_loss=0.05621, over 3234618.68 frames. ], batch size: 246, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:49:29,053 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:31,524 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:14,421 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:37,268 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8704, 2.1167, 2.2006, 4.6727, 2.1784, 2.7494, 2.2729, 2.4856], device='cuda:3'), covar=tensor([0.0814, 0.3306, 0.2043, 0.0284, 0.3548, 0.2066, 0.2723, 0.3189], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0371, 0.0314, 0.0324, 0.0400, 0.0411, 0.0330, 0.0436], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:50:37,833 INFO [train.py:904] (3/8) Epoch 9, batch 800, loss[loss=0.1735, simple_loss=0.263, pruned_loss=0.04202, over 17124.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2717, pruned_loss=0.05595, over 3256711.26 frames. ], batch size: 48, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:50:39,916 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:42,267 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 23:51:40,252 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 850, loss[loss=0.1901, simple_loss=0.2595, pruned_loss=0.06037, over 16855.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2699, pruned_loss=0.0555, over 3272188.72 frames. ], batch size: 116, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:51:49,599 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:51:58,509 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6814, 2.6234, 2.3063, 3.6438, 3.0557, 3.8053, 1.4277, 2.7593], device='cuda:3'), covar=tensor([0.1262, 0.0594, 0.1151, 0.0162, 0.0215, 0.0378, 0.1413, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0153, 0.0174, 0.0125, 0.0191, 0.0208, 0.0174, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 23:52:51,584 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6747, 3.1732, 2.6195, 4.8432, 3.9516, 4.5047, 1.5896, 3.2607], device='cuda:3'), covar=tensor([0.1341, 0.0557, 0.1144, 0.0127, 0.0288, 0.0346, 0.1437, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0154, 0.0175, 0.0126, 0.0192, 0.0210, 0.0175, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-28 23:52:54,645 INFO [train.py:904] (3/8) Epoch 9, batch 900, loss[loss=0.2004, simple_loss=0.2674, pruned_loss=0.06668, over 16356.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2693, pruned_loss=0.05481, over 3288846.17 frames. ], batch size: 165, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:53:13,857 INFO [optim.py:368] (3/8) 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 23:53:14,989 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:54:03,806 INFO [train.py:904] (3/8) Epoch 9, batch 950, loss[loss=0.1987, simple_loss=0.2834, pruned_loss=0.05705, over 17128.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2694, pruned_loss=0.05495, over 3301738.74 frames. ], batch size: 49, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:11,099 INFO [train.py:904] (3/8) Epoch 9, batch 1000, loss[loss=0.1794, simple_loss=0.2545, pruned_loss=0.05213, over 16882.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2686, pruned_loss=0.0545, over 3294867.43 frames. ], batch size: 96, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:31,949 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.342e+02 2.920e+02 3.464e+02 5.943e+02, threshold=5.839e+02, percent-clipped=0.0 2023-04-28 23:55:37,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1766, 2.4608, 2.4415, 4.8034, 2.3381, 2.9446, 2.5362, 2.7470], device='cuda:3'), covar=tensor([0.0682, 0.2981, 0.1934, 0.0314, 0.3412, 0.2046, 0.2501, 0.3156], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0327, 0.0401, 0.0416, 0.0333, 0.0439], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:56:11,952 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5084, 2.1622, 2.3629, 4.1500, 2.1595, 2.6659, 2.2570, 2.4399], device='cuda:3'), covar=tensor([0.0927, 0.3057, 0.1873, 0.0404, 0.3258, 0.1905, 0.2865, 0.2597], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0327, 0.0400, 0.0416, 0.0333, 0.0440], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:56:18,002 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0846, 2.5266, 1.9732, 2.2492, 2.9185, 2.6715, 3.2406, 3.0788], device='cuda:3'), covar=tensor([0.0117, 0.0240, 0.0338, 0.0280, 0.0148, 0.0220, 0.0155, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0194, 0.0189, 0.0189, 0.0189, 0.0192, 0.0194, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-28 23:56:20,445 INFO [train.py:904] (3/8) Epoch 9, batch 1050, loss[loss=0.1983, simple_loss=0.2679, pruned_loss=0.06434, over 12058.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2686, pruned_loss=0.05476, over 3304019.75 frames. ], batch size: 246, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:28,550 INFO [train.py:904] (3/8) Epoch 9, batch 1100, loss[loss=0.1764, simple_loss=0.2519, pruned_loss=0.05044, over 16732.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2677, pruned_loss=0.05378, over 3314993.01 frames. ], batch size: 102, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:38,397 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-28 23:57:47,218 INFO [optim.py:368] (3/8) 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 23:58:01,947 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:24,252 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:28,015 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:35,308 INFO [train.py:904] (3/8) Epoch 9, batch 1150, loss[loss=0.1807, simple_loss=0.271, pruned_loss=0.04523, over 17015.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2677, pruned_loss=0.05331, over 3327342.20 frames. ], batch size: 55, lr: 7.79e-03, grad_scale: 4.0 2023-04-28 23:58:54,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7880, 3.9487, 2.1876, 4.2822, 2.8448, 4.2619, 2.4379, 3.0261], device='cuda:3'), covar=tensor([0.0203, 0.0272, 0.1521, 0.0157, 0.0761, 0.0401, 0.1299, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0163, 0.0185, 0.0114, 0.0165, 0.0202, 0.0193, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-28 23:59:04,894 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:44,518 INFO [train.py:904] (3/8) Epoch 9, batch 1200, loss[loss=0.1763, simple_loss=0.27, pruned_loss=0.04126, over 17066.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2673, pruned_loss=0.05264, over 3326741.11 frames. ], batch size: 53, lr: 7.79e-03, grad_scale: 8.0 2023-04-28 23:59:50,720 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 9, batch 1250, loss[loss=0.1609, simple_loss=0.2451, pruned_loss=0.03832, over 16847.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2678, pruned_loss=0.0538, over 3326697.24 frames. ], batch size: 42, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:00:57,319 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 00:01:47,013 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:01:58,480 INFO [train.py:904] (3/8) Epoch 9, batch 1300, loss[loss=0.1653, simple_loss=0.2471, pruned_loss=0.04175, over 16826.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.266, pruned_loss=0.05293, over 3323419.59 frames. ], batch size: 39, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:02:18,040 INFO [optim.py:368] (3/8) 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 00:03:05,224 INFO [train.py:904] (3/8) Epoch 9, batch 1350, loss[loss=0.1917, simple_loss=0.2745, pruned_loss=0.05444, over 16534.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2667, pruned_loss=0.05277, over 3333441.74 frames. ], batch size: 68, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:03:07,976 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 00:03:10,422 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0653, 1.8773, 2.4554, 2.9112, 2.8001, 3.4990, 2.2494, 3.3113], device='cuda:3'), covar=tensor([0.0125, 0.0301, 0.0197, 0.0179, 0.0164, 0.0102, 0.0269, 0.0085], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0163, 0.0149, 0.0150, 0.0154, 0.0111, 0.0163, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 00:03:11,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7901, 4.0549, 4.2811, 3.1057, 3.7013, 4.1240, 3.7805, 2.6213], device='cuda:3'), covar=tensor([0.0330, 0.0072, 0.0031, 0.0218, 0.0055, 0.0075, 0.0055, 0.0281], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0067, 0.0065, 0.0121, 0.0070, 0.0080, 0.0071, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:03:29,881 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:03:45,941 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 00:04:12,576 INFO [train.py:904] (3/8) Epoch 9, batch 1400, loss[loss=0.1943, simple_loss=0.2606, pruned_loss=0.06401, over 16350.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2657, pruned_loss=0.0519, over 3331409.12 frames. ], batch size: 165, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:04:33,568 INFO [optim.py:368] (3/8) 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 00:04:53,120 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:09,577 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:22,655 INFO [train.py:904] (3/8) Epoch 9, batch 1450, loss[loss=0.1609, simple_loss=0.2515, pruned_loss=0.03513, over 17211.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2651, pruned_loss=0.05207, over 3319313.38 frames. ], batch size: 44, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:16,062 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 1500, loss[loss=0.1936, simple_loss=0.2647, pruned_loss=0.06123, over 16224.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2648, pruned_loss=0.05246, over 3319493.36 frames. ], batch size: 165, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:29,980 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:42,762 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:51,612 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.540e+02 3.048e+02 3.562e+02 8.559e+02, threshold=6.096e+02, percent-clipped=4.0 2023-04-29 00:07:31,403 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1403, 4.1855, 4.6088, 4.5824, 4.6242, 4.2416, 4.3400, 4.1443], device='cuda:3'), covar=tensor([0.0365, 0.0641, 0.0379, 0.0407, 0.0389, 0.0385, 0.0765, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0317, 0.0323, 0.0327, 0.0311, 0.0368, 0.0343, 0.0444, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 00:07:34,670 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5592, 2.1187, 2.3994, 4.2302, 2.1742, 2.8381, 2.2605, 2.3811], device='cuda:3'), covar=tensor([0.0827, 0.2856, 0.1664, 0.0353, 0.3040, 0.1590, 0.2580, 0.2387], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0375, 0.0317, 0.0327, 0.0401, 0.0421, 0.0335, 0.0442], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:07:39,171 INFO [train.py:904] (3/8) Epoch 9, batch 1550, loss[loss=0.2141, simple_loss=0.2761, pruned_loss=0.07604, over 16696.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2661, pruned_loss=0.05316, over 3325576.85 frames. ], batch size: 89, lr: 7.77e-03, grad_scale: 4.0 2023-04-29 00:07:49,786 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:07:51,012 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0678, 3.9148, 4.1268, 4.2738, 4.3557, 3.9235, 4.0906, 4.3342], device='cuda:3'), covar=tensor([0.1163, 0.0920, 0.1107, 0.0561, 0.0472, 0.1325, 0.1902, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0526, 0.0640, 0.0807, 0.0658, 0.0492, 0.0486, 0.0512, 0.0570], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:07:52,795 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 00:07:55,144 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 00:08:48,744 INFO [train.py:904] (3/8) Epoch 9, batch 1600, loss[loss=0.2044, simple_loss=0.2948, pruned_loss=0.057, over 17027.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2686, pruned_loss=0.05437, over 3330289.43 frames. ], batch size: 55, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:02,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1930, 4.0678, 4.3855, 1.9417, 4.7059, 4.6520, 3.4176, 3.6424], device='cuda:3'), covar=tensor([0.0566, 0.0156, 0.0157, 0.1134, 0.0042, 0.0104, 0.0311, 0.0338], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0098, 0.0086, 0.0141, 0.0070, 0.0097, 0.0119, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 00:09:09,710 INFO [optim.py:368] (3/8) 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 00:09:11,369 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9845, 4.1938, 2.2986, 4.6070, 2.8812, 4.5910, 2.2967, 3.3022], device='cuda:3'), covar=tensor([0.0201, 0.0275, 0.1489, 0.0179, 0.0828, 0.0449, 0.1501, 0.0627], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0164, 0.0186, 0.0118, 0.0166, 0.0205, 0.0196, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 00:09:31,317 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 00:09:52,304 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:09:56,031 INFO [train.py:904] (3/8) Epoch 9, batch 1650, loss[loss=0.2005, simple_loss=0.2864, pruned_loss=0.05731, over 16644.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2709, pruned_loss=0.05516, over 3336820.91 frames. ], batch size: 57, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:10:36,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1355, 5.7052, 5.8718, 5.6212, 5.7144, 6.2290, 5.8779, 5.5431], device='cuda:3'), covar=tensor([0.0752, 0.1491, 0.1587, 0.1832, 0.2482, 0.0904, 0.1199, 0.2183], device='cuda:3'), in_proj_covar=tensor([0.0341, 0.0474, 0.0504, 0.0420, 0.0558, 0.0532, 0.0402, 0.0560], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:11:03,605 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 00:11:05,716 INFO [train.py:904] (3/8) Epoch 9, batch 1700, loss[loss=0.1683, simple_loss=0.256, pruned_loss=0.04029, over 16774.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2721, pruned_loss=0.0558, over 3337165.04 frames. ], batch size: 39, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:11:24,536 INFO [optim.py:368] (3/8) 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 00:11:32,121 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4698, 3.8024, 4.0752, 2.7660, 3.7513, 4.0923, 3.8655, 2.3409], device='cuda:3'), covar=tensor([0.0357, 0.0122, 0.0032, 0.0261, 0.0051, 0.0056, 0.0040, 0.0303], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0067, 0.0065, 0.0121, 0.0070, 0.0080, 0.0071, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:11:37,166 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:11:40,646 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3070, 4.2759, 4.2178, 3.7570, 4.2587, 1.8125, 4.0101, 3.9256], device='cuda:3'), covar=tensor([0.0080, 0.0069, 0.0117, 0.0254, 0.0068, 0.2099, 0.0108, 0.0156], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0111, 0.0161, 0.0155, 0.0130, 0.0176, 0.0145, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:11:40,868 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 00:12:06,747 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1477, 1.9566, 2.5492, 2.9854, 2.8801, 3.4919, 2.2171, 3.3848], device='cuda:3'), covar=tensor([0.0136, 0.0320, 0.0196, 0.0184, 0.0169, 0.0091, 0.0295, 0.0097], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0164, 0.0151, 0.0153, 0.0156, 0.0113, 0.0164, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 00:12:13,191 INFO [train.py:904] (3/8) Epoch 9, batch 1750, loss[loss=0.1808, simple_loss=0.2603, pruned_loss=0.05069, over 15953.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2719, pruned_loss=0.05546, over 3344451.00 frames. ], batch size: 35, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:12:28,810 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:12:32,351 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-04-29 00:12:33,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4164, 2.0463, 2.2635, 4.1440, 2.1058, 2.6284, 2.1741, 2.3107], device='cuda:3'), covar=tensor([0.0860, 0.2977, 0.1816, 0.0338, 0.3101, 0.1901, 0.2855, 0.2594], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0373, 0.0313, 0.0323, 0.0399, 0.0420, 0.0334, 0.0440], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:13:15,118 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5979, 3.2449, 2.6998, 4.9964, 4.2676, 4.7462, 1.3755, 3.3172], device='cuda:3'), covar=tensor([0.1405, 0.0587, 0.1116, 0.0161, 0.0248, 0.0318, 0.1531, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0152, 0.0174, 0.0126, 0.0196, 0.0210, 0.0171, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 00:13:19,614 INFO [train.py:904] (3/8) Epoch 9, batch 1800, loss[loss=0.1992, simple_loss=0.275, pruned_loss=0.06167, over 16421.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2729, pruned_loss=0.05541, over 3340268.22 frames. ], batch size: 146, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:13:20,535 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:13:50,196 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:26,663 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 1850, loss[loss=0.2236, simple_loss=0.3013, pruned_loss=0.07291, over 15632.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2745, pruned_loss=0.05612, over 3334153.33 frames. ], batch size: 191, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:37,693 INFO [train.py:904] (3/8) Epoch 9, batch 1900, loss[loss=0.1892, simple_loss=0.2683, pruned_loss=0.05509, over 16841.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2727, pruned_loss=0.05522, over 3336375.44 frames. ], batch size: 102, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:59,152 INFO [optim.py:368] (3/8) 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 00:16:17,066 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 00:16:27,388 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8490, 3.3666, 2.4181, 4.6115, 3.9430, 4.2386, 1.6522, 2.9806], device='cuda:3'), covar=tensor([0.1072, 0.0396, 0.1036, 0.0133, 0.0235, 0.0419, 0.1182, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0151, 0.0174, 0.0125, 0.0194, 0.0209, 0.0170, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 00:16:42,382 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:16:46,825 INFO [train.py:904] (3/8) Epoch 9, batch 1950, loss[loss=0.1887, simple_loss=0.2778, pruned_loss=0.04986, over 16637.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2727, pruned_loss=0.05485, over 3335058.97 frames. ], batch size: 57, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:17:35,710 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 00:17:48,402 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:17:49,290 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 2000, loss[loss=0.2046, simple_loss=0.2766, pruned_loss=0.06627, over 16720.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2732, pruned_loss=0.05448, over 3321646.17 frames. ], batch size: 89, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:18:17,512 INFO [optim.py:368] (3/8) 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 00:18:26,840 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:18:33,614 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-29 00:19:04,138 INFO [train.py:904] (3/8) Epoch 9, batch 2050, loss[loss=0.2245, simple_loss=0.2875, pruned_loss=0.08073, over 16879.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2731, pruned_loss=0.05509, over 3313527.35 frames. ], batch size: 116, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:19:12,744 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:33,895 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:48,932 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:20:13,198 INFO [train.py:904] (3/8) Epoch 9, batch 2100, loss[loss=0.1591, simple_loss=0.2485, pruned_loss=0.03482, over 17245.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2744, pruned_loss=0.05612, over 3308097.21 frames. ], batch size: 45, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:20:35,021 INFO [optim.py:368] (3/8) 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 00:20:37,089 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:11,822 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:20,737 INFO [train.py:904] (3/8) Epoch 9, batch 2150, loss[loss=0.2023, simple_loss=0.2744, pruned_loss=0.06513, over 15639.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2764, pruned_loss=0.05784, over 3298558.87 frames. ], batch size: 191, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:22:26,664 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0215, 2.5587, 2.6329, 1.8083, 2.8138, 2.7736, 2.3623, 2.3719], device='cuda:3'), covar=tensor([0.0710, 0.0206, 0.0235, 0.0924, 0.0092, 0.0187, 0.0452, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0098, 0.0087, 0.0142, 0.0070, 0.0098, 0.0121, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 00:22:31,366 INFO [train.py:904] (3/8) Epoch 9, batch 2200, loss[loss=0.2182, simple_loss=0.2936, pruned_loss=0.07139, over 16288.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2766, pruned_loss=0.05791, over 3303379.53 frames. ], batch size: 165, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:22:54,054 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.797e+02 3.398e+02 4.283e+02 9.169e+02, threshold=6.797e+02, percent-clipped=3.0 2023-04-29 00:23:21,315 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0890, 5.0027, 4.9883, 4.5988, 4.5342, 5.0067, 4.9344, 4.6091], device='cuda:3'), covar=tensor([0.0596, 0.0471, 0.0225, 0.0251, 0.1034, 0.0412, 0.0303, 0.0671], device='cuda:3'), in_proj_covar=tensor([0.0245, 0.0294, 0.0284, 0.0258, 0.0312, 0.0294, 0.0196, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:23:41,045 INFO [train.py:904] (3/8) Epoch 9, batch 2250, loss[loss=0.1937, simple_loss=0.2865, pruned_loss=0.05049, over 17023.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2776, pruned_loss=0.05776, over 3312319.86 frames. ], batch size: 55, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:24:49,200 INFO [train.py:904] (3/8) Epoch 9, batch 2300, loss[loss=0.2205, simple_loss=0.3026, pruned_loss=0.06925, over 16666.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.278, pruned_loss=0.05739, over 3311891.20 frames. ], batch size: 62, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:12,015 INFO [optim.py:368] (3/8) 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 00:25:33,804 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 00:25:59,045 INFO [train.py:904] (3/8) Epoch 9, batch 2350, loss[loss=0.2008, simple_loss=0.2746, pruned_loss=0.06351, over 16461.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2777, pruned_loss=0.05709, over 3320375.71 frames. ], batch size: 68, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:59,397 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:26:40,860 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 00:27:06,423 INFO [train.py:904] (3/8) Epoch 9, batch 2400, loss[loss=0.1721, simple_loss=0.2608, pruned_loss=0.04169, over 17216.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2772, pruned_loss=0.05633, over 3328571.23 frames. ], batch size: 45, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:27:29,672 INFO [optim.py:368] (3/8) 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 00:27:31,276 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:56,917 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1707, 5.1393, 4.8986, 4.3961, 4.9527, 2.1133, 4.7343, 5.0341], device='cuda:3'), covar=tensor([0.0059, 0.0055, 0.0140, 0.0294, 0.0069, 0.1985, 0.0102, 0.0130], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0111, 0.0161, 0.0153, 0.0129, 0.0175, 0.0145, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:27:59,096 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:28:14,402 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5993, 2.7490, 2.3804, 3.8595, 3.1828, 3.9438, 1.4748, 2.6474], device='cuda:3'), covar=tensor([0.1299, 0.0575, 0.1017, 0.0162, 0.0171, 0.0350, 0.1360, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0153, 0.0174, 0.0128, 0.0198, 0.0210, 0.0172, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 00:28:14,955 INFO [train.py:904] (3/8) Epoch 9, batch 2450, loss[loss=0.2085, simple_loss=0.2788, pruned_loss=0.06915, over 16447.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2782, pruned_loss=0.05663, over 3327470.49 frames. ], batch size: 146, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:28:35,102 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 2500, loss[loss=0.2158, simple_loss=0.3002, pruned_loss=0.06565, over 16653.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2786, pruned_loss=0.05671, over 3321219.44 frames. ], batch size: 57, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:29:23,005 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3400, 1.9966, 2.1950, 3.9981, 2.0189, 2.5883, 2.0927, 2.2294], device='cuda:3'), covar=tensor([0.0885, 0.3098, 0.1898, 0.0401, 0.3201, 0.1814, 0.2940, 0.2615], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0323, 0.0398, 0.0424, 0.0334, 0.0442], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:29:44,587 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.505e+02 2.999e+02 3.639e+02 7.354e+02, threshold=5.999e+02, percent-clipped=3.0 2023-04-29 00:30:28,782 INFO [train.py:904] (3/8) Epoch 9, batch 2550, loss[loss=0.1846, simple_loss=0.2824, pruned_loss=0.04339, over 17079.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2791, pruned_loss=0.05714, over 3324719.76 frames. ], batch size: 53, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:30:41,806 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:30:43,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8602, 2.2015, 2.2669, 4.6750, 2.0088, 2.8877, 2.2776, 2.4667], device='cuda:3'), covar=tensor([0.0736, 0.3106, 0.1991, 0.0299, 0.3629, 0.1946, 0.2623, 0.3084], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0375, 0.0316, 0.0324, 0.0398, 0.0425, 0.0334, 0.0442], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:31:05,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2202, 5.6774, 5.8276, 5.5279, 5.6317, 6.1799, 5.7834, 5.4550], device='cuda:3'), covar=tensor([0.0754, 0.1720, 0.1469, 0.1909, 0.2391, 0.0946, 0.1190, 0.2180], device='cuda:3'), in_proj_covar=tensor([0.0336, 0.0473, 0.0504, 0.0416, 0.0551, 0.0528, 0.0399, 0.0556], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:31:20,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6653, 6.0094, 5.7254, 5.8607, 5.3909, 5.2730, 5.5390, 6.1734], device='cuda:3'), covar=tensor([0.0936, 0.0848, 0.1167, 0.0654, 0.0825, 0.0581, 0.0821, 0.0835], device='cuda:3'), in_proj_covar=tensor([0.0509, 0.0648, 0.0543, 0.0440, 0.0400, 0.0410, 0.0532, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:31:38,616 INFO [train.py:904] (3/8) Epoch 9, batch 2600, loss[loss=0.2022, simple_loss=0.2845, pruned_loss=0.05994, over 15910.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2786, pruned_loss=0.057, over 3331173.02 frames. ], batch size: 35, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:31:59,378 INFO [optim.py:368] (3/8) 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 00:32:04,206 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:36,939 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 00:32:45,547 INFO [train.py:904] (3/8) Epoch 9, batch 2650, loss[loss=0.1995, simple_loss=0.2891, pruned_loss=0.05498, over 16536.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2789, pruned_loss=0.05648, over 3333353.62 frames. ], batch size: 68, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:32:45,933 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:54,737 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2884, 4.2119, 4.7086, 4.6849, 4.7044, 4.3073, 4.3529, 4.1930], device='cuda:3'), covar=tensor([0.0275, 0.0500, 0.0337, 0.0364, 0.0425, 0.0348, 0.0831, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0328, 0.0327, 0.0314, 0.0373, 0.0349, 0.0455, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 00:33:51,886 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 2700, loss[loss=0.1692, simple_loss=0.2672, pruned_loss=0.03561, over 17107.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2789, pruned_loss=0.05546, over 3336158.02 frames. ], batch size: 49, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:34:17,451 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.501e+02 2.856e+02 3.411e+02 7.585e+02, threshold=5.711e+02, percent-clipped=1.0 2023-04-29 00:34:34,676 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 00:34:47,071 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:34:58,111 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:02,459 INFO [train.py:904] (3/8) Epoch 9, batch 2750, loss[loss=0.2036, simple_loss=0.2863, pruned_loss=0.0605, over 16869.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2796, pruned_loss=0.05613, over 3331026.44 frames. ], batch size: 90, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:35:48,386 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:51,691 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:56,234 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8688, 2.1649, 2.2357, 4.7179, 2.0319, 2.8200, 2.3790, 2.4988], device='cuda:3'), covar=tensor([0.0735, 0.3128, 0.1974, 0.0286, 0.3640, 0.2026, 0.2638, 0.2867], device='cuda:3'), in_proj_covar=tensor([0.0356, 0.0377, 0.0317, 0.0326, 0.0401, 0.0428, 0.0335, 0.0446], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:36:15,269 INFO [train.py:904] (3/8) Epoch 9, batch 2800, loss[loss=0.1525, simple_loss=0.2457, pruned_loss=0.02966, over 17220.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2793, pruned_loss=0.05611, over 3325965.11 frames. ], batch size: 44, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:36:20,893 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0176, 3.4428, 2.9182, 1.7897, 2.4365, 2.0518, 3.2374, 3.4708], device='cuda:3'), covar=tensor([0.0201, 0.0545, 0.0649, 0.1790, 0.0947, 0.0941, 0.0604, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0143, 0.0157, 0.0142, 0.0135, 0.0125, 0.0136, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 00:36:25,637 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 00:36:36,295 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.751e+02 3.290e+02 3.928e+02 7.223e+02, threshold=6.581e+02, percent-clipped=1.0 2023-04-29 00:36:38,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5363, 3.9976, 4.1894, 2.9126, 3.6740, 4.0759, 3.7837, 2.5274], device='cuda:3'), covar=tensor([0.0374, 0.0062, 0.0029, 0.0249, 0.0062, 0.0068, 0.0057, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0069, 0.0067, 0.0124, 0.0075, 0.0084, 0.0075, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:37:16,059 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 2850, loss[loss=0.1797, simple_loss=0.2562, pruned_loss=0.0516, over 17038.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2779, pruned_loss=0.05602, over 3326225.01 frames. ], batch size: 41, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:25,966 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 2900, loss[loss=0.1633, simple_loss=0.2446, pruned_loss=0.041, over 17004.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2772, pruned_loss=0.0573, over 3321724.32 frames. ], batch size: 41, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:47,759 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 00:38:52,445 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:54,442 INFO [optim.py:368] (3/8) 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 00:39:43,178 INFO [train.py:904] (3/8) Epoch 9, batch 2950, loss[loss=0.2863, simple_loss=0.3442, pruned_loss=0.1142, over 11966.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2777, pruned_loss=0.05859, over 3300069.61 frames. ], batch size: 248, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:39:51,094 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:39:54,416 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-29 00:40:42,193 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7857, 3.1160, 2.8120, 4.7572, 3.9261, 4.4916, 1.4611, 3.3195], device='cuda:3'), covar=tensor([0.1266, 0.0590, 0.0975, 0.0176, 0.0269, 0.0347, 0.1467, 0.0666], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0153, 0.0175, 0.0129, 0.0200, 0.0210, 0.0172, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 00:40:47,973 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:48,091 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2962, 3.8106, 3.9016, 2.0655, 2.8819, 2.3462, 3.7300, 3.9644], device='cuda:3'), covar=tensor([0.0281, 0.0625, 0.0414, 0.1600, 0.0757, 0.0912, 0.0593, 0.0858], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0142, 0.0156, 0.0140, 0.0133, 0.0123, 0.0134, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 00:40:52,809 INFO [train.py:904] (3/8) Epoch 9, batch 3000, loss[loss=0.1738, simple_loss=0.2635, pruned_loss=0.04207, over 17232.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2781, pruned_loss=0.05874, over 3295060.61 frames. ], batch size: 45, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:40:52,809 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 00:41:02,063 INFO [train.py:938] (3/8) Epoch 9, validation: loss=0.1444, simple_loss=0.2507, pruned_loss=0.019, over 944034.00 frames. 2023-04-29 00:41:02,063 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 00:41:11,757 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7703, 3.8854, 2.0054, 4.3445, 2.7644, 4.3763, 2.2169, 3.0345], device='cuda:3'), covar=tensor([0.0185, 0.0306, 0.1582, 0.0180, 0.0748, 0.0353, 0.1378, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0164, 0.0184, 0.0121, 0.0164, 0.0206, 0.0191, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 00:41:23,147 INFO [optim.py:368] (3/8) 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 00:41:29,378 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3992, 1.9871, 2.2964, 4.1319, 2.0960, 2.5741, 2.1128, 2.2093], device='cuda:3'), covar=tensor([0.0937, 0.3124, 0.1838, 0.0376, 0.3072, 0.1820, 0.2904, 0.2797], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0380, 0.0318, 0.0328, 0.0403, 0.0430, 0.0339, 0.0449], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:41:47,893 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:09,946 INFO [train.py:904] (3/8) Epoch 9, batch 3050, loss[loss=0.1918, simple_loss=0.2677, pruned_loss=0.0579, over 16894.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2775, pruned_loss=0.05827, over 3296869.29 frames. ], batch size: 109, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:42:20,918 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:43:11,823 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:43:17,163 INFO [train.py:904] (3/8) Epoch 9, batch 3100, loss[loss=0.2194, simple_loss=0.3139, pruned_loss=0.0625, over 17054.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2767, pruned_loss=0.05765, over 3302821.37 frames. ], batch size: 55, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:43:22,112 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:43:57,190 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5720, 3.5935, 2.8187, 2.1483, 2.4544, 2.1074, 3.5804, 3.3728], device='cuda:3'), covar=tensor([0.2323, 0.0612, 0.1386, 0.2199, 0.2297, 0.1771, 0.0455, 0.1094], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0257, 0.0279, 0.0269, 0.0282, 0.0216, 0.0263, 0.0292], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:44:06,110 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:12,239 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:28,373 INFO [train.py:904] (3/8) Epoch 9, batch 3150, loss[loss=0.2305, simple_loss=0.2912, pruned_loss=0.08489, over 16798.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2762, pruned_loss=0.05784, over 3307015.97 frames. ], batch size: 124, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:44:40,117 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5895, 2.5447, 2.0237, 2.2658, 2.9215, 2.7176, 3.4495, 3.1597], device='cuda:3'), covar=tensor([0.0064, 0.0274, 0.0350, 0.0347, 0.0187, 0.0255, 0.0178, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0193, 0.0187, 0.0190, 0.0190, 0.0193, 0.0200, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:44:47,608 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 00:45:30,300 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:36,485 INFO [train.py:904] (3/8) Epoch 9, batch 3200, loss[loss=0.178, simple_loss=0.2749, pruned_loss=0.04056, over 17270.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2739, pruned_loss=0.05627, over 3312926.31 frames. ], batch size: 52, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:56,355 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:59,092 INFO [optim.py:368] (3/8) 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 00:46:45,525 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 00:46:45,891 INFO [train.py:904] (3/8) Epoch 9, batch 3250, loss[loss=0.2455, simple_loss=0.3191, pruned_loss=0.08598, over 12029.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2748, pruned_loss=0.05648, over 3311943.34 frames. ], batch size: 247, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:46:47,424 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:03,266 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:32,642 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 00:47:55,749 INFO [train.py:904] (3/8) Epoch 9, batch 3300, loss[loss=0.1887, simple_loss=0.281, pruned_loss=0.04821, over 16806.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.275, pruned_loss=0.05625, over 3318470.00 frames. ], batch size: 57, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:48:18,090 INFO [optim.py:368] (3/8) 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 00:48:27,869 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:04,554 INFO [train.py:904] (3/8) Epoch 9, batch 3350, loss[loss=0.1886, simple_loss=0.2611, pruned_loss=0.058, over 16774.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2758, pruned_loss=0.05647, over 3323499.87 frames. ], batch size: 83, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:49:10,249 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:53,293 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:56,520 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 00:50:01,510 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:16,095 INFO [train.py:904] (3/8) Epoch 9, batch 3400, loss[loss=0.2142, simple_loss=0.283, pruned_loss=0.07269, over 16906.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2754, pruned_loss=0.05626, over 3324441.82 frames. ], batch size: 116, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:50:20,119 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:38,676 INFO [optim.py:368] (3/8) 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 00:50:59,423 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:00,483 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7463, 4.6938, 4.5936, 4.3885, 4.2283, 4.6900, 4.5240, 4.4430], device='cuda:3'), covar=tensor([0.0517, 0.0558, 0.0263, 0.0260, 0.0891, 0.0429, 0.0396, 0.0588], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0304, 0.0294, 0.0266, 0.0320, 0.0303, 0.0201, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 00:51:10,771 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:24,588 INFO [train.py:904] (3/8) Epoch 9, batch 3450, loss[loss=0.1747, simple_loss=0.2718, pruned_loss=0.03884, over 17050.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2736, pruned_loss=0.05536, over 3314847.14 frames. ], batch size: 50, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:51:26,678 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:05,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5046, 5.9257, 5.6380, 5.8173, 5.3005, 5.0761, 5.4544, 6.0203], device='cuda:3'), covar=tensor([0.1002, 0.0870, 0.0971, 0.0538, 0.0738, 0.0652, 0.0804, 0.0766], device='cuda:3'), in_proj_covar=tensor([0.0525, 0.0662, 0.0549, 0.0449, 0.0410, 0.0420, 0.0546, 0.0492], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 00:52:16,068 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:20,253 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:24,208 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:35,102 INFO [train.py:904] (3/8) Epoch 9, batch 3500, loss[loss=0.2294, simple_loss=0.2875, pruned_loss=0.08561, over 11942.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2729, pruned_loss=0.0548, over 3318267.03 frames. ], batch size: 247, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:52:35,805 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 00:52:56,975 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.287e+02 2.792e+02 3.603e+02 8.093e+02, threshold=5.583e+02, percent-clipped=2.0 2023-04-29 00:53:44,890 INFO [train.py:904] (3/8) Epoch 9, batch 3550, loss[loss=0.2046, simple_loss=0.2737, pruned_loss=0.06776, over 16623.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2718, pruned_loss=0.05439, over 3314912.70 frames. ], batch size: 134, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:53:46,549 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:53,078 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:54,907 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:55,845 INFO [train.py:904] (3/8) Epoch 9, batch 3600, loss[loss=0.1606, simple_loss=0.246, pruned_loss=0.03754, over 17216.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2722, pruned_loss=0.05467, over 3322972.10 frames. ], batch size: 46, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:55:08,869 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 00:55:17,924 INFO [optim.py:368] (3/8) 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 00:55:59,050 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 00:56:07,140 INFO [train.py:904] (3/8) Epoch 9, batch 3650, loss[loss=0.2018, simple_loss=0.2693, pruned_loss=0.0672, over 16808.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2697, pruned_loss=0.0546, over 3314553.15 frames. ], batch size: 124, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:56:12,772 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:21,165 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:50,505 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:58,077 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 00:57:07,396 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 3700, loss[loss=0.168, simple_loss=0.2432, pruned_loss=0.04636, over 16489.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2686, pruned_loss=0.05655, over 3274979.00 frames. ], batch size: 146, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:57:23,506 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:45,917 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 00:57:45,941 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 00:57:46,406 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.579e+02 2.947e+02 3.773e+02 6.988e+02, threshold=5.894e+02, percent-clipped=1.0 2023-04-29 00:58:19,041 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:34,969 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 3750, loss[loss=0.1798, simple_loss=0.2538, pruned_loss=0.05291, over 16471.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2693, pruned_loss=0.05796, over 3263250.70 frames. ], batch size: 75, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 00:59:21,842 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-29 00:59:28,296 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:31,311 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.86 vs. limit=5.0 2023-04-29 00:59:33,933 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 3800, loss[loss=0.2043, simple_loss=0.2849, pruned_loss=0.06187, over 16681.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.271, pruned_loss=0.05984, over 3266814.05 frames. ], batch size: 62, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:00:03,326 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:00:10,085 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 01:00:11,015 INFO [optim.py:368] (3/8) 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 01:00:36,786 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 01:00:43,934 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:01:00,353 INFO [train.py:904] (3/8) Epoch 9, batch 3850, loss[loss=0.1713, simple_loss=0.2393, pruned_loss=0.05161, over 16444.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2717, pruned_loss=0.06097, over 3253619.66 frames. ], batch size: 75, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:13,000 INFO [train.py:904] (3/8) Epoch 9, batch 3900, loss[loss=0.1976, simple_loss=0.2731, pruned_loss=0.06108, over 16868.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2714, pruned_loss=0.06128, over 3261505.75 frames. ], batch size: 116, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:21,710 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:32,602 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:03:18,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-29 01:03:20,615 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2526, 4.2387, 4.3457, 4.2973, 4.3025, 4.7974, 4.3948, 4.1324], device='cuda:3'), covar=tensor([0.1839, 0.1734, 0.1806, 0.1848, 0.2603, 0.1089, 0.1466, 0.2525], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0474, 0.0508, 0.0420, 0.0551, 0.0527, 0.0407, 0.0563], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:03:24,358 INFO [train.py:904] (3/8) Epoch 9, batch 3950, loss[loss=0.2303, simple_loss=0.308, pruned_loss=0.07625, over 15502.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2712, pruned_loss=0.06158, over 3272961.96 frames. ], batch size: 190, lr: 7.66e-03, grad_scale: 8.0 2023-04-29 01:03:32,139 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:03:50,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7803, 2.7873, 2.7974, 1.8797, 2.5758, 2.8021, 2.5949, 1.7262], device='cuda:3'), covar=tensor([0.0341, 0.0071, 0.0039, 0.0276, 0.0072, 0.0065, 0.0068, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0067, 0.0066, 0.0121, 0.0073, 0.0082, 0.0074, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:03:50,075 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:04:00,737 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:07,277 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 4000, loss[loss=0.1918, simple_loss=0.2635, pruned_loss=0.06003, over 16481.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2707, pruned_loss=0.06158, over 3272560.03 frames. ], batch size: 146, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:04:40,573 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3632, 5.3504, 5.0637, 4.5550, 5.2475, 1.8324, 4.9350, 4.9394], device='cuda:3'), covar=tensor([0.0045, 0.0031, 0.0101, 0.0249, 0.0051, 0.2195, 0.0082, 0.0124], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0112, 0.0161, 0.0154, 0.0131, 0.0173, 0.0147, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:05:00,769 INFO [optim.py:368] (3/8) 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 01:05:17,724 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:05:51,192 INFO [train.py:904] (3/8) Epoch 9, batch 4050, loss[loss=0.1721, simple_loss=0.257, pruned_loss=0.0436, over 16726.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2705, pruned_loss=0.06023, over 3282484.98 frames. ], batch size: 124, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:06:45,535 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:04,624 INFO [train.py:904] (3/8) Epoch 9, batch 4100, loss[loss=0.2152, simple_loss=0.3009, pruned_loss=0.06473, over 16762.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2711, pruned_loss=0.05896, over 3279489.01 frames. ], batch size: 124, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:07:12,414 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:28,042 INFO [optim.py:368] (3/8) 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 01:07:54,823 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-29 01:07:57,983 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:08:20,247 INFO [train.py:904] (3/8) Epoch 9, batch 4150, loss[loss=0.2727, simple_loss=0.3319, pruned_loss=0.1067, over 11603.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2788, pruned_loss=0.06218, over 3241176.72 frames. ], batch size: 247, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:09:37,254 INFO [train.py:904] (3/8) Epoch 9, batch 4200, loss[loss=0.2212, simple_loss=0.3065, pruned_loss=0.068, over 16916.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2865, pruned_loss=0.06436, over 3219309.51 frames. ], batch size: 109, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:02,492 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.639e+02 3.175e+02 3.973e+02 9.081e+02, threshold=6.349e+02, percent-clipped=14.0 2023-04-29 01:10:17,721 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:10:45,446 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3017, 3.8037, 3.7347, 2.5551, 3.3295, 3.6506, 3.4621, 2.2214], device='cuda:3'), covar=tensor([0.0349, 0.0032, 0.0044, 0.0246, 0.0052, 0.0081, 0.0051, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0065, 0.0066, 0.0120, 0.0072, 0.0082, 0.0072, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:10:52,278 INFO [train.py:904] (3/8) Epoch 9, batch 4250, loss[loss=0.1774, simple_loss=0.2747, pruned_loss=0.04009, over 16833.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.29, pruned_loss=0.06417, over 3198616.94 frames. ], batch size: 102, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:59,121 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:08,336 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:38,968 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:48,345 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:07,379 INFO [train.py:904] (3/8) Epoch 9, batch 4300, loss[loss=0.215, simple_loss=0.3069, pruned_loss=0.06157, over 16795.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2909, pruned_loss=0.06323, over 3203003.90 frames. ], batch size: 124, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:12:11,406 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:12:49,209 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:10,318 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:21,370 INFO [train.py:904] (3/8) Epoch 9, batch 4350, loss[loss=0.2217, simple_loss=0.3081, pruned_loss=0.06765, over 16495.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2944, pruned_loss=0.06436, over 3186691.85 frames. ], batch size: 75, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:13:24,226 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0439, 5.3872, 5.0525, 5.1254, 4.7827, 4.6142, 4.6792, 5.4299], device='cuda:3'), covar=tensor([0.0709, 0.0622, 0.0899, 0.0587, 0.0665, 0.0756, 0.0892, 0.0736], device='cuda:3'), in_proj_covar=tensor([0.0489, 0.0616, 0.0514, 0.0424, 0.0385, 0.0400, 0.0513, 0.0463], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:14:18,975 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:30,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5513, 4.6744, 4.6268, 2.7705, 4.0279, 4.4070, 3.8944, 2.3267], device='cuda:3'), covar=tensor([0.0354, 0.0012, 0.0021, 0.0288, 0.0049, 0.0052, 0.0042, 0.0335], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0064, 0.0064, 0.0119, 0.0071, 0.0080, 0.0071, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:14:34,738 INFO [train.py:904] (3/8) Epoch 9, batch 4400, loss[loss=0.2015, simple_loss=0.2828, pruned_loss=0.06008, over 16642.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2962, pruned_loss=0.06523, over 3187400.54 frames. ], batch size: 57, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:14:41,685 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:56,972 INFO [optim.py:368] (3/8) 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 01:15:06,904 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 4450, loss[loss=0.2191, simple_loss=0.3111, pruned_loss=0.06355, over 17254.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2985, pruned_loss=0.06568, over 3195989.38 frames. ], batch size: 52, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:15:50,414 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:16:33,356 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:16:44,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3010, 3.3765, 3.5668, 1.6594, 3.9336, 3.9176, 2.9119, 3.0006], device='cuda:3'), covar=tensor([0.0810, 0.0206, 0.0219, 0.1174, 0.0053, 0.0078, 0.0381, 0.0372], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0096, 0.0085, 0.0138, 0.0067, 0.0096, 0.0118, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 01:16:56,347 INFO [train.py:904] (3/8) Epoch 9, batch 4500, loss[loss=0.2066, simple_loss=0.2989, pruned_loss=0.05713, over 16798.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.299, pruned_loss=0.06614, over 3204449.36 frames. ], batch size: 102, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:17:20,316 INFO [optim.py:368] (3/8) 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 01:17:33,584 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9897, 2.7210, 2.0363, 2.5191, 3.3042, 2.8304, 3.6693, 3.4632], device='cuda:3'), covar=tensor([0.0027, 0.0265, 0.0394, 0.0301, 0.0123, 0.0246, 0.0113, 0.0151], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0188, 0.0187, 0.0185, 0.0185, 0.0190, 0.0190, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:17:45,737 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2220, 3.1755, 1.6781, 3.5323, 2.3695, 3.4889, 1.8987, 2.5401], device='cuda:3'), covar=tensor([0.0182, 0.0330, 0.1713, 0.0079, 0.0789, 0.0296, 0.1387, 0.0607], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0161, 0.0183, 0.0112, 0.0163, 0.0200, 0.0191, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 01:18:07,071 INFO [train.py:904] (3/8) Epoch 9, batch 4550, loss[loss=0.2605, simple_loss=0.3323, pruned_loss=0.09431, over 16696.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2994, pruned_loss=0.06677, over 3205294.91 frames. ], batch size: 57, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:18:23,709 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:34,938 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 9, batch 4600, loss[loss=0.1639, simple_loss=0.2601, pruned_loss=0.03384, over 16859.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2996, pruned_loss=0.06584, over 3217796.32 frames. ], batch size: 102, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:19:32,653 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:42,422 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:12,996 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:30,861 INFO [train.py:904] (3/8) Epoch 9, batch 4650, loss[loss=0.2137, simple_loss=0.2978, pruned_loss=0.06485, over 15251.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2987, pruned_loss=0.06551, over 3217526.67 frames. ], batch size: 190, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:21:04,775 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:11,048 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5798, 4.5983, 5.1312, 5.0889, 5.1458, 4.6427, 4.6931, 4.2676], device='cuda:3'), covar=tensor([0.0260, 0.0342, 0.0266, 0.0334, 0.0338, 0.0293, 0.0698, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0295, 0.0298, 0.0307, 0.0293, 0.0346, 0.0320, 0.0422, 0.0257], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 01:21:20,345 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:41,768 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-04-29 01:21:42,723 INFO [train.py:904] (3/8) Epoch 9, batch 4700, loss[loss=0.214, simple_loss=0.2959, pruned_loss=0.06606, over 15272.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2958, pruned_loss=0.06438, over 3225190.81 frames. ], batch size: 190, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:21:50,660 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5950, 2.6443, 2.3610, 4.1147, 2.8107, 4.0293, 1.3048, 2.7743], device='cuda:3'), covar=tensor([0.1362, 0.0709, 0.1213, 0.0135, 0.0296, 0.0349, 0.1645, 0.0859], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0153, 0.0175, 0.0126, 0.0200, 0.0206, 0.0173, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 01:21:54,808 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7085, 4.5073, 4.7376, 4.9598, 5.0486, 4.5668, 5.0497, 5.0592], device='cuda:3'), covar=tensor([0.1235, 0.0966, 0.1174, 0.0450, 0.0430, 0.0648, 0.0368, 0.0395], device='cuda:3'), in_proj_covar=tensor([0.0484, 0.0593, 0.0737, 0.0600, 0.0459, 0.0461, 0.0472, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:22:06,316 INFO [optim.py:368] (3/8) 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 01:22:25,116 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:22:32,864 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:22:55,504 INFO [train.py:904] (3/8) Epoch 9, batch 4750, loss[loss=0.1715, simple_loss=0.2512, pruned_loss=0.04587, over 16610.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2914, pruned_loss=0.0624, over 3230385.27 frames. ], batch size: 62, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:22:59,912 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8694, 5.2912, 5.4180, 5.2577, 5.2933, 5.7744, 5.3419, 5.0688], device='cuda:3'), covar=tensor([0.0781, 0.1342, 0.1359, 0.1601, 0.2223, 0.0842, 0.1010, 0.2211], device='cuda:3'), in_proj_covar=tensor([0.0331, 0.0447, 0.0473, 0.0395, 0.0523, 0.0506, 0.0385, 0.0532], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:23:04,380 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:23:39,896 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:23:54,625 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:24:13,771 INFO [train.py:904] (3/8) Epoch 9, batch 4800, loss[loss=0.2357, simple_loss=0.3001, pruned_loss=0.08567, over 11703.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2879, pruned_loss=0.06079, over 3216487.64 frames. ], batch size: 246, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:24:14,246 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:24:30,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7100, 4.7068, 5.1470, 5.1154, 5.1606, 4.7093, 4.7475, 4.4434], device='cuda:3'), covar=tensor([0.0228, 0.0383, 0.0279, 0.0317, 0.0358, 0.0270, 0.0861, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0303, 0.0314, 0.0300, 0.0353, 0.0326, 0.0432, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 01:24:37,221 INFO [optim.py:368] (3/8) 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 01:24:38,439 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:00,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9600, 4.0460, 3.8077, 3.6349, 3.4611, 3.9460, 3.5949, 3.6769], device='cuda:3'), covar=tensor([0.0529, 0.0335, 0.0271, 0.0246, 0.0801, 0.0357, 0.0986, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0222, 0.0271, 0.0264, 0.0239, 0.0288, 0.0271, 0.0181, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:25:28,106 INFO [train.py:904] (3/8) Epoch 9, batch 4850, loss[loss=0.2367, simple_loss=0.3062, pruned_loss=0.08361, over 12161.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2893, pruned_loss=0.06041, over 3201482.11 frames. ], batch size: 247, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:25:28,597 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:45,081 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:54,720 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:16,450 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:42,281 INFO [train.py:904] (3/8) Epoch 9, batch 4900, loss[loss=0.1836, simple_loss=0.2673, pruned_loss=0.04998, over 16593.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2896, pruned_loss=0.06003, over 3171678.51 frames. ], batch size: 62, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:26:59,805 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:27:23,779 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:26,529 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:37,257 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:55,672 INFO [train.py:904] (3/8) Epoch 9, batch 4950, loss[loss=0.1831, simple_loss=0.2757, pruned_loss=0.04525, over 16413.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2891, pruned_loss=0.0593, over 3168057.00 frames. ], batch size: 75, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:28:46,024 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:28:46,995 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:08,679 INFO [train.py:904] (3/8) Epoch 9, batch 5000, loss[loss=0.1938, simple_loss=0.2764, pruned_loss=0.05553, over 17242.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2905, pruned_loss=0.05909, over 3191822.74 frames. ], batch size: 52, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:29:27,959 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 2023-04-29 01:29:32,026 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.365e+02 2.931e+02 3.414e+02 6.733e+02, threshold=5.862e+02, percent-clipped=2.0 2023-04-29 01:29:50,433 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:54,568 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:11,843 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:19,717 INFO [train.py:904] (3/8) Epoch 9, batch 5050, loss[loss=0.1975, simple_loss=0.2893, pruned_loss=0.05284, over 16878.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2903, pruned_loss=0.05866, over 3201081.73 frames. ], batch size: 96, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:31:01,265 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:07,942 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:31:29,531 INFO [train.py:904] (3/8) Epoch 9, batch 5100, loss[loss=0.2052, simple_loss=0.2852, pruned_loss=0.06256, over 16198.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2884, pruned_loss=0.05796, over 3206055.60 frames. ], batch size: 165, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:31:37,436 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:45,612 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:32:08,053 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:32:41,540 INFO [train.py:904] (3/8) Epoch 9, batch 5150, loss[loss=0.2215, simple_loss=0.3115, pruned_loss=0.06579, over 15331.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2887, pruned_loss=0.05759, over 3192636.20 frames. ], batch size: 191, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:32:50,286 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:33:11,100 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9729, 3.9099, 3.8903, 3.2151, 3.8718, 1.6614, 3.6401, 3.5255], device='cuda:3'), covar=tensor([0.0093, 0.0092, 0.0116, 0.0399, 0.0086, 0.2309, 0.0137, 0.0209], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0106, 0.0153, 0.0146, 0.0122, 0.0168, 0.0137, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:33:54,430 INFO [train.py:904] (3/8) Epoch 9, batch 5200, loss[loss=0.198, simple_loss=0.2943, pruned_loss=0.05082, over 16835.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2873, pruned_loss=0.05731, over 3189637.03 frames. ], batch size: 90, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:34:02,828 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:04,501 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-04-29 01:34:16,849 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-04-29 01:34:17,300 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 2.298e+02 2.670e+02 3.115e+02 5.719e+02, threshold=5.340e+02, percent-clipped=1.0 2023-04-29 01:34:27,688 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:50,091 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:57,844 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:35:06,328 INFO [train.py:904] (3/8) Epoch 9, batch 5250, loss[loss=0.2269, simple_loss=0.2986, pruned_loss=0.07754, over 12487.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2846, pruned_loss=0.05693, over 3187925.02 frames. ], batch size: 248, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:35:41,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6161, 3.7186, 2.9997, 2.2252, 2.5751, 2.3314, 3.9223, 3.6025], device='cuda:3'), covar=tensor([0.2437, 0.0690, 0.1281, 0.2036, 0.2015, 0.1549, 0.0435, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0252, 0.0274, 0.0268, 0.0274, 0.0212, 0.0261, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:36:16,603 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 01:36:17,589 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 5300, loss[loss=0.186, simple_loss=0.2647, pruned_loss=0.0537, over 16901.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2805, pruned_loss=0.05517, over 3203924.54 frames. ], batch size: 109, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:36:25,581 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.448e+02 2.767e+02 3.292e+02 5.111e+02, threshold=5.534e+02, percent-clipped=0.0 2023-04-29 01:36:47,607 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3498, 3.6191, 3.8992, 1.7300, 4.1517, 4.1363, 3.0156, 2.9410], device='cuda:3'), covar=tensor([0.0834, 0.0158, 0.0143, 0.1212, 0.0041, 0.0071, 0.0367, 0.0459], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0096, 0.0084, 0.0137, 0.0067, 0.0094, 0.0118, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-29 01:36:58,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0412, 2.8505, 2.7079, 2.0132, 2.6150, 2.1446, 2.8641, 2.8923], device='cuda:3'), covar=tensor([0.0234, 0.0555, 0.0508, 0.1468, 0.0652, 0.0834, 0.0518, 0.0601], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0138, 0.0156, 0.0140, 0.0134, 0.0124, 0.0135, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 01:37:01,876 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:37:32,912 INFO [train.py:904] (3/8) Epoch 9, batch 5350, loss[loss=0.2139, simple_loss=0.3023, pruned_loss=0.06273, over 16938.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2792, pruned_loss=0.0544, over 3207823.36 frames. ], batch size: 109, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:11,780 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:13,779 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:23,813 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:38:45,867 INFO [train.py:904] (3/8) Epoch 9, batch 5400, loss[loss=0.2288, simple_loss=0.3014, pruned_loss=0.0781, over 11826.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2819, pruned_loss=0.05511, over 3217663.54 frames. ], batch size: 246, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:46,249 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:01,689 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:09,034 INFO [optim.py:368] (3/8) 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 01:39:12,837 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7289, 4.4609, 4.4831, 4.9753, 5.0666, 4.6573, 5.1385, 5.0966], device='cuda:3'), covar=tensor([0.1497, 0.1132, 0.2062, 0.0730, 0.0704, 0.0730, 0.0508, 0.0649], device='cuda:3'), in_proj_covar=tensor([0.0499, 0.0606, 0.0752, 0.0614, 0.0465, 0.0466, 0.0483, 0.0536], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:39:31,533 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:39:41,805 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:53,928 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 01:40:00,484 INFO [train.py:904] (3/8) Epoch 9, batch 5450, loss[loss=0.23, simple_loss=0.314, pruned_loss=0.07302, over 16763.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.285, pruned_loss=0.05659, over 3209653.48 frames. ], batch size: 89, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:40:03,990 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 01:40:11,005 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:15,341 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 5500, loss[loss=0.2186, simple_loss=0.3031, pruned_loss=0.06707, over 16814.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2935, pruned_loss=0.06253, over 3181244.08 frames. ], batch size: 102, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:41:24,318 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:27,618 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:43,592 INFO [optim.py:368] (3/8) 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 01:41:54,520 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:42:36,877 INFO [train.py:904] (3/8) Epoch 9, batch 5550, loss[loss=0.2175, simple_loss=0.3034, pruned_loss=0.06586, over 17233.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3022, pruned_loss=0.06869, over 3168267.53 frames. ], batch size: 52, lr: 7.59e-03, grad_scale: 16.0 2023-04-29 01:42:43,065 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:09,698 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:44,627 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:53,943 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:43:54,665 INFO [train.py:904] (3/8) Epoch 9, batch 5600, loss[loss=0.2579, simple_loss=0.3219, pruned_loss=0.09692, over 16518.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3084, pruned_loss=0.074, over 3128627.72 frames. ], batch size: 68, lr: 7.59e-03, grad_scale: 8.0 2023-04-29 01:44:14,659 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.990e+02 4.962e+02 6.279e+02 1.585e+03, threshold=9.923e+02, percent-clipped=6.0 2023-04-29 01:44:22,450 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 01:45:17,290 INFO [train.py:904] (3/8) Epoch 9, batch 5650, loss[loss=0.1968, simple_loss=0.2848, pruned_loss=0.05443, over 17039.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3135, pruned_loss=0.07864, over 3093425.78 frames. ], batch size: 55, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:45:30,238 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2984, 4.1067, 4.0707, 2.7799, 3.6631, 4.0446, 3.7069, 2.2498], device='cuda:3'), covar=tensor([0.0397, 0.0023, 0.0032, 0.0271, 0.0060, 0.0080, 0.0039, 0.0324], device='cuda:3'), in_proj_covar=tensor([0.0121, 0.0064, 0.0064, 0.0120, 0.0071, 0.0081, 0.0070, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:45:31,711 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-04-29 01:45:53,438 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:45:58,557 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:11,320 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 01:46:33,118 INFO [train.py:904] (3/8) Epoch 9, batch 5700, loss[loss=0.2165, simple_loss=0.303, pruned_loss=0.06499, over 16658.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3149, pruned_loss=0.07988, over 3095128.93 frames. ], batch size: 57, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:46:33,518 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:59,880 INFO [optim.py:368] (3/8) 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 01:47:05,572 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6924, 3.3562, 3.1167, 1.8335, 2.6578, 2.2250, 3.1530, 3.3444], device='cuda:3'), covar=tensor([0.0310, 0.0592, 0.0541, 0.1757, 0.0833, 0.0920, 0.0653, 0.0839], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0139, 0.0157, 0.0141, 0.0134, 0.0124, 0.0135, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 01:47:25,187 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:31,002 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:48,289 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 5750, loss[loss=0.2057, simple_loss=0.3017, pruned_loss=0.05486, over 16845.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3174, pruned_loss=0.08161, over 3051016.58 frames. ], batch size: 102, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:48:31,406 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 01:49:12,515 INFO [train.py:904] (3/8) Epoch 9, batch 5800, loss[loss=0.2063, simple_loss=0.2973, pruned_loss=0.0577, over 16328.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3167, pruned_loss=0.07981, over 3060886.34 frames. ], batch size: 146, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:49:13,831 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:49:40,720 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 3.181e+02 3.841e+02 4.855e+02 7.782e+02, threshold=7.681e+02, percent-clipped=0.0 2023-04-29 01:50:30,438 INFO [train.py:904] (3/8) Epoch 9, batch 5850, loss[loss=0.2169, simple_loss=0.3028, pruned_loss=0.06551, over 16231.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3145, pruned_loss=0.07799, over 3072681.59 frames. ], batch size: 165, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:50:47,456 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:51:40,412 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-29 01:51:42,080 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:51:50,038 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:51:50,871 INFO [train.py:904] (3/8) Epoch 9, batch 5900, loss[loss=0.2303, simple_loss=0.3236, pruned_loss=0.0685, over 17025.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3143, pruned_loss=0.07826, over 3077390.37 frames. ], batch size: 41, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:52:22,906 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 3.002e+02 3.675e+02 4.500e+02 8.851e+02, threshold=7.350e+02, percent-clipped=1.0 2023-04-29 01:52:27,300 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 01:52:40,730 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 01:53:01,035 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:09,061 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:13,003 INFO [train.py:904] (3/8) Epoch 9, batch 5950, loss[loss=0.2141, simple_loss=0.3044, pruned_loss=0.06187, over 17046.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3151, pruned_loss=0.07684, over 3083963.70 frames. ], batch size: 50, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:53:18,213 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9883, 3.2442, 3.3594, 2.1159, 3.0764, 3.3220, 3.1524, 1.8036], device='cuda:3'), covar=tensor([0.0377, 0.0042, 0.0032, 0.0308, 0.0066, 0.0073, 0.0054, 0.0345], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0065, 0.0066, 0.0123, 0.0073, 0.0083, 0.0072, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 01:53:42,077 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:54:33,094 INFO [train.py:904] (3/8) Epoch 9, batch 6000, loss[loss=0.2238, simple_loss=0.3021, pruned_loss=0.07272, over 16520.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3133, pruned_loss=0.07571, over 3104656.13 frames. ], batch size: 75, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:54:33,095 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 01:54:44,303 INFO [train.py:938] (3/8) Epoch 9, validation: loss=0.1674, simple_loss=0.2809, pruned_loss=0.02692, over 944034.00 frames. 2023-04-29 01:54:44,303 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 01:55:11,802 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.247e+02 3.795e+02 4.870e+02 1.523e+03, threshold=7.589e+02, percent-clipped=3.0 2023-04-29 01:55:28,617 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:37,530 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:56:03,544 INFO [train.py:904] (3/8) Epoch 9, batch 6050, loss[loss=0.2947, simple_loss=0.3456, pruned_loss=0.1219, over 11761.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3118, pruned_loss=0.07475, over 3115284.81 frames. ], batch size: 246, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:56:08,367 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-29 01:56:33,590 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4467, 4.5360, 4.9651, 4.9274, 4.8953, 4.5638, 4.5327, 4.3009], device='cuda:3'), covar=tensor([0.0305, 0.0526, 0.0333, 0.0342, 0.0393, 0.0299, 0.0919, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0309, 0.0321, 0.0305, 0.0358, 0.0331, 0.0442, 0.0270], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 01:56:52,088 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:56:59,503 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0747, 3.8874, 4.0925, 4.2664, 4.3491, 3.9111, 4.3037, 4.3563], device='cuda:3'), covar=tensor([0.1222, 0.0987, 0.1086, 0.0521, 0.0486, 0.1245, 0.0615, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0483, 0.0596, 0.0727, 0.0605, 0.0461, 0.0456, 0.0483, 0.0532], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 01:57:06,987 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:23,337 INFO [train.py:904] (3/8) Epoch 9, batch 6100, loss[loss=0.234, simple_loss=0.3198, pruned_loss=0.07412, over 16741.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3113, pruned_loss=0.07356, over 3133350.56 frames. ], batch size: 124, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:57:52,499 INFO [optim.py:368] (3/8) 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 01:58:42,519 INFO [train.py:904] (3/8) Epoch 9, batch 6150, loss[loss=0.2334, simple_loss=0.3214, pruned_loss=0.07273, over 16485.00 frames. ], tot_loss[loss=0.227, simple_loss=0.309, pruned_loss=0.0725, over 3125479.57 frames. ], batch size: 75, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:58:52,851 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:58:59,257 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:59:16,488 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5519, 2.5714, 1.6199, 2.7121, 2.1033, 2.7452, 1.8711, 2.2616], device='cuda:3'), covar=tensor([0.0205, 0.0298, 0.1239, 0.0147, 0.0588, 0.0384, 0.1143, 0.0497], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0161, 0.0186, 0.0113, 0.0165, 0.0201, 0.0191, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 01:59:51,461 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 02:00:00,627 INFO [train.py:904] (3/8) Epoch 9, batch 6200, loss[loss=0.2082, simple_loss=0.2981, pruned_loss=0.05916, over 16782.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3066, pruned_loss=0.07182, over 3140653.38 frames. ], batch size: 83, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 02:00:19,118 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 02:00:28,302 INFO [optim.py:368] (3/8) 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 02:00:34,940 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:05,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3939, 3.3095, 3.3778, 3.5075, 3.5297, 3.2346, 3.4756, 3.5732], device='cuda:3'), covar=tensor([0.0960, 0.0875, 0.1005, 0.0512, 0.0591, 0.2029, 0.0898, 0.0654], device='cuda:3'), in_proj_covar=tensor([0.0487, 0.0604, 0.0738, 0.0607, 0.0466, 0.0460, 0.0487, 0.0538], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:01:18,822 INFO [train.py:904] (3/8) Epoch 9, batch 6250, loss[loss=0.2191, simple_loss=0.3055, pruned_loss=0.06629, over 16473.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3065, pruned_loss=0.07164, over 3141048.63 frames. ], batch size: 146, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:01:32,679 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:45,536 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:33,519 INFO [train.py:904] (3/8) Epoch 9, batch 6300, loss[loss=0.2366, simple_loss=0.3147, pruned_loss=0.07923, over 16410.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3059, pruned_loss=0.07125, over 3126653.75 frames. ], batch size: 146, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:02:38,991 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:59,938 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:03:06,791 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:24,796 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 6350, loss[loss=0.2214, simple_loss=0.3036, pruned_loss=0.06958, over 16175.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3076, pruned_loss=0.07319, over 3101889.13 frames. ], batch size: 165, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:04:11,664 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3718, 1.4660, 1.9716, 2.3593, 2.4590, 2.5706, 1.6746, 2.5023], device='cuda:3'), covar=tensor([0.0120, 0.0348, 0.0201, 0.0176, 0.0177, 0.0112, 0.0308, 0.0082], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0162, 0.0145, 0.0148, 0.0156, 0.0112, 0.0163, 0.0104], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 02:04:13,023 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:04:37,474 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:40,660 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:44,420 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:49,442 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:08,849 INFO [train.py:904] (3/8) Epoch 9, batch 6400, loss[loss=0.2193, simple_loss=0.2861, pruned_loss=0.07621, over 17030.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3074, pruned_loss=0.07393, over 3106131.25 frames. ], batch size: 55, lr: 7.56e-03, grad_scale: 8.0 2023-04-29 02:05:13,522 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:37,572 INFO [optim.py:368] (3/8) 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 02:06:14,965 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:06:22,248 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 6450, loss[loss=0.2001, simple_loss=0.2861, pruned_loss=0.05703, over 16826.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3075, pruned_loss=0.07326, over 3100544.65 frames. ], batch size: 116, lr: 7.55e-03, grad_scale: 4.0 2023-04-29 02:06:34,324 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:06:46,438 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:07:05,534 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1699, 2.3598, 1.8535, 2.1248, 2.7492, 2.4151, 3.0910, 3.0192], device='cuda:3'), covar=tensor([0.0069, 0.0279, 0.0422, 0.0324, 0.0161, 0.0280, 0.0149, 0.0151], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0186, 0.0186, 0.0183, 0.0183, 0.0187, 0.0187, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:07:11,437 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5226, 4.5799, 4.9789, 4.9702, 4.9454, 4.6086, 4.6342, 4.3087], device='cuda:3'), covar=tensor([0.0261, 0.0384, 0.0303, 0.0334, 0.0440, 0.0311, 0.0833, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0305, 0.0316, 0.0303, 0.0356, 0.0326, 0.0435, 0.0266], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 02:07:41,885 INFO [train.py:904] (3/8) Epoch 9, batch 6500, loss[loss=0.2145, simple_loss=0.3007, pruned_loss=0.06412, over 16851.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3056, pruned_loss=0.07229, over 3111603.00 frames. ], batch size: 96, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:07:48,425 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:08:06,443 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:08:13,252 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 3.236e+02 4.021e+02 5.179e+02 1.078e+03, threshold=8.043e+02, percent-clipped=2.0 2023-04-29 02:08:24,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0585, 2.6415, 2.6621, 1.8633, 2.8306, 2.8383, 2.4112, 2.4061], device='cuda:3'), covar=tensor([0.0659, 0.0186, 0.0197, 0.0897, 0.0077, 0.0184, 0.0416, 0.0404], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0100, 0.0085, 0.0140, 0.0066, 0.0097, 0.0120, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 02:09:00,847 INFO [train.py:904] (3/8) Epoch 9, batch 6550, loss[loss=0.2195, simple_loss=0.3258, pruned_loss=0.05657, over 16724.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3082, pruned_loss=0.07368, over 3086577.96 frames. ], batch size: 89, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:09:19,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7713, 1.2708, 1.6401, 1.6577, 1.8468, 1.8466, 1.4956, 1.8760], device='cuda:3'), covar=tensor([0.0129, 0.0249, 0.0135, 0.0171, 0.0164, 0.0108, 0.0259, 0.0071], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0159, 0.0142, 0.0146, 0.0153, 0.0110, 0.0161, 0.0102], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 02:10:18,518 INFO [train.py:904] (3/8) Epoch 9, batch 6600, loss[loss=0.2606, simple_loss=0.3266, pruned_loss=0.09731, over 11393.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3107, pruned_loss=0.07464, over 3066046.67 frames. ], batch size: 246, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:10:42,558 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:11:36,975 INFO [train.py:904] (3/8) Epoch 9, batch 6650, loss[loss=0.267, simple_loss=0.3284, pruned_loss=0.1029, over 11515.00 frames. ], tot_loss[loss=0.2322, simple_loss=0.3115, pruned_loss=0.07649, over 3039224.16 frames. ], batch size: 247, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:11:50,957 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:12:06,421 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7270, 1.6527, 2.2776, 2.8107, 2.6378, 3.0581, 1.7293, 3.0351], device='cuda:3'), covar=tensor([0.0124, 0.0348, 0.0191, 0.0167, 0.0176, 0.0112, 0.0363, 0.0075], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0161, 0.0143, 0.0146, 0.0154, 0.0111, 0.0163, 0.0103], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 02:12:28,605 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 6700, loss[loss=0.2671, simple_loss=0.3231, pruned_loss=0.1055, over 11565.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3099, pruned_loss=0.07622, over 3050263.89 frames. ], batch size: 246, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:13:26,701 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.383e+02 4.194e+02 5.361e+02 9.838e+02, threshold=8.388e+02, percent-clipped=4.0 2023-04-29 02:13:44,327 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:13:52,490 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:14:00,202 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:14:10,415 INFO [train.py:904] (3/8) Epoch 9, batch 6750, loss[loss=0.2144, simple_loss=0.2934, pruned_loss=0.06764, over 16730.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3089, pruned_loss=0.07562, over 3074927.49 frames. ], batch size: 124, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:14:24,856 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:15:29,294 INFO [train.py:904] (3/8) Epoch 9, batch 6800, loss[loss=0.2219, simple_loss=0.3118, pruned_loss=0.06601, over 16382.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3082, pruned_loss=0.07514, over 3078431.09 frames. ], batch size: 165, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:15:45,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1371, 4.1973, 4.3302, 4.2152, 4.1922, 4.7438, 4.3526, 4.0525], device='cuda:3'), covar=tensor([0.1705, 0.3519, 0.3593, 0.2404, 0.4353, 0.1486, 0.2314, 0.3612], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0448, 0.0479, 0.0395, 0.0522, 0.0508, 0.0388, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 02:15:54,167 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:16:02,234 INFO [optim.py:368] (3/8) 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 02:16:20,984 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5711, 3.5058, 2.7834, 2.2235, 2.5735, 2.3144, 3.6966, 3.3998], device='cuda:3'), covar=tensor([0.2616, 0.0804, 0.1657, 0.2081, 0.2104, 0.1634, 0.0503, 0.0919], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0253, 0.0279, 0.0269, 0.0281, 0.0213, 0.0262, 0.0279], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:16:30,693 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 02:16:45,306 INFO [train.py:904] (3/8) Epoch 9, batch 6850, loss[loss=0.2153, simple_loss=0.3095, pruned_loss=0.06051, over 16715.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3103, pruned_loss=0.07559, over 3088329.63 frames. ], batch size: 134, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:17:06,649 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:17:19,389 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 02:17:59,091 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 6900, loss[loss=0.1979, simple_loss=0.2909, pruned_loss=0.05245, over 16759.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3122, pruned_loss=0.07449, over 3109244.31 frames. ], batch size: 83, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:18:24,579 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:33,100 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 3.211e+02 3.844e+02 4.463e+02 9.504e+02, threshold=7.687e+02, percent-clipped=4.0 2023-04-29 02:19:03,976 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 02:19:18,074 INFO [train.py:904] (3/8) Epoch 9, batch 6950, loss[loss=0.2081, simple_loss=0.3, pruned_loss=0.05815, over 16834.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.315, pruned_loss=0.07741, over 3075760.42 frames. ], batch size: 96, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:19:31,038 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1843, 4.0344, 4.5014, 2.2140, 4.8145, 4.7288, 3.2530, 3.6470], device='cuda:3'), covar=tensor([0.0618, 0.0159, 0.0120, 0.1026, 0.0034, 0.0075, 0.0323, 0.0352], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0098, 0.0085, 0.0139, 0.0066, 0.0096, 0.0119, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 02:19:32,202 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:19:32,230 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:19:38,179 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:20:31,301 INFO [train.py:904] (3/8) Epoch 9, batch 7000, loss[loss=0.2194, simple_loss=0.3157, pruned_loss=0.06153, over 16351.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3149, pruned_loss=0.07673, over 3076553.54 frames. ], batch size: 146, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:20:43,255 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:20:54,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6673, 3.1977, 2.4281, 4.7825, 3.7175, 4.2240, 1.7197, 2.9030], device='cuda:3'), covar=tensor([0.1311, 0.0551, 0.1184, 0.0116, 0.0288, 0.0431, 0.1308, 0.0891], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0154, 0.0178, 0.0128, 0.0202, 0.0207, 0.0175, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 02:21:03,406 INFO [optim.py:368] (3/8) 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 02:21:18,568 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8922, 3.5419, 3.1602, 1.7560, 2.6389, 2.4050, 3.2286, 3.5231], device='cuda:3'), covar=tensor([0.0369, 0.0641, 0.0682, 0.1954, 0.0963, 0.0878, 0.0935, 0.0927], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0137, 0.0156, 0.0141, 0.0133, 0.0124, 0.0135, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 02:21:29,012 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:36,015 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 7050, loss[loss=0.2478, simple_loss=0.3139, pruned_loss=0.0909, over 11332.00 frames. ], tot_loss[loss=0.2339, simple_loss=0.3151, pruned_loss=0.0764, over 3062103.31 frames. ], batch size: 246, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:22:00,431 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:22:41,150 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:22:45,496 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:22:47,885 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 7100, loss[loss=0.2523, simple_loss=0.3134, pruned_loss=0.09563, over 11360.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3137, pruned_loss=0.07632, over 3050997.23 frames. ], batch size: 248, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:23:09,529 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 02:23:12,558 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:33,098 INFO [optim.py:368] (3/8) 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 02:24:04,331 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4257, 4.4170, 4.3243, 3.1448, 4.3491, 1.4380, 4.0825, 4.0709], device='cuda:3'), covar=tensor([0.0149, 0.0121, 0.0201, 0.0673, 0.0125, 0.3223, 0.0183, 0.0356], device='cuda:3'), in_proj_covar=tensor([0.0116, 0.0103, 0.0150, 0.0144, 0.0119, 0.0166, 0.0134, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:24:15,977 INFO [train.py:904] (3/8) Epoch 9, batch 7150, loss[loss=0.2332, simple_loss=0.3147, pruned_loss=0.07582, over 16884.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3109, pruned_loss=0.07513, over 3062977.88 frames. ], batch size: 90, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:24:17,087 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:24:24,685 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 02:25:00,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8403, 2.7131, 2.5778, 1.8559, 2.5844, 2.6367, 2.5194, 1.7351], device='cuda:3'), covar=tensor([0.0336, 0.0051, 0.0064, 0.0287, 0.0076, 0.0089, 0.0078, 0.0368], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0064, 0.0065, 0.0124, 0.0072, 0.0084, 0.0073, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 02:25:03,278 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9069, 2.9018, 2.6660, 4.4234, 3.3740, 4.3106, 1.5365, 3.1789], device='cuda:3'), covar=tensor([0.1238, 0.0620, 0.1024, 0.0119, 0.0289, 0.0321, 0.1492, 0.0713], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0154, 0.0176, 0.0127, 0.0200, 0.0205, 0.0175, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 02:25:21,134 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 02:25:28,100 INFO [train.py:904] (3/8) Epoch 9, batch 7200, loss[loss=0.1983, simple_loss=0.2851, pruned_loss=0.05581, over 16673.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3087, pruned_loss=0.07315, over 3073039.12 frames. ], batch size: 57, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:00,150 INFO [optim.py:368] (3/8) 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 02:26:47,216 INFO [train.py:904] (3/8) Epoch 9, batch 7250, loss[loss=0.1947, simple_loss=0.2745, pruned_loss=0.05745, over 16614.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3062, pruned_loss=0.07164, over 3070971.04 frames. ], batch size: 62, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:53,243 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:28:00,998 INFO [train.py:904] (3/8) Epoch 9, batch 7300, loss[loss=0.238, simple_loss=0.315, pruned_loss=0.08049, over 16152.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3063, pruned_loss=0.07244, over 3061283.66 frames. ], batch size: 165, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:28:33,453 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.255e+02 4.092e+02 5.788e+02 1.345e+03, threshold=8.184e+02, percent-clipped=12.0 2023-04-29 02:29:14,111 INFO [train.py:904] (3/8) Epoch 9, batch 7350, loss[loss=0.1843, simple_loss=0.2783, pruned_loss=0.04511, over 16832.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3065, pruned_loss=0.07267, over 3053559.83 frames. ], batch size: 102, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:29:29,565 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5824, 4.9282, 5.0656, 4.9606, 4.9521, 5.4934, 4.9942, 4.7847], device='cuda:3'), covar=tensor([0.1058, 0.1669, 0.1411, 0.1726, 0.2105, 0.0911, 0.1446, 0.2468], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0452, 0.0484, 0.0400, 0.0528, 0.0513, 0.0391, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 02:29:48,170 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6217, 4.4564, 4.6624, 4.8609, 4.9766, 4.4627, 4.9312, 4.9323], device='cuda:3'), covar=tensor([0.1379, 0.0921, 0.1184, 0.0503, 0.0468, 0.0798, 0.0471, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0471, 0.0584, 0.0721, 0.0586, 0.0454, 0.0452, 0.0473, 0.0522], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:30:18,070 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6191, 3.7310, 3.1228, 2.1977, 2.7870, 2.3050, 4.0425, 3.6109], device='cuda:3'), covar=tensor([0.2550, 0.0703, 0.1390, 0.2203, 0.2066, 0.1757, 0.0420, 0.0905], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0257, 0.0283, 0.0271, 0.0282, 0.0216, 0.0264, 0.0285], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:30:27,926 INFO [train.py:904] (3/8) Epoch 9, batch 7400, loss[loss=0.2709, simple_loss=0.3341, pruned_loss=0.1039, over 11727.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3075, pruned_loss=0.07287, over 3070758.91 frames. ], batch size: 248, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:31:01,751 INFO [optim.py:368] (3/8) 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 02:31:04,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6300, 4.5933, 5.1525, 5.1207, 5.1299, 4.7781, 4.7436, 4.3588], device='cuda:3'), covar=tensor([0.0314, 0.0487, 0.0387, 0.0437, 0.0441, 0.0323, 0.0860, 0.0479], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0304, 0.0316, 0.0298, 0.0350, 0.0326, 0.0431, 0.0266], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 02:31:38,445 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:31:44,621 INFO [train.py:904] (3/8) Epoch 9, batch 7450, loss[loss=0.2123, simple_loss=0.3041, pruned_loss=0.06026, over 16766.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3085, pruned_loss=0.07389, over 3074525.66 frames. ], batch size: 102, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:03,515 INFO [train.py:904] (3/8) Epoch 9, batch 7500, loss[loss=0.2386, simple_loss=0.3154, pruned_loss=0.08091, over 15359.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3086, pruned_loss=0.0733, over 3085480.67 frames. ], batch size: 190, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:36,929 INFO [optim.py:368] (3/8) 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 02:34:09,636 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1703, 1.9309, 1.6418, 1.7151, 2.2252, 1.9834, 2.1917, 2.3982], device='cuda:3'), covar=tensor([0.0081, 0.0236, 0.0307, 0.0283, 0.0139, 0.0229, 0.0122, 0.0139], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0187, 0.0185, 0.0184, 0.0185, 0.0187, 0.0186, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:34:18,158 INFO [train.py:904] (3/8) Epoch 9, batch 7550, loss[loss=0.2887, simple_loss=0.3366, pruned_loss=0.1204, over 11690.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.308, pruned_loss=0.07388, over 3072204.98 frames. ], batch size: 246, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:34:23,600 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:34:45,570 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 02:35:25,650 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4352, 3.0918, 2.7648, 1.8030, 2.6007, 2.2874, 2.9364, 3.1608], device='cuda:3'), covar=tensor([0.0352, 0.0584, 0.0725, 0.1853, 0.0830, 0.0901, 0.0773, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0140, 0.0157, 0.0142, 0.0134, 0.0126, 0.0137, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 02:35:33,596 INFO [train.py:904] (3/8) Epoch 9, batch 7600, loss[loss=0.2207, simple_loss=0.3036, pruned_loss=0.06888, over 16773.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3069, pruned_loss=0.07359, over 3089876.08 frames. ], batch size: 124, lr: 7.51e-03, grad_scale: 8.0 2023-04-29 02:35:37,866 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:36:05,553 INFO [optim.py:368] (3/8) 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 02:36:28,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6702, 4.9811, 5.0691, 5.0455, 4.9274, 5.5509, 5.0240, 4.8162], device='cuda:3'), covar=tensor([0.1022, 0.1748, 0.1710, 0.1600, 0.2430, 0.0900, 0.1398, 0.2485], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0449, 0.0483, 0.0400, 0.0524, 0.0513, 0.0390, 0.0540], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 02:36:45,919 INFO [train.py:904] (3/8) Epoch 9, batch 7650, loss[loss=0.2349, simple_loss=0.313, pruned_loss=0.07836, over 17059.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3075, pruned_loss=0.07416, over 3090222.44 frames. ], batch size: 55, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:37:15,891 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4794, 2.4859, 1.8262, 2.2336, 3.0196, 2.5705, 3.3103, 3.2901], device='cuda:3'), covar=tensor([0.0053, 0.0269, 0.0452, 0.0367, 0.0161, 0.0277, 0.0145, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0188, 0.0186, 0.0183, 0.0185, 0.0187, 0.0186, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:38:01,880 INFO [train.py:904] (3/8) Epoch 9, batch 7700, loss[loss=0.2369, simple_loss=0.309, pruned_loss=0.08244, over 16348.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.3069, pruned_loss=0.07425, over 3099162.14 frames. ], batch size: 35, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:04,003 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 02:38:34,565 INFO [optim.py:368] (3/8) 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 02:39:08,827 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:39:16,827 INFO [train.py:904] (3/8) Epoch 9, batch 7750, loss[loss=0.248, simple_loss=0.3125, pruned_loss=0.09178, over 11505.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3079, pruned_loss=0.07481, over 3078436.27 frames. ], batch size: 246, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:39:58,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1321, 2.8949, 2.6545, 2.0019, 2.5456, 2.1650, 2.8264, 2.9698], device='cuda:3'), covar=tensor([0.0288, 0.0578, 0.0652, 0.1660, 0.0873, 0.0972, 0.0590, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0139, 0.0157, 0.0142, 0.0134, 0.0125, 0.0137, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 02:40:20,776 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:40:29,345 INFO [train.py:904] (3/8) Epoch 9, batch 7800, loss[loss=0.2115, simple_loss=0.3046, pruned_loss=0.05919, over 17192.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.309, pruned_loss=0.0759, over 3058429.74 frames. ], batch size: 44, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:41:02,673 INFO [optim.py:368] (3/8) 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 02:41:07,702 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:41:42,270 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1134, 4.1817, 4.3617, 4.1734, 4.2139, 4.7427, 4.3339, 4.0008], device='cuda:3'), covar=tensor([0.1680, 0.1972, 0.2031, 0.2002, 0.2393, 0.1123, 0.1603, 0.2717], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0451, 0.0485, 0.0402, 0.0524, 0.0515, 0.0391, 0.0541], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 02:41:44,457 INFO [train.py:904] (3/8) Epoch 9, batch 7850, loss[loss=0.2155, simple_loss=0.3012, pruned_loss=0.06487, over 16381.00 frames. ], tot_loss[loss=0.2316, simple_loss=0.3106, pruned_loss=0.07626, over 3050062.63 frames. ], batch size: 146, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:42:39,018 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:42:57,229 INFO [train.py:904] (3/8) Epoch 9, batch 7900, loss[loss=0.2534, simple_loss=0.3308, pruned_loss=0.08802, over 16863.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3098, pruned_loss=0.07513, over 3076332.13 frames. ], batch size: 116, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:43:28,442 INFO [optim.py:368] (3/8) 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 02:44:05,657 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6835, 4.7659, 5.2362, 5.2273, 5.1988, 4.8616, 4.8320, 4.6161], device='cuda:3'), covar=tensor([0.0274, 0.0381, 0.0321, 0.0344, 0.0421, 0.0265, 0.0783, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0303, 0.0310, 0.0297, 0.0349, 0.0325, 0.0429, 0.0261], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-29 02:44:12,923 INFO [train.py:904] (3/8) Epoch 9, batch 7950, loss[loss=0.2839, simple_loss=0.3338, pruned_loss=0.117, over 11522.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3094, pruned_loss=0.07454, over 3093540.06 frames. ], batch size: 248, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:44:39,099 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:45:26,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0839, 2.2790, 1.7363, 2.1302, 2.7746, 2.4334, 3.0135, 3.0125], device='cuda:3'), covar=tensor([0.0064, 0.0234, 0.0353, 0.0293, 0.0145, 0.0239, 0.0137, 0.0137], device='cuda:3'), in_proj_covar=tensor([0.0119, 0.0187, 0.0185, 0.0184, 0.0185, 0.0187, 0.0188, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:45:26,762 INFO [train.py:904] (3/8) Epoch 9, batch 8000, loss[loss=0.2103, simple_loss=0.2954, pruned_loss=0.06256, over 16705.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3093, pruned_loss=0.07488, over 3086346.12 frames. ], batch size: 89, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:45:53,908 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2574, 1.9164, 1.5722, 1.7016, 2.2325, 2.0183, 2.2502, 2.4292], device='cuda:3'), covar=tensor([0.0071, 0.0192, 0.0294, 0.0269, 0.0114, 0.0197, 0.0120, 0.0127], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0187, 0.0186, 0.0185, 0.0186, 0.0188, 0.0188, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:45:59,520 INFO [optim.py:368] (3/8) 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 02:46:09,170 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 8050, loss[loss=0.2379, simple_loss=0.321, pruned_loss=0.07735, over 16852.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.3087, pruned_loss=0.07386, over 3111413.06 frames. ], batch size: 102, lr: 7.49e-03, grad_scale: 4.0 2023-04-29 02:47:28,822 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-29 02:47:55,804 INFO [train.py:904] (3/8) Epoch 9, batch 8100, loss[loss=0.2003, simple_loss=0.2848, pruned_loss=0.05787, over 16230.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3087, pruned_loss=0.0735, over 3102096.20 frames. ], batch size: 35, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:48:29,433 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.308e+02 2.948e+02 3.621e+02 4.670e+02 8.254e+02, threshold=7.243e+02, percent-clipped=0.0 2023-04-29 02:49:10,971 INFO [train.py:904] (3/8) Epoch 9, batch 8150, loss[loss=0.1955, simple_loss=0.2801, pruned_loss=0.05541, over 16912.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3066, pruned_loss=0.07296, over 3091672.37 frames. ], batch size: 96, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:49:45,544 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:49:59,392 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:50:11,781 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6301, 4.4498, 4.6973, 4.8707, 5.0057, 4.4737, 4.9834, 4.9280], device='cuda:3'), covar=tensor([0.1372, 0.1038, 0.1209, 0.0551, 0.0466, 0.0894, 0.0444, 0.0545], device='cuda:3'), in_proj_covar=tensor([0.0483, 0.0599, 0.0729, 0.0602, 0.0465, 0.0461, 0.0485, 0.0541], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:50:27,524 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 02:50:28,004 INFO [train.py:904] (3/8) Epoch 9, batch 8200, loss[loss=0.2168, simple_loss=0.2836, pruned_loss=0.07503, over 11662.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3043, pruned_loss=0.07257, over 3086166.45 frames. ], batch size: 247, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:50:51,635 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6784, 2.6923, 2.3474, 3.3663, 2.2138, 3.7209, 1.3193, 2.8840], device='cuda:3'), covar=tensor([0.1385, 0.0527, 0.1050, 0.0149, 0.0117, 0.0366, 0.1590, 0.0666], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0155, 0.0178, 0.0129, 0.0205, 0.0207, 0.0177, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 02:51:05,356 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.992e+02 3.313e+02 4.005e+02 4.576e+02 8.683e+02, threshold=8.011e+02, percent-clipped=3.0 2023-04-29 02:51:20,804 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:27,498 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:51:37,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6325, 3.7543, 2.9923, 2.1149, 2.4914, 2.3189, 3.9243, 3.5977], device='cuda:3'), covar=tensor([0.2321, 0.0537, 0.1316, 0.2237, 0.2224, 0.1717, 0.0341, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0257, 0.0284, 0.0273, 0.0280, 0.0216, 0.0264, 0.0283], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:51:47,430 INFO [train.py:904] (3/8) Epoch 9, batch 8250, loss[loss=0.1921, simple_loss=0.289, pruned_loss=0.04765, over 16395.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3038, pruned_loss=0.07075, over 3075527.73 frames. ], batch size: 75, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:51:54,951 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4932, 3.0374, 2.6579, 2.2271, 2.2799, 2.0974, 3.0246, 2.9790], device='cuda:3'), covar=tensor([0.2422, 0.0670, 0.1432, 0.2179, 0.2394, 0.2017, 0.0495, 0.1067], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0257, 0.0283, 0.0273, 0.0280, 0.0217, 0.0264, 0.0283], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:52:43,204 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:52:46,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9925, 2.4247, 2.3470, 2.9294, 2.1111, 3.3125, 1.6300, 2.8304], device='cuda:3'), covar=tensor([0.1118, 0.0486, 0.0859, 0.0102, 0.0087, 0.0336, 0.1264, 0.0592], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0152, 0.0175, 0.0126, 0.0200, 0.0202, 0.0174, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 02:53:06,322 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:53:08,808 INFO [train.py:904] (3/8) Epoch 9, batch 8300, loss[loss=0.2028, simple_loss=0.2965, pruned_loss=0.05457, over 16920.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3007, pruned_loss=0.06767, over 3059135.29 frames. ], batch size: 116, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:53:21,211 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:44,171 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9219, 3.7801, 4.0194, 4.1534, 4.2640, 3.8123, 4.1898, 4.2382], device='cuda:3'), covar=tensor([0.1403, 0.1119, 0.1231, 0.0603, 0.0479, 0.1366, 0.0603, 0.0634], device='cuda:3'), in_proj_covar=tensor([0.0481, 0.0595, 0.0723, 0.0596, 0.0461, 0.0458, 0.0481, 0.0538], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:53:47,008 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 9, batch 8350, loss[loss=0.2251, simple_loss=0.2982, pruned_loss=0.07594, over 12033.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.3001, pruned_loss=0.06535, over 3075861.56 frames. ], batch size: 246, lr: 7.47e-03, grad_scale: 2.0 2023-04-29 02:55:01,207 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:55:01,428 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 02:55:07,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 02:55:52,200 INFO [train.py:904] (3/8) Epoch 9, batch 8400, loss[loss=0.2087, simple_loss=0.2967, pruned_loss=0.0604, over 16669.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2971, pruned_loss=0.06291, over 3056010.93 frames. ], batch size: 134, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:56:15,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4742, 3.0643, 2.5877, 2.2773, 2.3227, 2.1032, 3.0578, 2.9567], device='cuda:3'), covar=tensor([0.2432, 0.0862, 0.1454, 0.1969, 0.2345, 0.2084, 0.0539, 0.1182], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0249, 0.0277, 0.0266, 0.0269, 0.0212, 0.0256, 0.0275], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 02:56:31,315 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.203e+02 2.904e+02 3.369e+02 3.930e+02 8.032e+02, threshold=6.737e+02, percent-clipped=2.0 2023-04-29 02:57:13,775 INFO [train.py:904] (3/8) Epoch 9, batch 8450, loss[loss=0.1764, simple_loss=0.2745, pruned_loss=0.03911, over 16658.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2951, pruned_loss=0.06092, over 3074718.85 frames. ], batch size: 89, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:58:00,114 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:04,828 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:34,963 INFO [train.py:904] (3/8) Epoch 9, batch 8500, loss[loss=0.1771, simple_loss=0.2644, pruned_loss=0.04488, over 16660.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2904, pruned_loss=0.05783, over 3070684.80 frames. ], batch size: 83, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:58:53,536 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6569, 2.6709, 1.7959, 2.7912, 2.1148, 2.8218, 2.0022, 2.4574], device='cuda:3'), covar=tensor([0.0219, 0.0284, 0.1191, 0.0164, 0.0644, 0.0397, 0.1172, 0.0512], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0155, 0.0181, 0.0110, 0.0161, 0.0194, 0.0190, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 02:59:14,097 INFO [optim.py:368] (3/8) 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 02:59:20,471 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:38,940 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:58,593 INFO [train.py:904] (3/8) Epoch 9, batch 8550, loss[loss=0.2297, simple_loss=0.3281, pruned_loss=0.06562, over 16686.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2877, pruned_loss=0.05699, over 3039796.61 frames. ], batch size: 134, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 03:00:55,136 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:01:24,771 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:01:37,952 INFO [train.py:904] (3/8) Epoch 9, batch 8600, loss[loss=0.1845, simple_loss=0.2619, pruned_loss=0.05354, over 12205.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2876, pruned_loss=0.05617, over 3024262.46 frames. ], batch size: 247, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:02:25,744 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:02:26,428 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 9, batch 8650, loss[loss=0.174, simple_loss=0.2666, pruned_loss=0.04064, over 16686.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2855, pruned_loss=0.05448, over 3022649.96 frames. ], batch size: 134, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:03:47,294 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:47,452 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:04:03,131 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-29 03:04:05,356 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:05:02,347 INFO [train.py:904] (3/8) Epoch 9, batch 8700, loss[loss=0.1843, simple_loss=0.2737, pruned_loss=0.04746, over 16666.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2825, pruned_loss=0.05271, over 3037789.06 frames. ], batch size: 134, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:05:45,060 INFO [optim.py:368] (3/8) 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 03:05:45,831 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:06:10,381 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4187, 3.8494, 3.9393, 2.5359, 3.4634, 3.8837, 3.6797, 2.2359], device='cuda:3'), covar=tensor([0.0349, 0.0020, 0.0022, 0.0285, 0.0062, 0.0049, 0.0047, 0.0346], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0063, 0.0064, 0.0122, 0.0072, 0.0082, 0.0071, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 03:06:36,272 INFO [train.py:904] (3/8) Epoch 9, batch 8750, loss[loss=0.1795, simple_loss=0.2629, pruned_loss=0.04807, over 12305.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2827, pruned_loss=0.05231, over 3041488.38 frames. ], batch size: 249, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:06:40,861 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 03:06:52,865 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2803, 4.2964, 4.1476, 3.9151, 3.8172, 4.2623, 3.9815, 3.9866], device='cuda:3'), covar=tensor([0.0504, 0.0382, 0.0280, 0.0254, 0.0825, 0.0338, 0.0546, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0219, 0.0265, 0.0253, 0.0236, 0.0274, 0.0266, 0.0177, 0.0295], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:06:52,984 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7639, 3.3968, 3.2474, 1.7582, 2.7754, 2.2888, 3.1604, 3.3585], device='cuda:3'), covar=tensor([0.0291, 0.0564, 0.0527, 0.1881, 0.0744, 0.0862, 0.0752, 0.0790], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0131, 0.0152, 0.0138, 0.0129, 0.0122, 0.0132, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 03:07:34,398 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-29 03:08:32,638 INFO [train.py:904] (3/8) Epoch 9, batch 8800, loss[loss=0.1991, simple_loss=0.2835, pruned_loss=0.05736, over 16822.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2805, pruned_loss=0.05113, over 3044480.29 frames. ], batch size: 124, lr: 7.46e-03, grad_scale: 8.0 2023-04-29 03:09:21,968 INFO [optim.py:368] (3/8) 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 03:09:31,736 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:44,403 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:54,914 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 03:10:04,220 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4965, 3.6519, 2.7542, 2.0387, 2.4423, 2.1797, 3.8434, 3.3304], device='cuda:3'), covar=tensor([0.2568, 0.0645, 0.1459, 0.2169, 0.2155, 0.1666, 0.0432, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0239, 0.0266, 0.0256, 0.0252, 0.0204, 0.0248, 0.0263], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:10:17,857 INFO [train.py:904] (3/8) Epoch 9, batch 8850, loss[loss=0.1605, simple_loss=0.2531, pruned_loss=0.034, over 12052.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2824, pruned_loss=0.05024, over 3037779.01 frames. ], batch size: 248, lr: 7.45e-03, grad_scale: 8.0 2023-04-29 03:11:12,096 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:49,621 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 03:12:02,853 INFO [train.py:904] (3/8) Epoch 9, batch 8900, loss[loss=0.1854, simple_loss=0.2711, pruned_loss=0.04984, over 12495.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2831, pruned_loss=0.04978, over 3040848.37 frames. ], batch size: 247, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:12:56,436 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7749, 3.6779, 3.8713, 3.9829, 4.0567, 3.6593, 4.0185, 4.0526], device='cuda:3'), covar=tensor([0.1266, 0.0965, 0.1061, 0.0540, 0.0474, 0.1473, 0.0559, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0454, 0.0559, 0.0677, 0.0565, 0.0437, 0.0435, 0.0455, 0.0510], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:12:57,513 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 2.641e+02 3.095e+02 3.699e+02 6.742e+02, threshold=6.190e+02, percent-clipped=4.0 2023-04-29 03:13:07,419 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:10,504 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:14:08,713 INFO [train.py:904] (3/8) Epoch 9, batch 8950, loss[loss=0.2026, simple_loss=0.2813, pruned_loss=0.06197, over 12720.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2825, pruned_loss=0.05002, over 3052170.50 frames. ], batch size: 247, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:14:38,300 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:14:49,206 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7545, 5.1136, 5.2806, 5.1061, 5.0938, 5.6829, 5.2561, 4.9853], device='cuda:3'), covar=tensor([0.0820, 0.1459, 0.1528, 0.1764, 0.2200, 0.0893, 0.1291, 0.2053], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0435, 0.0465, 0.0383, 0.0501, 0.0490, 0.0380, 0.0509], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:15:34,995 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:57,359 INFO [train.py:904] (3/8) Epoch 9, batch 9000, loss[loss=0.1896, simple_loss=0.2698, pruned_loss=0.05473, over 12248.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.279, pruned_loss=0.0484, over 3046697.21 frames. ], batch size: 248, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:15:57,360 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 03:16:07,534 INFO [train.py:938] (3/8) Epoch 9, validation: loss=0.1581, simple_loss=0.2623, pruned_loss=0.02697, over 944034.00 frames. 2023-04-29 03:16:07,534 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 03:16:31,272 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:16:47,982 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:16:58,704 INFO [optim.py:368] (3/8) 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 03:17:08,238 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:17:51,878 INFO [train.py:904] (3/8) Epoch 9, batch 9050, loss[loss=0.1687, simple_loss=0.2602, pruned_loss=0.03856, over 16648.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2796, pruned_loss=0.04868, over 3056721.07 frames. ], batch size: 89, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:18:24,987 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4031, 2.0066, 2.0288, 3.8910, 1.9831, 2.4299, 2.1075, 2.1441], device='cuda:3'), covar=tensor([0.0697, 0.3057, 0.2200, 0.0342, 0.3755, 0.2011, 0.3105, 0.3059], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0363, 0.0308, 0.0304, 0.0395, 0.0406, 0.0327, 0.0422], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:19:19,191 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:19:22,971 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 03:19:36,842 INFO [train.py:904] (3/8) Epoch 9, batch 9100, loss[loss=0.1707, simple_loss=0.2722, pruned_loss=0.03464, over 16909.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.28, pruned_loss=0.04926, over 3060624.05 frames. ], batch size: 96, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:19:41,951 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8303, 3.3654, 2.5788, 4.8980, 3.9614, 4.3096, 1.7088, 3.0218], device='cuda:3'), covar=tensor([0.1268, 0.0476, 0.1120, 0.0099, 0.0160, 0.0378, 0.1300, 0.0763], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0150, 0.0174, 0.0123, 0.0186, 0.0201, 0.0173, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 03:19:50,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 03:19:56,079 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5494, 4.5117, 4.9544, 4.9656, 4.9331, 4.6377, 4.6281, 4.5198], device='cuda:3'), covar=tensor([0.0222, 0.0373, 0.0336, 0.0315, 0.0323, 0.0241, 0.0666, 0.0272], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0297, 0.0304, 0.0295, 0.0342, 0.0319, 0.0417, 0.0257], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-29 03:20:27,097 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5542, 1.5246, 2.0122, 2.5918, 2.4570, 2.8354, 1.9046, 2.7002], device='cuda:3'), covar=tensor([0.0134, 0.0361, 0.0227, 0.0174, 0.0199, 0.0127, 0.0299, 0.0093], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0159, 0.0143, 0.0143, 0.0153, 0.0109, 0.0159, 0.0099], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 03:20:34,108 INFO [optim.py:368] (3/8) 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 03:20:58,614 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:33,802 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 9150, loss[loss=0.1883, simple_loss=0.2696, pruned_loss=0.0535, over 12205.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2805, pruned_loss=0.04897, over 3050458.58 frames. ], batch size: 248, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:22:45,425 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 9200, loss[loss=0.163, simple_loss=0.2527, pruned_loss=0.03665, over 16725.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2758, pruned_loss=0.04784, over 3056999.53 frames. ], batch size: 39, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:23:42,358 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.608e+02 2.952e+02 3.728e+02 7.219e+02, threshold=5.904e+02, percent-clipped=3.0 2023-04-29 03:24:58,585 INFO [train.py:904] (3/8) Epoch 9, batch 9250, loss[loss=0.1553, simple_loss=0.253, pruned_loss=0.02874, over 16856.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2752, pruned_loss=0.04797, over 3037273.98 frames. ], batch size: 96, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:25:14,824 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:13,013 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:38,551 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 9300, loss[loss=0.1752, simple_loss=0.2695, pruned_loss=0.04042, over 16692.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2742, pruned_loss=0.04746, over 3035778.53 frames. ], batch size: 83, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:27:35,072 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:27:36,967 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:27:46,913 INFO [optim.py:368] (3/8) 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 03:28:32,111 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 03:28:36,077 INFO [train.py:904] (3/8) Epoch 9, batch 9350, loss[loss=0.1931, simple_loss=0.281, pruned_loss=0.05265, over 16691.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2738, pruned_loss=0.04732, over 3044437.62 frames. ], batch size: 134, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:28:44,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3436, 1.9619, 1.6416, 1.6308, 2.2019, 1.8450, 2.1232, 2.3139], device='cuda:3'), covar=tensor([0.0068, 0.0237, 0.0345, 0.0300, 0.0141, 0.0247, 0.0109, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0185, 0.0183, 0.0182, 0.0181, 0.0185, 0.0176, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:28:48,616 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:14,438 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:50,812 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 9400, loss[loss=0.1539, simple_loss=0.2408, pruned_loss=0.03348, over 12515.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2742, pruned_loss=0.04691, over 3052188.58 frames. ], batch size: 249, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:31:09,564 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 9, batch 9450, loss[loss=0.1991, simple_loss=0.288, pruned_loss=0.05508, over 16214.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2761, pruned_loss=0.04717, over 3053595.05 frames. ], batch size: 165, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:32:38,473 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1835, 3.2643, 3.3148, 1.6539, 3.5118, 3.5796, 2.8369, 2.6903], device='cuda:3'), covar=tensor([0.0753, 0.0166, 0.0165, 0.1202, 0.0053, 0.0123, 0.0351, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0092, 0.0078, 0.0133, 0.0063, 0.0089, 0.0113, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-29 03:33:43,967 INFO [train.py:904] (3/8) Epoch 9, batch 9500, loss[loss=0.1537, simple_loss=0.2418, pruned_loss=0.03281, over 17032.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2748, pruned_loss=0.04624, over 3070122.70 frames. ], batch size: 50, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:58,714 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:34:09,294 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 03:34:35,317 INFO [optim.py:368] (3/8) 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 03:35:25,207 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4198, 3.2573, 2.6568, 2.0596, 2.1841, 2.1262, 3.3390, 3.0686], device='cuda:3'), covar=tensor([0.2619, 0.0757, 0.1665, 0.2329, 0.2156, 0.1826, 0.0588, 0.1053], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0243, 0.0270, 0.0259, 0.0248, 0.0207, 0.0252, 0.0265], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:35:30,036 INFO [train.py:904] (3/8) Epoch 9, batch 9550, loss[loss=0.2018, simple_loss=0.2887, pruned_loss=0.05748, over 15350.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2739, pruned_loss=0.04646, over 3071680.20 frames. ], batch size: 191, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:36:43,185 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:11,180 INFO [train.py:904] (3/8) Epoch 9, batch 9600, loss[loss=0.2014, simple_loss=0.287, pruned_loss=0.05788, over 16708.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2755, pruned_loss=0.0472, over 3077533.10 frames. ], batch size: 76, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:37:37,149 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:59,117 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.542e+02 3.036e+02 4.023e+02 8.440e+02, threshold=6.073e+02, percent-clipped=4.0 2023-04-29 03:38:17,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:38:52,296 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1038, 3.0704, 3.1309, 1.6450, 3.3252, 3.3949, 2.7837, 2.6129], device='cuda:3'), covar=tensor([0.0781, 0.0184, 0.0146, 0.1180, 0.0065, 0.0118, 0.0378, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0093, 0.0080, 0.0135, 0.0064, 0.0090, 0.0114, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-29 03:38:55,293 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 9650, loss[loss=0.2097, simple_loss=0.2974, pruned_loss=0.06097, over 15308.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2773, pruned_loss=0.04783, over 3054236.71 frames. ], batch size: 191, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:39:03,429 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:39:27,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1321, 3.2388, 1.6412, 3.4939, 2.2293, 3.4461, 1.7457, 2.5931], device='cuda:3'), covar=tensor([0.0259, 0.0324, 0.1833, 0.0132, 0.1021, 0.0455, 0.1784, 0.0701], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0154, 0.0181, 0.0109, 0.0161, 0.0191, 0.0190, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 03:39:34,853 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6221, 2.6710, 1.7629, 2.8135, 2.0791, 2.7817, 1.9425, 2.4201], device='cuda:3'), covar=tensor([0.0233, 0.0291, 0.1217, 0.0167, 0.0697, 0.0396, 0.1255, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0154, 0.0181, 0.0109, 0.0161, 0.0190, 0.0190, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 03:39:41,726 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8473, 2.3211, 2.2434, 3.0670, 2.0276, 3.2787, 1.5540, 2.6671], device='cuda:3'), covar=tensor([0.1282, 0.0600, 0.1084, 0.0134, 0.0093, 0.0425, 0.1499, 0.0724], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0151, 0.0177, 0.0123, 0.0182, 0.0203, 0.0176, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 03:40:18,969 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 9, batch 9700, loss[loss=0.1636, simple_loss=0.2606, pruned_loss=0.03332, over 16797.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2761, pruned_loss=0.04742, over 3051100.47 frames. ], batch size: 83, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:41:04,646 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:41:40,522 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.339e+02 3.062e+02 3.716e+02 7.920e+02, threshold=6.123e+02, percent-clipped=1.0 2023-04-29 03:41:59,658 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:42:28,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6269, 3.5226, 2.8457, 2.1538, 2.3357, 2.1663, 3.6834, 3.2537], device='cuda:3'), covar=tensor([0.2284, 0.0561, 0.1389, 0.2160, 0.2205, 0.1739, 0.0347, 0.0826], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0244, 0.0269, 0.0261, 0.0247, 0.0207, 0.0252, 0.0265], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:42:31,488 INFO [train.py:904] (3/8) Epoch 9, batch 9750, loss[loss=0.1731, simple_loss=0.2689, pruned_loss=0.03867, over 16497.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.275, pruned_loss=0.04762, over 3039031.82 frames. ], batch size: 68, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:43:57,949 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-04-29 03:44:10,868 INFO [train.py:904] (3/8) Epoch 9, batch 9800, loss[loss=0.177, simple_loss=0.2721, pruned_loss=0.04095, over 16552.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2752, pruned_loss=0.04694, over 3039921.03 frames. ], batch size: 57, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:44:21,832 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:44:40,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6617, 4.7114, 5.1408, 5.0813, 5.0817, 4.7974, 4.7193, 4.5492], device='cuda:3'), covar=tensor([0.0242, 0.0390, 0.0311, 0.0360, 0.0394, 0.0271, 0.0813, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0292, 0.0297, 0.0286, 0.0333, 0.0312, 0.0402, 0.0251], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-29 03:44:57,914 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.313e+02 2.751e+02 3.431e+02 5.847e+02, threshold=5.502e+02, percent-clipped=0.0 2023-04-29 03:45:57,961 INFO [train.py:904] (3/8) Epoch 9, batch 9850, loss[loss=0.1667, simple_loss=0.2563, pruned_loss=0.03851, over 12972.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2761, pruned_loss=0.04658, over 3046347.08 frames. ], batch size: 248, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:46:04,899 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:46:47,374 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4658, 3.5519, 2.6453, 2.0918, 2.3916, 2.0304, 3.7090, 3.2464], device='cuda:3'), covar=tensor([0.2721, 0.0702, 0.1659, 0.2318, 0.2343, 0.1948, 0.0431, 0.1002], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0244, 0.0268, 0.0260, 0.0245, 0.0207, 0.0251, 0.0264], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:47:48,700 INFO [train.py:904] (3/8) Epoch 9, batch 9900, loss[loss=0.1844, simple_loss=0.2883, pruned_loss=0.0403, over 16860.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2765, pruned_loss=0.04669, over 3041306.14 frames. ], batch size: 102, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:48:20,590 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:48:47,935 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 9, batch 9950, loss[loss=0.1888, simple_loss=0.2731, pruned_loss=0.05227, over 12042.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2785, pruned_loss=0.04705, over 3037780.44 frames. ], batch size: 248, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:49:49,172 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:49:59,592 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6711, 2.0652, 1.7063, 1.8108, 2.4437, 2.2180, 2.5067, 2.6249], device='cuda:3'), covar=tensor([0.0064, 0.0290, 0.0328, 0.0335, 0.0139, 0.0226, 0.0136, 0.0141], device='cuda:3'), in_proj_covar=tensor([0.0117, 0.0190, 0.0184, 0.0184, 0.0184, 0.0187, 0.0178, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:50:14,540 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:50,888 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 03:51:44,349 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:47,236 INFO [train.py:904] (3/8) Epoch 9, batch 10000, loss[loss=0.1835, simple_loss=0.27, pruned_loss=0.04846, over 12592.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2771, pruned_loss=0.04652, over 3051758.54 frames. ], batch size: 248, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:51:53,873 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:56,267 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 03:52:35,744 INFO [optim.py:368] (3/8) 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 03:52:37,488 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-29 03:53:27,231 INFO [train.py:904] (3/8) Epoch 9, batch 10050, loss[loss=0.1948, simple_loss=0.2805, pruned_loss=0.05452, over 11619.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2769, pruned_loss=0.04632, over 3039534.32 frames. ], batch size: 248, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:53:33,746 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4477, 3.9918, 4.1131, 2.8012, 3.7439, 4.1628, 3.8811, 2.3648], device='cuda:3'), covar=tensor([0.0337, 0.0023, 0.0027, 0.0236, 0.0053, 0.0035, 0.0038, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0062, 0.0064, 0.0120, 0.0072, 0.0078, 0.0071, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 03:54:00,780 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2262, 1.9394, 2.0836, 3.7279, 1.8452, 2.3394, 2.1204, 2.0929], device='cuda:3'), covar=tensor([0.0811, 0.3209, 0.2184, 0.0401, 0.3982, 0.2169, 0.2813, 0.2998], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0360, 0.0306, 0.0304, 0.0390, 0.0400, 0.0322, 0.0415], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:54:30,790 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:55:01,276 INFO [train.py:904] (3/8) Epoch 9, batch 10100, loss[loss=0.1784, simple_loss=0.2587, pruned_loss=0.04906, over 12786.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2775, pruned_loss=0.04649, over 3064159.74 frames. ], batch size: 248, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:55:52,059 INFO [optim.py:368] (3/8) 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 03:56:14,363 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:56:45,258 INFO [train.py:904] (3/8) Epoch 10, batch 0, loss[loss=0.2916, simple_loss=0.3309, pruned_loss=0.1262, over 16829.00 frames. ], tot_loss[loss=0.2916, simple_loss=0.3309, pruned_loss=0.1262, over 16829.00 frames. ], batch size: 102, lr: 7.04e-03, grad_scale: 8.0 2023-04-29 03:56:45,259 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 03:56:50,514 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2868, 5.6208, 5.4217, 5.5079, 5.1941, 5.1285, 4.9949, 5.6486], device='cuda:3'), covar=tensor([0.0793, 0.0665, 0.0632, 0.0485, 0.0721, 0.0281, 0.0791, 0.0635], device='cuda:3'), in_proj_covar=tensor([0.0467, 0.0596, 0.0485, 0.0407, 0.0373, 0.0388, 0.0496, 0.0443], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:56:52,895 INFO [train.py:938] (3/8) Epoch 10, validation: loss=0.158, simple_loss=0.2614, pruned_loss=0.02732, over 944034.00 frames. 2023-04-29 03:56:52,895 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 03:58:02,505 INFO [train.py:904] (3/8) Epoch 10, batch 50, loss[loss=0.2256, simple_loss=0.2898, pruned_loss=0.0807, over 16828.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.2961, pruned_loss=0.07074, over 744448.05 frames. ], batch size: 102, lr: 7.04e-03, grad_scale: 2.0 2023-04-29 03:58:39,941 INFO [optim.py:368] (3/8) 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 03:58:53,064 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5403, 2.1116, 2.3096, 4.2624, 2.1207, 2.6977, 2.2625, 2.3862], device='cuda:3'), covar=tensor([0.0808, 0.3107, 0.1945, 0.0498, 0.3325, 0.1750, 0.2734, 0.2472], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0365, 0.0310, 0.0308, 0.0395, 0.0406, 0.0328, 0.0423], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 03:59:08,801 INFO [train.py:904] (3/8) Epoch 10, batch 100, loss[loss=0.2104, simple_loss=0.2757, pruned_loss=0.07254, over 16859.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2864, pruned_loss=0.06316, over 1326231.00 frames. ], batch size: 116, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 03:59:56,269 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5692, 4.4627, 4.9977, 4.9794, 5.0207, 4.5835, 4.6411, 4.4264], device='cuda:3'), covar=tensor([0.0325, 0.0557, 0.0459, 0.0483, 0.0444, 0.0405, 0.0887, 0.0479], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0303, 0.0307, 0.0293, 0.0342, 0.0325, 0.0417, 0.0261], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-29 04:00:16,886 INFO [train.py:904] (3/8) Epoch 10, batch 150, loss[loss=0.202, simple_loss=0.2754, pruned_loss=0.06428, over 16805.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.282, pruned_loss=0.06111, over 1772299.72 frames. ], batch size: 83, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:00:22,930 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:00:56,392 INFO [optim.py:368] (3/8) 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 04:01:12,429 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:01:26,571 INFO [train.py:904] (3/8) Epoch 10, batch 200, loss[loss=0.1736, simple_loss=0.2444, pruned_loss=0.05145, over 16979.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2824, pruned_loss=0.06122, over 2120707.33 frames. ], batch size: 41, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:01:28,066 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:02:34,674 INFO [train.py:904] (3/8) Epoch 10, batch 250, loss[loss=0.1658, simple_loss=0.24, pruned_loss=0.04576, over 16968.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2802, pruned_loss=0.05966, over 2390301.83 frames. ], batch size: 41, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:02:36,443 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:02:54,331 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.52 vs. limit=5.0 2023-04-29 04:03:07,411 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 04:03:11,342 INFO [optim.py:368] (3/8) 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 04:03:27,295 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-04-29 04:03:28,029 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 300, loss[loss=0.1974, simple_loss=0.2864, pruned_loss=0.05419, over 17205.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2767, pruned_loss=0.0579, over 2601630.14 frames. ], batch size: 44, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:03:58,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1586, 5.6281, 5.7843, 5.4808, 5.5915, 6.1147, 5.6781, 5.4476], device='cuda:3'), covar=tensor([0.0765, 0.1593, 0.1707, 0.1752, 0.2637, 0.0903, 0.1326, 0.2259], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0478, 0.0501, 0.0416, 0.0549, 0.0529, 0.0410, 0.0552], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 04:04:51,283 INFO [train.py:904] (3/8) Epoch 10, batch 350, loss[loss=0.1781, simple_loss=0.2524, pruned_loss=0.0519, over 15431.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2739, pruned_loss=0.05576, over 2769868.25 frames. ], batch size: 190, lr: 7.02e-03, grad_scale: 1.0 2023-04-29 04:05:28,619 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.384e+02 2.940e+02 3.585e+02 5.710e+02, threshold=5.881e+02, percent-clipped=0.0 2023-04-29 04:05:35,488 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-29 04:05:59,422 INFO [train.py:904] (3/8) Epoch 10, batch 400, loss[loss=0.2052, simple_loss=0.2773, pruned_loss=0.06653, over 16773.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2716, pruned_loss=0.05543, over 2894080.51 frames. ], batch size: 89, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:11,414 INFO [train.py:904] (3/8) Epoch 10, batch 450, loss[loss=0.1901, simple_loss=0.2697, pruned_loss=0.05526, over 16473.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2704, pruned_loss=0.05505, over 2987630.61 frames. ], batch size: 68, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:50,758 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 500, loss[loss=0.1879, simple_loss=0.2825, pruned_loss=0.04665, over 16609.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2685, pruned_loss=0.05368, over 3065725.97 frames. ], batch size: 62, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:09:15,099 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:09:23,880 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:09:29,464 INFO [train.py:904] (3/8) Epoch 10, batch 550, loss[loss=0.223, simple_loss=0.2834, pruned_loss=0.08127, over 16792.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2673, pruned_loss=0.05305, over 3122242.14 frames. ], batch size: 102, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:10:07,828 INFO [optim.py:368] (3/8) 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 04:10:22,796 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:38,344 INFO [train.py:904] (3/8) Epoch 10, batch 600, loss[loss=0.1646, simple_loss=0.2384, pruned_loss=0.04542, over 16806.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2676, pruned_loss=0.05403, over 3164717.05 frames. ], batch size: 116, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:10:38,833 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:42,878 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:18,691 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:11:30,056 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:51,294 INFO [train.py:904] (3/8) Epoch 10, batch 650, loss[loss=0.1987, simple_loss=0.2892, pruned_loss=0.05416, over 17208.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2664, pruned_loss=0.05344, over 3209397.55 frames. ], batch size: 44, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:12:12,148 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:12:31,800 INFO [optim.py:368] (3/8) 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 04:12:49,497 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 04:13:03,065 INFO [train.py:904] (3/8) Epoch 10, batch 700, loss[loss=0.1532, simple_loss=0.2409, pruned_loss=0.03274, over 16824.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2668, pruned_loss=0.05386, over 3231601.67 frames. ], batch size: 42, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:13:36,305 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7843, 3.5046, 3.0122, 5.1309, 4.3783, 4.7759, 1.8071, 3.4233], device='cuda:3'), covar=tensor([0.1393, 0.0529, 0.1037, 0.0123, 0.0228, 0.0293, 0.1445, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0152, 0.0177, 0.0131, 0.0189, 0.0208, 0.0176, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 04:14:12,201 INFO [train.py:904] (3/8) Epoch 10, batch 750, loss[loss=0.1677, simple_loss=0.2559, pruned_loss=0.03982, over 17231.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2665, pruned_loss=0.05317, over 3241271.82 frames. ], batch size: 44, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:14:30,199 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7635, 4.7060, 4.7071, 4.1137, 4.6985, 1.9965, 4.4511, 4.5605], device='cuda:3'), covar=tensor([0.0087, 0.0073, 0.0118, 0.0328, 0.0079, 0.2190, 0.0102, 0.0157], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0110, 0.0161, 0.0150, 0.0130, 0.0180, 0.0147, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:14:52,015 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 800, loss[loss=0.1996, simple_loss=0.2619, pruned_loss=0.06868, over 16755.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2656, pruned_loss=0.05297, over 3254868.28 frames. ], batch size: 124, lr: 7.01e-03, grad_scale: 4.0 2023-04-29 04:16:27,560 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:16:32,818 INFO [train.py:904] (3/8) Epoch 10, batch 850, loss[loss=0.2007, simple_loss=0.2755, pruned_loss=0.06288, over 16692.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2658, pruned_loss=0.05294, over 3271882.47 frames. ], batch size: 89, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:17:10,136 INFO [optim.py:368] (3/8) 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 04:17:33,336 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:17:34,355 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 900, loss[loss=0.1724, simple_loss=0.2597, pruned_loss=0.04255, over 17111.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2652, pruned_loss=0.05278, over 3277116.99 frames. ], batch size: 49, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:18:34,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5352, 5.9154, 5.6934, 5.7449, 5.3109, 5.1302, 5.3359, 6.0664], device='cuda:3'), covar=tensor([0.1104, 0.0797, 0.0825, 0.0612, 0.0763, 0.0606, 0.0968, 0.0812], device='cuda:3'), in_proj_covar=tensor([0.0526, 0.0666, 0.0548, 0.0456, 0.0419, 0.0424, 0.0562, 0.0500], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:18:50,838 INFO [train.py:904] (3/8) Epoch 10, batch 950, loss[loss=0.1916, simple_loss=0.2615, pruned_loss=0.06089, over 16250.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.265, pruned_loss=0.05188, over 3293961.55 frames. ], batch size: 165, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:19:01,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6194, 4.6718, 4.7220, 4.6298, 4.6574, 5.2469, 4.8158, 4.5027], device='cuda:3'), covar=tensor([0.1192, 0.1897, 0.1782, 0.2044, 0.2820, 0.1051, 0.1373, 0.2354], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0491, 0.0514, 0.0420, 0.0563, 0.0537, 0.0417, 0.0564], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 04:19:04,204 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:19:29,762 INFO [optim.py:368] (3/8) 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 04:19:39,576 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:19:58,793 INFO [train.py:904] (3/8) Epoch 10, batch 1000, loss[loss=0.205, simple_loss=0.2773, pruned_loss=0.06635, over 16466.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2639, pruned_loss=0.0524, over 3298636.29 frames. ], batch size: 75, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:20:59,411 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1763, 1.9209, 2.4067, 2.9034, 2.8764, 3.5073, 2.1138, 3.3352], device='cuda:3'), covar=tensor([0.0145, 0.0329, 0.0233, 0.0183, 0.0184, 0.0108, 0.0306, 0.0094], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0167, 0.0152, 0.0155, 0.0164, 0.0119, 0.0168, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 04:21:03,501 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4982, 3.8553, 3.9760, 2.0682, 3.2089, 2.6201, 3.8838, 3.9951], device='cuda:3'), covar=tensor([0.0284, 0.0776, 0.0460, 0.1774, 0.0756, 0.0871, 0.0634, 0.1029], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0141, 0.0157, 0.0142, 0.0135, 0.0125, 0.0135, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 04:21:09,129 INFO [train.py:904] (3/8) Epoch 10, batch 1050, loss[loss=0.197, simple_loss=0.2633, pruned_loss=0.06538, over 12203.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.263, pruned_loss=0.05195, over 3307013.81 frames. ], batch size: 248, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:21:48,359 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 1100, loss[loss=0.1964, simple_loss=0.2941, pruned_loss=0.0493, over 17090.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2627, pruned_loss=0.05136, over 3302503.03 frames. ], batch size: 53, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:23:28,210 INFO [train.py:904] (3/8) Epoch 10, batch 1150, loss[loss=0.1519, simple_loss=0.238, pruned_loss=0.03294, over 16856.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2623, pruned_loss=0.05108, over 3312461.91 frames. ], batch size: 42, lr: 6.99e-03, grad_scale: 4.0 2023-04-29 04:23:56,888 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2099, 4.3089, 2.5219, 4.8681, 3.1648, 4.8109, 2.7506, 3.6307], device='cuda:3'), covar=tensor([0.0197, 0.0303, 0.1438, 0.0152, 0.0715, 0.0371, 0.1227, 0.0549], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0163, 0.0187, 0.0124, 0.0165, 0.0203, 0.0191, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 04:24:08,393 INFO [optim.py:368] (3/8) 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 04:24:32,814 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 1200, loss[loss=0.1656, simple_loss=0.2511, pruned_loss=0.04001, over 15941.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2618, pruned_loss=0.05065, over 3315337.00 frames. ], batch size: 35, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:25:16,987 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2374, 3.3949, 3.6092, 3.5700, 3.5741, 3.3930, 3.4010, 3.4457], device='cuda:3'), covar=tensor([0.0386, 0.0661, 0.0424, 0.0470, 0.0483, 0.0446, 0.0764, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0339, 0.0338, 0.0326, 0.0380, 0.0356, 0.0464, 0.0286], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 04:25:39,054 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:25:47,779 INFO [train.py:904] (3/8) Epoch 10, batch 1250, loss[loss=0.1721, simple_loss=0.2701, pruned_loss=0.037, over 17244.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2623, pruned_loss=0.05104, over 3312809.48 frames. ], batch size: 52, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:26:01,768 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:20,453 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:27,271 INFO [optim.py:368] (3/8) 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 04:26:37,458 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:26:58,797 INFO [train.py:904] (3/8) Epoch 10, batch 1300, loss[loss=0.1951, simple_loss=0.2807, pruned_loss=0.05475, over 17122.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2626, pruned_loss=0.05093, over 3318722.14 frames. ], batch size: 48, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:27:07,401 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:27:44,064 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:27:45,261 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:07,566 INFO [train.py:904] (3/8) Epoch 10, batch 1350, loss[loss=0.1794, simple_loss=0.2601, pruned_loss=0.04937, over 16819.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2626, pruned_loss=0.05014, over 3324872.03 frames. ], batch size: 102, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:28:34,041 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:47,041 INFO [optim.py:368] (3/8) 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 04:29:07,345 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 04:29:18,974 INFO [train.py:904] (3/8) Epoch 10, batch 1400, loss[loss=0.17, simple_loss=0.2594, pruned_loss=0.04029, over 17022.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.263, pruned_loss=0.05012, over 3329293.06 frames. ], batch size: 50, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:29:30,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5160, 4.4200, 4.4644, 3.9516, 4.4589, 1.6349, 4.1856, 4.1821], device='cuda:3'), covar=tensor([0.0087, 0.0075, 0.0133, 0.0305, 0.0079, 0.2345, 0.0123, 0.0183], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0112, 0.0164, 0.0153, 0.0131, 0.0178, 0.0149, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:30:00,665 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:30:28,851 INFO [train.py:904] (3/8) Epoch 10, batch 1450, loss[loss=0.2234, simple_loss=0.3151, pruned_loss=0.06582, over 17094.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2626, pruned_loss=0.05009, over 3320504.68 frames. ], batch size: 55, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:30:32,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0738, 2.0510, 2.4126, 2.9940, 2.7811, 3.3841, 2.2096, 3.3599], device='cuda:3'), covar=tensor([0.0146, 0.0314, 0.0218, 0.0195, 0.0202, 0.0164, 0.0313, 0.0107], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0169, 0.0154, 0.0157, 0.0165, 0.0121, 0.0170, 0.0110], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 04:31:07,982 INFO [optim.py:368] (3/8) 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 04:31:09,660 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:31:38,221 INFO [train.py:904] (3/8) Epoch 10, batch 1500, loss[loss=0.201, simple_loss=0.2931, pruned_loss=0.05446, over 16647.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2624, pruned_loss=0.05041, over 3319408.69 frames. ], batch size: 62, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:32:05,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7105, 2.7241, 2.5744, 4.0175, 3.4640, 4.0967, 1.3774, 2.8984], device='cuda:3'), covar=tensor([0.1288, 0.0570, 0.0983, 0.0141, 0.0148, 0.0329, 0.1481, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0154, 0.0177, 0.0135, 0.0195, 0.0211, 0.0177, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 04:32:20,404 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-29 04:32:34,113 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:45,549 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:48,539 INFO [train.py:904] (3/8) Epoch 10, batch 1550, loss[loss=0.1555, simple_loss=0.2331, pruned_loss=0.03901, over 15755.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2636, pruned_loss=0.05136, over 3326902.46 frames. ], batch size: 35, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:33:26,143 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.944e+02 3.490e+02 4.097e+02 8.318e+02, threshold=6.980e+02, percent-clipped=5.0 2023-04-29 04:33:28,318 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:33:56,442 INFO [train.py:904] (3/8) Epoch 10, batch 1600, loss[loss=0.178, simple_loss=0.2713, pruned_loss=0.04234, over 17125.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2657, pruned_loss=0.05233, over 3314468.80 frames. ], batch size: 48, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:34:07,757 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:13,995 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9470, 2.5660, 2.1600, 2.1813, 2.9350, 2.6582, 3.1311, 3.0698], device='cuda:3'), covar=tensor([0.0120, 0.0277, 0.0334, 0.0343, 0.0157, 0.0253, 0.0163, 0.0157], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0202, 0.0198, 0.0196, 0.0199, 0.0201, 0.0208, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:34:28,987 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:37,372 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:39,571 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 04:34:52,651 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:06,231 INFO [train.py:904] (3/8) Epoch 10, batch 1650, loss[loss=0.2049, simple_loss=0.2762, pruned_loss=0.06676, over 16888.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2676, pruned_loss=0.05279, over 3313882.95 frames. ], batch size: 116, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:35:18,207 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:22,330 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5928, 2.5977, 2.2886, 2.2367, 2.9282, 2.6290, 3.4391, 3.1447], device='cuda:3'), covar=tensor([0.0081, 0.0260, 0.0299, 0.0324, 0.0188, 0.0268, 0.0161, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0204, 0.0200, 0.0198, 0.0201, 0.0203, 0.0210, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:35:45,199 INFO [optim.py:368] (3/8) 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 04:35:53,133 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:53,146 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:15,631 INFO [train.py:904] (3/8) Epoch 10, batch 1700, loss[loss=0.1897, simple_loss=0.2838, pruned_loss=0.04782, over 17129.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.269, pruned_loss=0.05285, over 3318340.12 frames. ], batch size: 49, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:36:42,602 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:47,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1358, 3.9597, 4.1508, 4.3360, 4.4488, 3.9883, 4.1765, 4.4131], device='cuda:3'), covar=tensor([0.1268, 0.1015, 0.1284, 0.0697, 0.0514, 0.1314, 0.1758, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0537, 0.0655, 0.0809, 0.0669, 0.0502, 0.0506, 0.0522, 0.0594], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:36:48,267 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:37:02,430 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 04:37:17,855 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:37:23,690 INFO [train.py:904] (3/8) Epoch 10, batch 1750, loss[loss=0.2509, simple_loss=0.3281, pruned_loss=0.08689, over 12484.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2707, pruned_loss=0.05422, over 3317081.59 frames. ], batch size: 246, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:37:42,802 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:38:01,414 INFO [optim.py:368] (3/8) 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 04:38:08,540 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0298, 4.1117, 4.4297, 2.0564, 4.7267, 4.6898, 3.3596, 3.6123], device='cuda:3'), covar=tensor([0.0666, 0.0155, 0.0155, 0.1130, 0.0039, 0.0096, 0.0327, 0.0327], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0099, 0.0088, 0.0143, 0.0070, 0.0101, 0.0122, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 04:38:16,380 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3561, 4.4155, 4.6668, 2.4693, 4.9841, 4.9926, 3.6075, 3.8950], device='cuda:3'), covar=tensor([0.0577, 0.0152, 0.0168, 0.0990, 0.0042, 0.0105, 0.0294, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0099, 0.0087, 0.0142, 0.0070, 0.0101, 0.0122, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 04:38:21,690 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5682, 5.9376, 5.6483, 5.7312, 5.3604, 5.2057, 5.4295, 6.0880], device='cuda:3'), covar=tensor([0.1152, 0.0827, 0.1090, 0.0668, 0.0830, 0.0623, 0.0993, 0.0775], device='cuda:3'), in_proj_covar=tensor([0.0523, 0.0670, 0.0552, 0.0461, 0.0421, 0.0428, 0.0558, 0.0504], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:38:32,562 INFO [train.py:904] (3/8) Epoch 10, batch 1800, loss[loss=0.1609, simple_loss=0.2571, pruned_loss=0.0324, over 17271.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2719, pruned_loss=0.05403, over 3317810.67 frames. ], batch size: 52, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:06,659 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:09,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9971, 4.9576, 4.7472, 4.2521, 4.8933, 1.9402, 4.6303, 4.7558], device='cuda:3'), covar=tensor([0.0064, 0.0062, 0.0137, 0.0321, 0.0069, 0.2080, 0.0101, 0.0147], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0115, 0.0166, 0.0156, 0.0134, 0.0179, 0.0152, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:39:11,086 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6148, 2.2585, 2.3473, 4.3664, 2.1361, 2.8443, 2.2700, 2.4994], device='cuda:3'), covar=tensor([0.0866, 0.3040, 0.2012, 0.0360, 0.3456, 0.1960, 0.2844, 0.2784], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0380, 0.0322, 0.0321, 0.0405, 0.0434, 0.0343, 0.0450], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:39:20,429 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 1850, loss[loss=0.1636, simple_loss=0.2434, pruned_loss=0.04188, over 16857.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2726, pruned_loss=0.05394, over 3321842.30 frames. ], batch size: 102, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:40:07,705 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9472, 3.2681, 2.9225, 5.1117, 4.2881, 4.7311, 1.6146, 3.3916], device='cuda:3'), covar=tensor([0.1244, 0.0577, 0.1021, 0.0148, 0.0252, 0.0325, 0.1445, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0156, 0.0179, 0.0137, 0.0197, 0.0213, 0.0178, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 04:40:21,084 INFO [optim.py:368] (3/8) 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 04:40:52,063 INFO [train.py:904] (3/8) Epoch 10, batch 1900, loss[loss=0.2055, simple_loss=0.2837, pruned_loss=0.06369, over 16272.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2723, pruned_loss=0.05354, over 3320714.53 frames. ], batch size: 165, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:40:56,693 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:25,698 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9965, 4.9534, 4.7840, 4.2568, 4.8737, 1.9480, 4.6127, 4.8099], device='cuda:3'), covar=tensor([0.0077, 0.0066, 0.0153, 0.0335, 0.0084, 0.2231, 0.0115, 0.0151], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0117, 0.0168, 0.0159, 0.0137, 0.0182, 0.0155, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:41:33,292 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:41,653 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:48,252 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9788, 4.0631, 2.0034, 4.7324, 3.0140, 4.6545, 2.2249, 3.3489], device='cuda:3'), covar=tensor([0.0200, 0.0335, 0.1616, 0.0126, 0.0700, 0.0297, 0.1464, 0.0546], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0166, 0.0185, 0.0126, 0.0165, 0.0204, 0.0191, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 04:42:02,725 INFO [train.py:904] (3/8) Epoch 10, batch 1950, loss[loss=0.1651, simple_loss=0.2469, pruned_loss=0.04164, over 17200.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2727, pruned_loss=0.0536, over 3310714.61 frames. ], batch size: 43, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:42:20,007 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 04:42:40,591 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.458e+02 3.015e+02 3.561e+02 8.313e+02, threshold=6.031e+02, percent-clipped=4.0 2023-04-29 04:42:43,567 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:04,732 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4391, 4.7746, 4.5552, 4.5563, 4.3240, 4.2820, 4.3187, 4.8498], device='cuda:3'), covar=tensor([0.1074, 0.0870, 0.1011, 0.0693, 0.0739, 0.1243, 0.1006, 0.0879], device='cuda:3'), in_proj_covar=tensor([0.0532, 0.0678, 0.0557, 0.0465, 0.0426, 0.0432, 0.0566, 0.0512], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:43:12,556 INFO [train.py:904] (3/8) Epoch 10, batch 2000, loss[loss=0.2027, simple_loss=0.2664, pruned_loss=0.0695, over 16757.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2719, pruned_loss=0.05284, over 3305945.37 frames. ], batch size: 83, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:43:31,503 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:44,523 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:44:08,367 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:44:20,886 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8555, 2.7566, 2.3326, 2.6659, 3.0884, 2.8728, 3.7046, 3.3863], device='cuda:3'), covar=tensor([0.0059, 0.0264, 0.0356, 0.0287, 0.0184, 0.0272, 0.0165, 0.0170], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0202, 0.0198, 0.0196, 0.0199, 0.0201, 0.0209, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:44:21,491 INFO [train.py:904] (3/8) Epoch 10, batch 2050, loss[loss=0.1548, simple_loss=0.2351, pruned_loss=0.03727, over 16775.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2715, pruned_loss=0.05335, over 3309371.59 frames. ], batch size: 39, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:44:51,420 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 04:45:29,948 INFO [train.py:904] (3/8) Epoch 10, batch 2100, loss[loss=0.1769, simple_loss=0.2577, pruned_loss=0.04805, over 16813.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2728, pruned_loss=0.05426, over 3310288.38 frames. ], batch size: 42, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:45:56,797 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:18,618 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 2150, loss[loss=0.1933, simple_loss=0.2649, pruned_loss=0.06083, over 16840.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2746, pruned_loss=0.05557, over 3298435.64 frames. ], batch size: 96, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:47:18,311 INFO [optim.py:368] (3/8) 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 04:47:23,944 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 04:47:24,632 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:47:39,966 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5328, 5.9466, 5.6622, 5.7544, 5.2307, 5.1498, 5.3665, 6.0805], device='cuda:3'), covar=tensor([0.1159, 0.0871, 0.0999, 0.0602, 0.0867, 0.0648, 0.0967, 0.0749], device='cuda:3'), in_proj_covar=tensor([0.0533, 0.0679, 0.0558, 0.0463, 0.0425, 0.0431, 0.0564, 0.0512], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:47:46,702 INFO [train.py:904] (3/8) Epoch 10, batch 2200, loss[loss=0.2112, simple_loss=0.2796, pruned_loss=0.07144, over 16901.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2744, pruned_loss=0.05572, over 3303137.56 frames. ], batch size: 109, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:47:52,060 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:47:53,567 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 04:47:55,788 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4419, 3.9620, 3.9474, 2.1071, 3.1878, 2.5614, 3.8774, 3.9201], device='cuda:3'), covar=tensor([0.0260, 0.0695, 0.0476, 0.1701, 0.0736, 0.0884, 0.0606, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0142, 0.0156, 0.0141, 0.0134, 0.0124, 0.0135, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 04:48:20,179 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7990, 3.9305, 2.2570, 4.2990, 2.8691, 4.3077, 2.2854, 3.2109], device='cuda:3'), covar=tensor([0.0197, 0.0306, 0.1437, 0.0191, 0.0809, 0.0378, 0.1471, 0.0562], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0166, 0.0186, 0.0127, 0.0167, 0.0207, 0.0194, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 04:48:35,434 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:54,025 INFO [train.py:904] (3/8) Epoch 10, batch 2250, loss[loss=0.1959, simple_loss=0.2779, pruned_loss=0.05699, over 16588.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2748, pruned_loss=0.05531, over 3315792.31 frames. ], batch size: 75, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:48:55,411 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:33,830 INFO [optim.py:368] (3/8) 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 04:49:34,848 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:04,003 INFO [train.py:904] (3/8) Epoch 10, batch 2300, loss[loss=0.1518, simple_loss=0.2424, pruned_loss=0.03058, over 16952.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2737, pruned_loss=0.05413, over 3322643.84 frames. ], batch size: 41, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:50:22,556 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:39,711 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:57,937 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:09,769 INFO [train.py:904] (3/8) Epoch 10, batch 2350, loss[loss=0.1984, simple_loss=0.2745, pruned_loss=0.06111, over 16902.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2731, pruned_loss=0.05402, over 3322590.72 frames. ], batch size: 96, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:51:27,699 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:43,462 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.77 vs. limit=5.0 2023-04-29 04:51:49,909 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.363e+02 2.805e+02 3.305e+02 9.718e+02, threshold=5.610e+02, percent-clipped=1.0 2023-04-29 04:52:03,150 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:17,487 INFO [train.py:904] (3/8) Epoch 10, batch 2400, loss[loss=0.2508, simple_loss=0.3286, pruned_loss=0.08655, over 12208.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2738, pruned_loss=0.05415, over 3318832.14 frames. ], batch size: 246, lr: 6.95e-03, grad_scale: 8.0 2023-04-29 04:52:41,432 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:53:00,979 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8126, 2.1683, 2.2233, 4.5380, 2.1458, 2.7162, 2.3181, 2.4297], device='cuda:3'), covar=tensor([0.0763, 0.3231, 0.2251, 0.0344, 0.3704, 0.2306, 0.2815, 0.3176], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0384, 0.0323, 0.0323, 0.0406, 0.0438, 0.0344, 0.0452], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 04:53:26,813 INFO [train.py:904] (3/8) Epoch 10, batch 2450, loss[loss=0.2324, simple_loss=0.3089, pruned_loss=0.07796, over 12390.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2749, pruned_loss=0.05427, over 3318707.09 frames. ], batch size: 245, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:53:49,967 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:05,084 INFO [optim.py:368] (3/8) 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 04:54:05,481 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 2500, loss[loss=0.192, simple_loss=0.2882, pruned_loss=0.04795, over 17076.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2745, pruned_loss=0.05401, over 3313548.92 frames. ], batch size: 55, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:54:50,185 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 04:55:43,655 INFO [train.py:904] (3/8) Epoch 10, batch 2550, loss[loss=0.231, simple_loss=0.2992, pruned_loss=0.08135, over 16708.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.275, pruned_loss=0.05487, over 3309854.92 frames. ], batch size: 124, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:55:50,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7656, 2.8358, 2.5553, 4.6399, 3.7825, 4.3586, 1.6189, 3.0421], device='cuda:3'), covar=tensor([0.1313, 0.0700, 0.1133, 0.0154, 0.0288, 0.0346, 0.1428, 0.0790], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0155, 0.0178, 0.0138, 0.0198, 0.0212, 0.0175, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 04:56:23,963 INFO [optim.py:368] (3/8) 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 04:56:48,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3533, 5.2166, 5.1676, 4.8431, 4.7560, 5.2409, 5.1658, 4.8036], device='cuda:3'), covar=tensor([0.0533, 0.0462, 0.0232, 0.0252, 0.1046, 0.0329, 0.0226, 0.0656], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0310, 0.0295, 0.0272, 0.0320, 0.0308, 0.0204, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 04:56:52,858 INFO [train.py:904] (3/8) Epoch 10, batch 2600, loss[loss=0.1811, simple_loss=0.2729, pruned_loss=0.04461, over 17043.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2739, pruned_loss=0.05381, over 3320734.02 frames. ], batch size: 53, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:57:09,856 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 04:57:32,301 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:57:42,490 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6136, 3.7472, 3.9629, 2.0419, 4.1671, 4.1950, 3.1907, 3.0551], device='cuda:3'), covar=tensor([0.0759, 0.0186, 0.0174, 0.1099, 0.0058, 0.0131, 0.0386, 0.0409], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0098, 0.0087, 0.0138, 0.0070, 0.0101, 0.0121, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 04:57:56,745 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1728, 3.3982, 3.5960, 3.5791, 3.5580, 3.3609, 3.4095, 3.3940], device='cuda:3'), covar=tensor([0.0475, 0.0622, 0.0519, 0.0540, 0.0564, 0.0482, 0.0882, 0.0573], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0344, 0.0347, 0.0330, 0.0388, 0.0362, 0.0469, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 04:58:03,850 INFO [train.py:904] (3/8) Epoch 10, batch 2650, loss[loss=0.1702, simple_loss=0.2614, pruned_loss=0.03953, over 17227.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2738, pruned_loss=0.05311, over 3323201.88 frames. ], batch size: 44, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:58:42,838 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-04-29 04:58:43,800 INFO [optim.py:368] (3/8) 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 04:59:00,281 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 2700, loss[loss=0.2462, simple_loss=0.3111, pruned_loss=0.0907, over 12373.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.275, pruned_loss=0.0533, over 3317663.15 frames. ], batch size: 246, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 05:00:23,305 INFO [train.py:904] (3/8) Epoch 10, batch 2750, loss[loss=0.202, simple_loss=0.2741, pruned_loss=0.06496, over 16756.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2752, pruned_loss=0.05295, over 3328101.19 frames. ], batch size: 83, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:00:55,433 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:01:01,089 INFO [optim.py:368] (3/8) 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 05:01:29,693 INFO [train.py:904] (3/8) Epoch 10, batch 2800, loss[loss=0.1776, simple_loss=0.2631, pruned_loss=0.04607, over 16830.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2755, pruned_loss=0.05323, over 3329229.29 frames. ], batch size: 42, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:02:39,399 INFO [train.py:904] (3/8) Epoch 10, batch 2850, loss[loss=0.2114, simple_loss=0.2796, pruned_loss=0.07156, over 16398.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.275, pruned_loss=0.0534, over 3322296.79 frames. ], batch size: 146, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:09,016 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:03:18,415 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 05:03:20,127 INFO [optim.py:368] (3/8) 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 05:03:49,047 INFO [train.py:904] (3/8) Epoch 10, batch 2900, loss[loss=0.2088, simple_loss=0.2938, pruned_loss=0.06192, over 17110.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2735, pruned_loss=0.0531, over 3321250.10 frames. ], batch size: 55, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:04:33,844 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4920, 1.6295, 2.0575, 2.4214, 2.4852, 2.4106, 1.5568, 2.6216], device='cuda:3'), covar=tensor([0.0129, 0.0331, 0.0220, 0.0202, 0.0185, 0.0180, 0.0340, 0.0093], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0155, 0.0160, 0.0167, 0.0122, 0.0169, 0.0112], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 05:04:33,854 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:04:58,055 INFO [train.py:904] (3/8) Epoch 10, batch 2950, loss[loss=0.1852, simple_loss=0.2642, pruned_loss=0.05311, over 16793.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2723, pruned_loss=0.05363, over 3330981.07 frames. ], batch size: 124, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:05:39,543 INFO [optim.py:368] (3/8) 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 05:05:49,207 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:03,114 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:08,006 INFO [train.py:904] (3/8) Epoch 10, batch 3000, loss[loss=0.1698, simple_loss=0.2644, pruned_loss=0.03761, over 17140.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2728, pruned_loss=0.05443, over 3327756.42 frames. ], batch size: 48, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:06:08,006 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 05:06:17,141 INFO [train.py:938] (3/8) Epoch 10, validation: loss=0.1426, simple_loss=0.2488, pruned_loss=0.01818, over 944034.00 frames. 2023-04-29 05:06:17,142 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 05:06:40,714 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8406, 5.1641, 4.9001, 4.9274, 4.6784, 4.5966, 4.7102, 5.2458], device='cuda:3'), covar=tensor([0.1058, 0.0822, 0.0973, 0.0689, 0.0782, 0.0887, 0.0950, 0.0872], device='cuda:3'), in_proj_covar=tensor([0.0528, 0.0674, 0.0554, 0.0458, 0.0420, 0.0428, 0.0558, 0.0509], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:07:26,696 INFO [train.py:904] (3/8) Epoch 10, batch 3050, loss[loss=0.1813, simple_loss=0.2746, pruned_loss=0.04404, over 17115.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2731, pruned_loss=0.05426, over 3332271.23 frames. ], batch size: 49, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:07:30,837 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-29 05:07:36,817 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:07:56,510 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9795, 5.4365, 5.6228, 5.3317, 5.3844, 5.9828, 5.6288, 5.3299], device='cuda:3'), covar=tensor([0.0883, 0.1900, 0.1725, 0.2199, 0.2779, 0.1059, 0.1213, 0.2401], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0497, 0.0520, 0.0432, 0.0574, 0.0548, 0.0419, 0.0581], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:07:57,737 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 05:08:10,072 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7311, 2.8137, 2.3200, 4.1318, 3.5456, 4.0731, 1.4988, 2.8469], device='cuda:3'), covar=tensor([0.1208, 0.0569, 0.1105, 0.0140, 0.0208, 0.0372, 0.1315, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0157, 0.0179, 0.0139, 0.0200, 0.0214, 0.0176, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:08:33,247 INFO [train.py:904] (3/8) Epoch 10, batch 3100, loss[loss=0.2058, simple_loss=0.2794, pruned_loss=0.0661, over 15630.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2729, pruned_loss=0.05407, over 3336819.02 frames. ], batch size: 191, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:09:04,482 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:09:43,503 INFO [train.py:904] (3/8) Epoch 10, batch 3150, loss[loss=0.1804, simple_loss=0.2712, pruned_loss=0.04484, over 16619.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2724, pruned_loss=0.05386, over 3340263.50 frames. ], batch size: 62, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:09:48,973 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 2023-04-29 05:09:59,563 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 05:10:23,683 INFO [optim.py:368] (3/8) 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 05:10:28,131 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2711, 5.6632, 5.3755, 5.4514, 5.0520, 4.9031, 5.1668, 5.7817], device='cuda:3'), covar=tensor([0.0991, 0.0776, 0.1095, 0.0674, 0.0813, 0.0706, 0.0981, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0540, 0.0686, 0.0567, 0.0469, 0.0429, 0.0435, 0.0569, 0.0519], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:10:36,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-04-29 05:10:52,169 INFO [train.py:904] (3/8) Epoch 10, batch 3200, loss[loss=0.1969, simple_loss=0.2701, pruned_loss=0.06191, over 16749.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2713, pruned_loss=0.05346, over 3334525.65 frames. ], batch size: 124, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:10:55,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7476, 4.9201, 5.0830, 4.9542, 4.9583, 5.5775, 5.1345, 4.8727], device='cuda:3'), covar=tensor([0.1108, 0.1857, 0.1689, 0.1877, 0.2814, 0.0911, 0.1202, 0.2030], device='cuda:3'), in_proj_covar=tensor([0.0346, 0.0498, 0.0522, 0.0431, 0.0573, 0.0548, 0.0423, 0.0579], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:11:32,214 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:12:04,565 INFO [train.py:904] (3/8) Epoch 10, batch 3250, loss[loss=0.1701, simple_loss=0.2472, pruned_loss=0.04655, over 15858.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2702, pruned_loss=0.05304, over 3335918.77 frames. ], batch size: 35, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:12:44,900 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:13:14,620 INFO [train.py:904] (3/8) Epoch 10, batch 3300, loss[loss=0.171, simple_loss=0.2583, pruned_loss=0.04186, over 15957.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2716, pruned_loss=0.05363, over 3331027.87 frames. ], batch size: 35, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:02,319 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:14:20,294 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8611, 3.9756, 2.3906, 4.5781, 2.8090, 4.5814, 2.1869, 3.1948], device='cuda:3'), covar=tensor([0.0218, 0.0326, 0.1425, 0.0174, 0.0831, 0.0411, 0.1548, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0132, 0.0168, 0.0210, 0.0196, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 05:14:24,568 INFO [train.py:904] (3/8) Epoch 10, batch 3350, loss[loss=0.1545, simple_loss=0.2376, pruned_loss=0.03567, over 15715.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2717, pruned_loss=0.05315, over 3337087.06 frames. ], batch size: 35, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:28,543 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:15:04,362 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7445, 3.8961, 2.4645, 4.2972, 2.7539, 4.4037, 2.3124, 3.1529], device='cuda:3'), covar=tensor([0.0221, 0.0319, 0.1277, 0.0241, 0.0826, 0.0329, 0.1326, 0.0599], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0167, 0.0187, 0.0131, 0.0168, 0.0210, 0.0195, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 05:15:05,008 INFO [optim.py:368] (3/8) 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 05:15:35,797 INFO [train.py:904] (3/8) Epoch 10, batch 3400, loss[loss=0.1756, simple_loss=0.2647, pruned_loss=0.04319, over 17198.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2723, pruned_loss=0.05315, over 3343840.75 frames. ], batch size: 45, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:15:55,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1282, 2.5731, 2.1436, 2.3852, 2.9771, 2.7591, 3.1606, 3.1913], device='cuda:3'), covar=tensor([0.0124, 0.0254, 0.0348, 0.0315, 0.0146, 0.0227, 0.0165, 0.0150], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0201, 0.0198, 0.0197, 0.0200, 0.0203, 0.0212, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:16:10,944 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8144, 3.0626, 2.6527, 4.5525, 3.8615, 4.3529, 1.6798, 3.0794], device='cuda:3'), covar=tensor([0.1171, 0.0555, 0.0935, 0.0116, 0.0187, 0.0310, 0.1243, 0.0646], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0158, 0.0178, 0.0141, 0.0201, 0.0214, 0.0176, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:16:44,906 INFO [train.py:904] (3/8) Epoch 10, batch 3450, loss[loss=0.2016, simple_loss=0.2732, pruned_loss=0.06494, over 16425.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2707, pruned_loss=0.05272, over 3338708.23 frames. ], batch size: 146, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:17:22,644 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 05:17:26,304 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.564e+02 2.957e+02 3.532e+02 6.981e+02, threshold=5.915e+02, percent-clipped=3.0 2023-04-29 05:17:56,543 INFO [train.py:904] (3/8) Epoch 10, batch 3500, loss[loss=0.2196, simple_loss=0.2957, pruned_loss=0.07173, over 16232.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2703, pruned_loss=0.05265, over 3342575.32 frames. ], batch size: 165, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:18:35,694 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:06,953 INFO [train.py:904] (3/8) Epoch 10, batch 3550, loss[loss=0.1504, simple_loss=0.2375, pruned_loss=0.0316, over 16864.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2683, pruned_loss=0.05152, over 3344867.48 frames. ], batch size: 42, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:19:41,448 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:47,556 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 3600, loss[loss=0.1801, simple_loss=0.2748, pruned_loss=0.04273, over 17125.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2669, pruned_loss=0.05093, over 3338386.13 frames. ], batch size: 48, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:28,701 INFO [train.py:904] (3/8) Epoch 10, batch 3650, loss[loss=0.1498, simple_loss=0.2337, pruned_loss=0.03299, over 16840.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2661, pruned_loss=0.05148, over 3316595.44 frames. ], batch size: 42, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:32,963 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:22:10,218 INFO [optim.py:368] (3/8) 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 05:22:43,025 INFO [train.py:904] (3/8) Epoch 10, batch 3700, loss[loss=0.1865, simple_loss=0.2565, pruned_loss=0.05829, over 11076.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2653, pruned_loss=0.05342, over 3280571.41 frames. ], batch size: 247, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:22:43,300 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:22:43,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5321, 3.6334, 2.2193, 3.8439, 2.7550, 3.7908, 2.1721, 2.9154], device='cuda:3'), covar=tensor([0.0219, 0.0364, 0.1292, 0.0190, 0.0662, 0.0609, 0.1357, 0.0610], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0166, 0.0186, 0.0131, 0.0166, 0.0210, 0.0193, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 05:22:54,840 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 05:23:01,415 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-29 05:23:18,062 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 05:23:56,092 INFO [train.py:904] (3/8) Epoch 10, batch 3750, loss[loss=0.1964, simple_loss=0.2653, pruned_loss=0.06371, over 16420.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2666, pruned_loss=0.05515, over 3265549.27 frames. ], batch size: 75, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:24:00,750 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 05:24:59,655 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6116, 2.6962, 2.1718, 3.7826, 3.0612, 3.8417, 1.3622, 2.6634], device='cuda:3'), covar=tensor([0.1339, 0.0592, 0.1219, 0.0160, 0.0190, 0.0376, 0.1476, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0158, 0.0179, 0.0142, 0.0203, 0.0213, 0.0177, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:25:07,896 INFO [train.py:904] (3/8) Epoch 10, batch 3800, loss[loss=0.1834, simple_loss=0.2562, pruned_loss=0.05536, over 16444.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2672, pruned_loss=0.05659, over 3269297.54 frames. ], batch size: 75, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:25:27,607 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7927, 4.0050, 3.0338, 2.2571, 2.5915, 2.2763, 4.0269, 3.5710], device='cuda:3'), covar=tensor([0.2212, 0.0470, 0.1456, 0.2518, 0.2423, 0.1834, 0.0464, 0.1128], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0260, 0.0283, 0.0278, 0.0288, 0.0222, 0.0271, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:25:28,633 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:25:56,961 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 05:26:14,158 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 05:26:20,912 INFO [train.py:904] (3/8) Epoch 10, batch 3850, loss[loss=0.1894, simple_loss=0.2573, pruned_loss=0.06077, over 16774.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2682, pruned_loss=0.05776, over 3252826.62 frames. ], batch size: 124, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:26:46,151 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3280, 2.0234, 2.1410, 4.0046, 2.0390, 2.4844, 2.0740, 2.1988], device='cuda:3'), covar=tensor([0.0918, 0.3180, 0.2185, 0.0408, 0.3352, 0.2154, 0.3242, 0.2711], device='cuda:3'), in_proj_covar=tensor([0.0363, 0.0390, 0.0325, 0.0326, 0.0409, 0.0447, 0.0353, 0.0462], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:26:47,457 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 05:27:00,986 INFO [optim.py:368] (3/8) 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 05:27:31,950 INFO [train.py:904] (3/8) Epoch 10, batch 3900, loss[loss=0.1742, simple_loss=0.2504, pruned_loss=0.04898, over 16791.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2683, pruned_loss=0.05854, over 3259533.26 frames. ], batch size: 102, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:27:43,914 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:27:56,168 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:28:45,288 INFO [train.py:904] (3/8) Epoch 10, batch 3950, loss[loss=0.1879, simple_loss=0.2598, pruned_loss=0.05803, over 16860.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2668, pruned_loss=0.0587, over 3276613.51 frames. ], batch size: 116, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:29:12,969 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:19,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4233, 3.5502, 3.8861, 2.5544, 3.5005, 3.8425, 3.6747, 2.2194], device='cuda:3'), covar=tensor([0.0340, 0.0084, 0.0029, 0.0265, 0.0058, 0.0070, 0.0048, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0067, 0.0067, 0.0121, 0.0076, 0.0083, 0.0074, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:29:23,165 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:23,356 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 05:29:25,657 INFO [optim.py:368] (3/8) 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 05:29:38,068 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:56,313 INFO [train.py:904] (3/8) Epoch 10, batch 4000, loss[loss=0.1965, simple_loss=0.2712, pruned_loss=0.06094, over 16750.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2666, pruned_loss=0.05868, over 3284580.32 frames. ], batch size: 134, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:30:23,519 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6060, 4.6038, 4.6260, 2.8702, 3.9539, 4.3131, 3.9878, 2.3272], device='cuda:3'), covar=tensor([0.0420, 0.0018, 0.0026, 0.0324, 0.0070, 0.0108, 0.0073, 0.0406], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0067, 0.0067, 0.0121, 0.0076, 0.0083, 0.0074, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:31:05,702 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 4050, loss[loss=0.1823, simple_loss=0.2604, pruned_loss=0.05206, over 16192.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2664, pruned_loss=0.05721, over 3285462.72 frames. ], batch size: 35, lr: 6.89e-03, grad_scale: 16.0 2023-04-29 05:31:42,362 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6129, 5.5987, 5.3699, 4.9071, 5.5534, 2.3252, 5.2481, 5.2521], device='cuda:3'), covar=tensor([0.0042, 0.0028, 0.0094, 0.0210, 0.0041, 0.1862, 0.0077, 0.0128], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0116, 0.0164, 0.0157, 0.0134, 0.0174, 0.0153, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:31:49,145 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 4100, loss[loss=0.1982, simple_loss=0.2835, pruned_loss=0.05648, over 16816.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2677, pruned_loss=0.05645, over 3284974.23 frames. ], batch size: 83, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:32:34,841 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:33:03,689 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6272, 2.5552, 2.2558, 3.3920, 2.5089, 3.6064, 1.3558, 2.6182], device='cuda:3'), covar=tensor([0.1322, 0.0628, 0.1175, 0.0149, 0.0188, 0.0369, 0.1510, 0.0843], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0157, 0.0179, 0.0141, 0.0204, 0.0210, 0.0176, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:33:33,944 INFO [train.py:904] (3/8) Epoch 10, batch 4150, loss[loss=0.2028, simple_loss=0.2897, pruned_loss=0.05797, over 16377.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.275, pruned_loss=0.05913, over 3264210.73 frames. ], batch size: 35, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:34:17,107 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.709e+02 3.150e+02 3.936e+02 7.135e+02, threshold=6.300e+02, percent-clipped=10.0 2023-04-29 05:34:49,630 INFO [train.py:904] (3/8) Epoch 10, batch 4200, loss[loss=0.2423, simple_loss=0.3296, pruned_loss=0.07755, over 16682.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2825, pruned_loss=0.06173, over 3221265.68 frames. ], batch size: 134, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:34:56,959 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9326, 4.9116, 5.3750, 5.2753, 5.3756, 4.8420, 4.9457, 4.5236], device='cuda:3'), covar=tensor([0.0258, 0.0318, 0.0262, 0.0366, 0.0336, 0.0307, 0.0781, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0330, 0.0332, 0.0317, 0.0377, 0.0353, 0.0455, 0.0280], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 05:36:04,061 INFO [train.py:904] (3/8) Epoch 10, batch 4250, loss[loss=0.2135, simple_loss=0.301, pruned_loss=0.06298, over 17002.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2863, pruned_loss=0.06199, over 3196106.78 frames. ], batch size: 55, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:36:24,739 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:26,121 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:27,842 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3845, 3.9472, 3.9242, 2.6553, 3.5721, 3.7737, 3.7078, 2.3294], device='cuda:3'), covar=tensor([0.0355, 0.0024, 0.0033, 0.0256, 0.0053, 0.0073, 0.0043, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0066, 0.0067, 0.0120, 0.0076, 0.0083, 0.0073, 0.0114], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:36:37,906 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:49,133 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 4300, loss[loss=0.2015, simple_loss=0.2918, pruned_loss=0.05562, over 17226.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2873, pruned_loss=0.0609, over 3188727.58 frames. ], batch size: 52, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:37:51,594 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7132, 2.9056, 2.8051, 4.9741, 3.9998, 4.4132, 1.5045, 3.1536], device='cuda:3'), covar=tensor([0.1309, 0.0660, 0.1088, 0.0135, 0.0347, 0.0298, 0.1528, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0157, 0.0177, 0.0139, 0.0201, 0.0207, 0.0176, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:37:59,719 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:24,131 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:33,663 INFO [train.py:904] (3/8) Epoch 10, batch 4350, loss[loss=0.2186, simple_loss=0.3054, pruned_loss=0.06592, over 16622.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2906, pruned_loss=0.06185, over 3191606.54 frames. ], batch size: 57, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:38:48,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0723, 5.3454, 5.0739, 5.1155, 4.8034, 4.6704, 4.8033, 5.4694], device='cuda:3'), covar=tensor([0.0749, 0.0695, 0.0963, 0.0690, 0.0667, 0.0772, 0.0869, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0502, 0.0642, 0.0530, 0.0440, 0.0401, 0.0415, 0.0530, 0.0489], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:39:18,687 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.523e+02 3.025e+02 3.679e+02 8.417e+02, threshold=6.050e+02, percent-clipped=3.0 2023-04-29 05:39:49,467 INFO [train.py:904] (3/8) Epoch 10, batch 4400, loss[loss=0.2068, simple_loss=0.296, pruned_loss=0.0588, over 16733.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2925, pruned_loss=0.06289, over 3186798.64 frames. ], batch size: 83, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:39:52,743 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-29 05:40:02,622 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:40:06,806 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 05:41:01,592 INFO [train.py:904] (3/8) Epoch 10, batch 4450, loss[loss=0.2153, simple_loss=0.3059, pruned_loss=0.06235, over 16892.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2956, pruned_loss=0.06373, over 3196639.93 frames. ], batch size: 116, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:41:12,948 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:17,822 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:41:46,152 INFO [optim.py:368] (3/8) 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 05:42:08,791 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:42:15,000 INFO [train.py:904] (3/8) Epoch 10, batch 4500, loss[loss=0.2035, simple_loss=0.2879, pruned_loss=0.05952, over 16696.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2962, pruned_loss=0.06404, over 3210040.73 frames. ], batch size: 134, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:42:36,186 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5719, 4.3900, 4.5676, 4.7292, 4.8866, 4.4686, 4.8473, 4.8922], device='cuda:3'), covar=tensor([0.1273, 0.0923, 0.1252, 0.0517, 0.0352, 0.0694, 0.0443, 0.0412], device='cuda:3'), in_proj_covar=tensor([0.0501, 0.0623, 0.0762, 0.0632, 0.0471, 0.0484, 0.0496, 0.0561], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:42:46,003 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:42:59,246 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 05:43:13,126 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6630, 3.0502, 2.7055, 4.8559, 3.5893, 4.3408, 1.5217, 3.0179], device='cuda:3'), covar=tensor([0.1429, 0.0685, 0.1121, 0.0084, 0.0313, 0.0276, 0.1634, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0157, 0.0178, 0.0138, 0.0202, 0.0206, 0.0177, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:43:27,209 INFO [train.py:904] (3/8) Epoch 10, batch 4550, loss[loss=0.2408, simple_loss=0.3157, pruned_loss=0.08296, over 16707.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2968, pruned_loss=0.06453, over 3226960.88 frames. ], batch size: 57, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:43:35,526 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:43:47,801 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:43:59,525 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:44:10,341 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 4600, loss[loss=0.1903, simple_loss=0.2817, pruned_loss=0.04948, over 15460.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2972, pruned_loss=0.06432, over 3221355.21 frames. ], batch size: 191, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:44:57,859 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:09,542 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:11,438 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:21,027 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:26,326 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0531, 3.0630, 1.6280, 3.2546, 2.2404, 3.3173, 1.8901, 2.5016], device='cuda:3'), covar=tensor([0.0216, 0.0344, 0.1611, 0.0118, 0.0839, 0.0358, 0.1417, 0.0656], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0163, 0.0185, 0.0120, 0.0165, 0.0201, 0.0189, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 05:45:43,254 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 4650, loss[loss=0.2169, simple_loss=0.2981, pruned_loss=0.06785, over 16675.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.296, pruned_loss=0.06419, over 3207078.76 frames. ], batch size: 134, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:46:07,671 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 05:46:12,176 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5170, 4.7865, 5.0043, 4.8139, 4.8516, 5.4216, 4.8932, 4.6315], device='cuda:3'), covar=tensor([0.1097, 0.1502, 0.1429, 0.1688, 0.2317, 0.0929, 0.1229, 0.2158], device='cuda:3'), in_proj_covar=tensor([0.0336, 0.0467, 0.0495, 0.0406, 0.0539, 0.0521, 0.0401, 0.0553], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:46:33,505 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 05:46:40,627 INFO [optim.py:368] (3/8) 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 05:46:50,643 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3425, 2.3882, 1.8560, 2.1756, 2.7543, 2.4224, 3.0910, 3.0305], device='cuda:3'), covar=tensor([0.0055, 0.0280, 0.0390, 0.0313, 0.0160, 0.0279, 0.0132, 0.0163], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0194, 0.0193, 0.0190, 0.0194, 0.0197, 0.0198, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:46:55,295 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:47:10,297 INFO [train.py:904] (3/8) Epoch 10, batch 4700, loss[loss=0.2189, simple_loss=0.302, pruned_loss=0.0679, over 15458.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2932, pruned_loss=0.06279, over 3199321.78 frames. ], batch size: 191, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:47:52,299 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:48:23,998 INFO [train.py:904] (3/8) Epoch 10, batch 4750, loss[loss=0.2002, simple_loss=0.2883, pruned_loss=0.05606, over 17041.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.289, pruned_loss=0.06076, over 3206069.98 frames. ], batch size: 50, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:48:33,785 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7872, 3.7999, 1.8940, 4.3671, 2.8217, 4.2715, 2.2508, 2.7565], device='cuda:3'), covar=tensor([0.0190, 0.0295, 0.1734, 0.0086, 0.0748, 0.0295, 0.1495, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0163, 0.0184, 0.0120, 0.0163, 0.0200, 0.0189, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 05:49:08,953 INFO [optim.py:368] (3/8) 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 05:49:22,629 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 4800, loss[loss=0.2038, simple_loss=0.2928, pruned_loss=0.05743, over 16281.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2854, pruned_loss=0.05864, over 3205967.04 frames. ], batch size: 165, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:58,747 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:50:03,055 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:50:39,517 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3612, 4.4319, 4.2358, 3.9669, 3.7745, 4.3146, 4.1122, 3.9541], device='cuda:3'), covar=tensor([0.0616, 0.0283, 0.0274, 0.0267, 0.1097, 0.0383, 0.0525, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0285, 0.0273, 0.0250, 0.0296, 0.0284, 0.0185, 0.0316], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:50:54,658 INFO [train.py:904] (3/8) Epoch 10, batch 4850, loss[loss=0.2413, simple_loss=0.3225, pruned_loss=0.08005, over 12129.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2869, pruned_loss=0.05858, over 3190366.36 frames. ], batch size: 248, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:50:56,266 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:51:32,266 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 05:52:10,085 INFO [train.py:904] (3/8) Epoch 10, batch 4900, loss[loss=0.1802, simple_loss=0.2755, pruned_loss=0.04247, over 16207.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2862, pruned_loss=0.0572, over 3178557.85 frames. ], batch size: 165, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:52:32,304 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5681, 2.4128, 1.9818, 3.5914, 2.2714, 3.6612, 1.3794, 2.5864], device='cuda:3'), covar=tensor([0.1448, 0.0822, 0.1449, 0.0154, 0.0201, 0.0410, 0.1683, 0.0959], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0153, 0.0175, 0.0134, 0.0196, 0.0202, 0.0175, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 05:52:42,998 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:24,548 INFO [train.py:904] (3/8) Epoch 10, batch 4950, loss[loss=0.1861, simple_loss=0.2774, pruned_loss=0.04737, over 17121.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2858, pruned_loss=0.05674, over 3180462.36 frames. ], batch size: 47, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:53:44,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7807, 4.5322, 4.7766, 4.9634, 5.1376, 4.6494, 5.1142, 5.1055], device='cuda:3'), covar=tensor([0.1239, 0.0983, 0.1272, 0.0565, 0.0372, 0.0597, 0.0415, 0.0445], device='cuda:3'), in_proj_covar=tensor([0.0498, 0.0619, 0.0755, 0.0630, 0.0469, 0.0486, 0.0492, 0.0559], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:53:52,126 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:56,327 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:03,909 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7247, 3.8257, 2.9514, 2.3003, 2.8954, 2.5032, 3.9510, 3.5734], device='cuda:3'), covar=tensor([0.2484, 0.0755, 0.1563, 0.2071, 0.1954, 0.1517, 0.0580, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0256, 0.0278, 0.0273, 0.0281, 0.0216, 0.0266, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:54:05,870 INFO [optim.py:368] (3/8) 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 05:54:12,439 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:32,657 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3116, 1.6004, 2.5998, 3.0968, 3.0720, 3.6261, 1.6815, 3.5186], device='cuda:3'), covar=tensor([0.0101, 0.0414, 0.0206, 0.0188, 0.0150, 0.0068, 0.0502, 0.0061], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0168, 0.0153, 0.0157, 0.0164, 0.0119, 0.0171, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 05:54:33,260 INFO [train.py:904] (3/8) Epoch 10, batch 5000, loss[loss=0.2079, simple_loss=0.2924, pruned_loss=0.06171, over 17154.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2872, pruned_loss=0.05686, over 3187239.89 frames. ], batch size: 46, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:54:38,346 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7793, 3.4665, 3.3735, 2.0767, 3.1172, 3.3745, 3.2110, 1.7852], device='cuda:3'), covar=tensor([0.0466, 0.0025, 0.0033, 0.0333, 0.0063, 0.0058, 0.0058, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0065, 0.0067, 0.0122, 0.0076, 0.0082, 0.0073, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 05:55:17,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2467, 2.0010, 2.0995, 3.9346, 1.8416, 2.4625, 2.1101, 2.1942], device='cuda:3'), covar=tensor([0.0974, 0.3140, 0.2146, 0.0408, 0.3886, 0.2248, 0.3044, 0.2957], device='cuda:3'), in_proj_covar=tensor([0.0356, 0.0385, 0.0319, 0.0319, 0.0405, 0.0438, 0.0346, 0.0451], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 05:55:20,963 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:28,000 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:44,798 INFO [train.py:904] (3/8) Epoch 10, batch 5050, loss[loss=0.2145, simple_loss=0.3, pruned_loss=0.06452, over 16927.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2887, pruned_loss=0.05731, over 3190738.24 frames. ], batch size: 109, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:56:27,889 INFO [optim.py:368] (3/8) 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 05:56:33,454 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:55,115 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:55,805 INFO [train.py:904] (3/8) Epoch 10, batch 5100, loss[loss=0.16, simple_loss=0.2459, pruned_loss=0.0371, over 16656.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2871, pruned_loss=0.05622, over 3205244.62 frames. ], batch size: 57, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:57:20,206 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:57:58,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0655, 4.9661, 5.4702, 5.4352, 5.4698, 5.1003, 5.0544, 4.8279], device='cuda:3'), covar=tensor([0.0230, 0.0474, 0.0270, 0.0286, 0.0298, 0.0266, 0.0755, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0319, 0.0322, 0.0327, 0.0310, 0.0373, 0.0347, 0.0446, 0.0277], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 05:58:08,611 INFO [train.py:904] (3/8) Epoch 10, batch 5150, loss[loss=0.1935, simple_loss=0.2838, pruned_loss=0.05156, over 16571.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2862, pruned_loss=0.05512, over 3206709.98 frames. ], batch size: 68, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:58:11,286 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:29,578 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.088e+02 2.479e+02 2.937e+02 7.130e+02, threshold=4.958e+02, percent-clipped=1.0 2023-04-29 05:59:21,921 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:59:22,748 INFO [train.py:904] (3/8) Epoch 10, batch 5200, loss[loss=0.1997, simple_loss=0.2774, pruned_loss=0.06102, over 16421.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2849, pruned_loss=0.05497, over 3213567.15 frames. ], batch size: 68, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:59:40,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2450, 3.7156, 3.8782, 1.5309, 4.0489, 4.1225, 2.9511, 2.6471], device='cuda:3'), covar=tensor([0.1147, 0.0136, 0.0116, 0.1452, 0.0058, 0.0069, 0.0347, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0099, 0.0085, 0.0138, 0.0069, 0.0096, 0.0120, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 05:59:55,691 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 06:00:35,342 INFO [train.py:904] (3/8) Epoch 10, batch 5250, loss[loss=0.1971, simple_loss=0.2817, pruned_loss=0.05626, over 16689.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2821, pruned_loss=0.05437, over 3221587.70 frames. ], batch size: 134, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:00:38,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5735, 2.8402, 2.4138, 4.0845, 2.9558, 4.0013, 1.5563, 2.8695], device='cuda:3'), covar=tensor([0.1336, 0.0598, 0.1120, 0.0088, 0.0194, 0.0332, 0.1454, 0.0792], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0154, 0.0175, 0.0134, 0.0197, 0.0203, 0.0175, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 06:01:21,023 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.301e+02 2.656e+02 3.144e+02 5.435e+02, threshold=5.311e+02, percent-clipped=2.0 2023-04-29 06:01:26,351 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:01:48,984 INFO [train.py:904] (3/8) Epoch 10, batch 5300, loss[loss=0.2072, simple_loss=0.2798, pruned_loss=0.06726, over 11930.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2794, pruned_loss=0.05379, over 3208003.38 frames. ], batch size: 246, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:02:30,150 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:02:36,966 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 5350, loss[loss=0.1939, simple_loss=0.2821, pruned_loss=0.05287, over 17128.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2769, pruned_loss=0.05247, over 3222421.60 frames. ], batch size: 47, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:03:21,040 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 06:03:21,072 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 06:03:22,335 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1880, 4.9966, 5.2067, 5.4321, 5.6042, 4.9585, 5.5424, 5.5100], device='cuda:3'), covar=tensor([0.1379, 0.1020, 0.1425, 0.0527, 0.0404, 0.0639, 0.0393, 0.0517], device='cuda:3'), in_proj_covar=tensor([0.0503, 0.0625, 0.0771, 0.0639, 0.0477, 0.0489, 0.0497, 0.0567], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:03:48,672 INFO [optim.py:368] (3/8) 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 06:03:52,800 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:08,473 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 5400, loss[loss=0.1925, simple_loss=0.2774, pruned_loss=0.05379, over 16511.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2798, pruned_loss=0.05333, over 3211383.57 frames. ], batch size: 75, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:04:52,055 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:02,543 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:31,404 INFO [train.py:904] (3/8) Epoch 10, batch 5450, loss[loss=0.2521, simple_loss=0.3294, pruned_loss=0.08741, over 16857.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2831, pruned_loss=0.05524, over 3193239.33 frames. ], batch size: 96, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:05:41,690 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 06:05:50,109 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 06:06:02,205 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 06:06:27,105 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:49,549 INFO [train.py:904] (3/8) Epoch 10, batch 5500, loss[loss=0.2889, simple_loss=0.3454, pruned_loss=0.1162, over 11393.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2916, pruned_loss=0.06121, over 3151588.94 frames. ], batch size: 247, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:07:17,927 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:07:25,362 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5734, 2.7580, 2.2067, 4.2050, 3.0983, 4.0188, 1.4218, 2.8785], device='cuda:3'), covar=tensor([0.1352, 0.0671, 0.1334, 0.0193, 0.0370, 0.0388, 0.1555, 0.0862], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0157, 0.0179, 0.0137, 0.0201, 0.0207, 0.0179, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 06:08:08,502 INFO [train.py:904] (3/8) Epoch 10, batch 5550, loss[loss=0.3168, simple_loss=0.3636, pruned_loss=0.1349, over 11171.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2994, pruned_loss=0.06669, over 3141686.23 frames. ], batch size: 248, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:09:01,353 INFO [optim.py:368] (3/8) 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 06:09:28,004 INFO [train.py:904] (3/8) Epoch 10, batch 5600, loss[loss=0.3157, simple_loss=0.3537, pruned_loss=0.1388, over 10764.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3055, pruned_loss=0.07239, over 3106398.64 frames. ], batch size: 248, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:09:37,431 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 06:10:15,968 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:51,174 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 5650, loss[loss=0.2621, simple_loss=0.3354, pruned_loss=0.09437, over 15377.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3112, pruned_loss=0.07696, over 3082879.10 frames. ], batch size: 190, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:11:34,446 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:11:43,490 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 06:11:43,932 INFO [optim.py:368] (3/8) 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 06:12:02,708 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:12:11,124 INFO [train.py:904] (3/8) Epoch 10, batch 5700, loss[loss=0.2397, simple_loss=0.3205, pruned_loss=0.07941, over 16445.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3123, pruned_loss=0.07843, over 3077372.51 frames. ], batch size: 146, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:12:25,266 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:12:37,592 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0312, 4.0053, 3.9080, 3.2614, 3.9458, 1.7938, 3.7297, 3.5749], device='cuda:3'), covar=tensor([0.0095, 0.0079, 0.0136, 0.0292, 0.0082, 0.2360, 0.0113, 0.0174], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0110, 0.0156, 0.0152, 0.0127, 0.0170, 0.0143, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:13:10,892 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8650, 4.8250, 4.7158, 4.4887, 4.4049, 4.7216, 4.6378, 4.4013], device='cuda:3'), covar=tensor([0.0512, 0.0347, 0.0223, 0.0227, 0.0815, 0.0394, 0.0308, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0236, 0.0292, 0.0277, 0.0254, 0.0300, 0.0292, 0.0189, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:13:11,333 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-29 06:13:16,961 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:29,512 INFO [train.py:904] (3/8) Epoch 10, batch 5750, loss[loss=0.2498, simple_loss=0.3298, pruned_loss=0.08492, over 15398.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3154, pruned_loss=0.08006, over 3072645.77 frames. ], batch size: 190, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:14:17,784 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:14:22,108 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 5800, loss[loss=0.2389, simple_loss=0.3198, pruned_loss=0.07899, over 12146.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3148, pruned_loss=0.07822, over 3083109.41 frames. ], batch size: 248, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:14:55,197 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-29 06:15:35,024 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6063, 4.8017, 4.9680, 4.8215, 4.7651, 5.3524, 4.9399, 4.7215], device='cuda:3'), covar=tensor([0.1036, 0.1592, 0.1528, 0.1677, 0.2411, 0.0955, 0.1349, 0.2303], device='cuda:3'), in_proj_covar=tensor([0.0342, 0.0472, 0.0497, 0.0410, 0.0548, 0.0535, 0.0407, 0.0562], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 06:16:07,864 INFO [train.py:904] (3/8) Epoch 10, batch 5850, loss[loss=0.241, simple_loss=0.3169, pruned_loss=0.0826, over 16255.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3124, pruned_loss=0.07625, over 3088453.50 frames. ], batch size: 165, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:00,834 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 2.965e+02 3.707e+02 4.622e+02 9.015e+02, threshold=7.415e+02, percent-clipped=1.0 2023-04-29 06:17:14,101 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 06:17:28,941 INFO [train.py:904] (3/8) Epoch 10, batch 5900, loss[loss=0.2245, simple_loss=0.2998, pruned_loss=0.07458, over 15392.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3126, pruned_loss=0.07677, over 3073065.88 frames. ], batch size: 191, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:37,286 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7240, 5.0306, 5.2784, 4.9819, 4.9857, 5.6320, 5.1226, 4.9363], device='cuda:3'), covar=tensor([0.0970, 0.1695, 0.1673, 0.1615, 0.2523, 0.0930, 0.1291, 0.2205], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0476, 0.0502, 0.0414, 0.0552, 0.0539, 0.0411, 0.0566], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 06:17:45,027 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:18:49,538 INFO [train.py:904] (3/8) Epoch 10, batch 5950, loss[loss=0.2229, simple_loss=0.3098, pruned_loss=0.06801, over 16679.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3124, pruned_loss=0.07502, over 3072243.57 frames. ], batch size: 134, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:19:03,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0635, 4.0587, 4.5680, 4.5426, 4.5015, 4.1933, 4.2388, 4.0916], device='cuda:3'), covar=tensor([0.0324, 0.0539, 0.0341, 0.0362, 0.0458, 0.0393, 0.0847, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0319, 0.0323, 0.0310, 0.0372, 0.0346, 0.0442, 0.0278], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 06:19:14,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9757, 2.9885, 3.0951, 1.6759, 3.3173, 3.3842, 2.6276, 2.5606], device='cuda:3'), covar=tensor([0.0851, 0.0215, 0.0183, 0.1180, 0.0064, 0.0116, 0.0391, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0098, 0.0086, 0.0137, 0.0068, 0.0096, 0.0120, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 06:19:20,386 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:19:31,588 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 06:19:41,580 INFO [optim.py:368] (3/8) 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 06:19:52,693 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:01,356 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 6000, loss[loss=0.2325, simple_loss=0.3149, pruned_loss=0.07501, over 16351.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3112, pruned_loss=0.0744, over 3081392.74 frames. ], batch size: 146, lr: 6.82e-03, grad_scale: 4.0 2023-04-29 06:20:09,122 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 06:20:23,729 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 06:20:30,440 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:58,006 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:21:15,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6328, 4.8203, 5.0163, 4.8021, 4.9223, 5.4592, 4.9295, 4.7878], device='cuda:3'), covar=tensor([0.1154, 0.1887, 0.1997, 0.1841, 0.2620, 0.1031, 0.1528, 0.2235], device='cuda:3'), in_proj_covar=tensor([0.0343, 0.0474, 0.0502, 0.0411, 0.0547, 0.0536, 0.0410, 0.0562], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 06:21:42,480 INFO [train.py:904] (3/8) Epoch 10, batch 6050, loss[loss=0.2242, simple_loss=0.3175, pruned_loss=0.06546, over 16849.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3085, pruned_loss=0.07288, over 3089433.65 frames. ], batch size: 102, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:21:45,496 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:21:52,223 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:30,433 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:34,965 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 10, batch 6100, loss[loss=0.2304, simple_loss=0.3116, pruned_loss=0.07466, over 15385.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3081, pruned_loss=0.07224, over 3097357.06 frames. ], batch size: 190, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:23:22,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5295, 3.1559, 2.8070, 1.7470, 2.5684, 2.0908, 3.0339, 3.2056], device='cuda:3'), covar=tensor([0.0331, 0.0567, 0.0726, 0.1935, 0.0867, 0.0960, 0.0769, 0.0778], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0139, 0.0154, 0.0140, 0.0133, 0.0123, 0.0134, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 06:23:41,517 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8269, 4.1409, 3.2672, 2.2910, 3.0370, 2.5698, 4.4000, 3.8497], device='cuda:3'), covar=tensor([0.2376, 0.0616, 0.1345, 0.2054, 0.2026, 0.1492, 0.0386, 0.0846], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0256, 0.0281, 0.0277, 0.0282, 0.0218, 0.0266, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:23:48,674 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:24:23,701 INFO [train.py:904] (3/8) Epoch 10, batch 6150, loss[loss=0.2283, simple_loss=0.3027, pruned_loss=0.0769, over 16192.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3062, pruned_loss=0.07126, over 3106343.26 frames. ], batch size: 165, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:17,520 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.229e+02 3.939e+02 5.020e+02 8.476e+02, threshold=7.879e+02, percent-clipped=2.0 2023-04-29 06:25:41,190 INFO [train.py:904] (3/8) Epoch 10, batch 6200, loss[loss=0.2108, simple_loss=0.2891, pruned_loss=0.06626, over 16696.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3044, pruned_loss=0.07097, over 3105634.53 frames. ], batch size: 62, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:46,650 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:26:14,433 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1413, 3.8391, 3.8131, 2.6667, 3.4724, 3.8309, 3.6369, 1.9238], device='cuda:3'), covar=tensor([0.0417, 0.0027, 0.0035, 0.0264, 0.0062, 0.0071, 0.0049, 0.0380], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0066, 0.0068, 0.0126, 0.0076, 0.0087, 0.0075, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 06:26:31,044 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1545, 3.3879, 3.5758, 3.5377, 3.5306, 3.3757, 3.3738, 3.4187], device='cuda:3'), covar=tensor([0.0427, 0.0659, 0.0476, 0.0517, 0.0516, 0.0531, 0.0832, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0321, 0.0328, 0.0329, 0.0316, 0.0377, 0.0351, 0.0452, 0.0281], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 06:26:57,945 INFO [train.py:904] (3/8) Epoch 10, batch 6250, loss[loss=0.2328, simple_loss=0.3163, pruned_loss=0.07469, over 16391.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3039, pruned_loss=0.0709, over 3102672.27 frames. ], batch size: 146, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:27:18,722 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:18,879 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:47,812 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 6300, loss[loss=0.2018, simple_loss=0.285, pruned_loss=0.05926, over 16724.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3037, pruned_loss=0.07024, over 3097088.87 frames. ], batch size: 134, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:28:18,581 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:28:38,527 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 06:28:48,932 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:24,484 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 06:29:30,239 INFO [train.py:904] (3/8) Epoch 10, batch 6350, loss[loss=0.2291, simple_loss=0.3092, pruned_loss=0.07445, over 16680.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3054, pruned_loss=0.0722, over 3088197.64 frames. ], batch size: 134, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:29:31,921 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:33,048 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:49,806 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 06:30:13,648 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:22,271 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 06:30:22,845 INFO [optim.py:368] (3/8) 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 06:30:46,254 INFO [train.py:904] (3/8) Epoch 10, batch 6400, loss[loss=0.2262, simple_loss=0.3005, pruned_loss=0.07593, over 16330.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3054, pruned_loss=0.07304, over 3086132.58 frames. ], batch size: 35, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:00,802 INFO [train.py:904] (3/8) Epoch 10, batch 6450, loss[loss=0.2098, simple_loss=0.2977, pruned_loss=0.061, over 16745.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3052, pruned_loss=0.07237, over 3079897.57 frames. ], batch size: 83, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:35,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5638, 2.2055, 2.3315, 4.2182, 2.1012, 2.7143, 2.3279, 2.3884], device='cuda:3'), covar=tensor([0.0858, 0.3087, 0.2104, 0.0357, 0.3671, 0.1977, 0.2863, 0.2898], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0386, 0.0323, 0.0318, 0.0411, 0.0438, 0.0349, 0.0452], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:32:56,059 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6481, 2.3607, 2.1458, 3.4131, 2.3531, 3.4981, 1.2953, 2.5929], device='cuda:3'), covar=tensor([0.1489, 0.0787, 0.1320, 0.0169, 0.0185, 0.0393, 0.1902, 0.0855], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0157, 0.0179, 0.0137, 0.0201, 0.0206, 0.0180, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 06:32:57,255 INFO [optim.py:368] (3/8) 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 06:33:21,801 INFO [train.py:904] (3/8) Epoch 10, batch 6500, loss[loss=0.2502, simple_loss=0.3282, pruned_loss=0.08614, over 16960.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3035, pruned_loss=0.0711, over 3103870.58 frames. ], batch size: 41, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:17,490 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:34:41,149 INFO [train.py:904] (3/8) Epoch 10, batch 6550, loss[loss=0.2805, simple_loss=0.3369, pruned_loss=0.1121, over 11409.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3061, pruned_loss=0.07161, over 3109290.11 frames. ], batch size: 246, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:55,678 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:05,010 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:20,787 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 06:35:35,375 INFO [optim.py:368] (3/8) 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 06:35:56,850 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:36:00,723 INFO [train.py:904] (3/8) Epoch 10, batch 6600, loss[loss=0.2048, simple_loss=0.2901, pruned_loss=0.05975, over 16384.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.308, pruned_loss=0.07143, over 3120686.78 frames. ], batch size: 35, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:36:19,349 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:14,719 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 6650, loss[loss=0.3057, simple_loss=0.3525, pruned_loss=0.1295, over 11395.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3084, pruned_loss=0.07241, over 3116070.93 frames. ], batch size: 247, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:37:24,774 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:30,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8675, 1.9938, 2.2869, 3.1197, 2.1144, 2.2761, 2.2091, 2.0878], device='cuda:3'), covar=tensor([0.0934, 0.2959, 0.1773, 0.0522, 0.3469, 0.1904, 0.2543, 0.2865], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0381, 0.0319, 0.0314, 0.0405, 0.0432, 0.0345, 0.0446], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:37:30,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 06:38:05,321 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:38:05,343 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 3.281e+02 3.856e+02 4.909e+02 8.809e+02, threshold=7.713e+02, percent-clipped=1.0 2023-04-29 06:38:31,003 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:38,424 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:39,792 INFO [train.py:904] (3/8) Epoch 10, batch 6700, loss[loss=0.2117, simple_loss=0.289, pruned_loss=0.0672, over 16603.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3066, pruned_loss=0.07244, over 3109663.62 frames. ], batch size: 57, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:39:08,484 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0744, 2.4197, 2.3519, 2.8650, 2.2041, 3.2131, 1.7919, 2.7445], device='cuda:3'), covar=tensor([0.0989, 0.0452, 0.0914, 0.0113, 0.0146, 0.0361, 0.1238, 0.0600], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0156, 0.0179, 0.0136, 0.0202, 0.0204, 0.0179, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 06:39:20,426 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:39:57,931 INFO [train.py:904] (3/8) Epoch 10, batch 6750, loss[loss=0.2252, simple_loss=0.302, pruned_loss=0.07422, over 16656.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.305, pruned_loss=0.07223, over 3106722.72 frames. ], batch size: 134, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:40:04,396 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:40:32,179 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 06:40:49,800 INFO [optim.py:368] (3/8) 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 06:41:12,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2567, 1.9604, 2.0781, 3.8672, 1.9114, 2.4138, 2.1019, 2.1807], device='cuda:3'), covar=tensor([0.0932, 0.3138, 0.2253, 0.0374, 0.3798, 0.2118, 0.2864, 0.3126], device='cuda:3'), in_proj_covar=tensor([0.0356, 0.0382, 0.0321, 0.0316, 0.0408, 0.0433, 0.0345, 0.0449], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:41:15,032 INFO [train.py:904] (3/8) Epoch 10, batch 6800, loss[loss=0.2379, simple_loss=0.3227, pruned_loss=0.07652, over 16333.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3059, pruned_loss=0.07262, over 3111503.42 frames. ], batch size: 165, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:41:39,192 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:41:53,641 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4447, 4.3919, 4.2276, 3.6033, 4.2997, 1.6330, 4.0771, 4.0913], device='cuda:3'), covar=tensor([0.0074, 0.0065, 0.0130, 0.0319, 0.0073, 0.2284, 0.0102, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0110, 0.0156, 0.0152, 0.0127, 0.0172, 0.0144, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:42:12,177 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5573, 3.5338, 2.7167, 2.1224, 2.4357, 2.2219, 3.6469, 3.2308], device='cuda:3'), covar=tensor([0.2635, 0.0730, 0.1618, 0.2284, 0.2344, 0.1798, 0.0514, 0.1178], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0256, 0.0280, 0.0276, 0.0281, 0.0217, 0.0264, 0.0288], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:42:33,968 INFO [train.py:904] (3/8) Epoch 10, batch 6850, loss[loss=0.2067, simple_loss=0.308, pruned_loss=0.05271, over 16825.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3078, pruned_loss=0.07367, over 3090885.99 frames. ], batch size: 96, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:42:47,635 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.168e+02 2.932e+02 3.586e+02 4.673e+02 8.284e+02, threshold=7.173e+02, percent-clipped=0.0 2023-04-29 06:43:37,275 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:43:49,037 INFO [train.py:904] (3/8) Epoch 10, batch 6900, loss[loss=0.2525, simple_loss=0.331, pruned_loss=0.087, over 16506.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3098, pruned_loss=0.07272, over 3110270.06 frames. ], batch size: 146, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:44:01,149 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:45:09,157 INFO [train.py:904] (3/8) Epoch 10, batch 6950, loss[loss=0.2935, simple_loss=0.346, pruned_loss=0.1205, over 11471.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3107, pruned_loss=0.07407, over 3101552.01 frames. ], batch size: 248, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:45:22,566 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4128, 3.4935, 1.8661, 3.8060, 2.4679, 3.8158, 1.9878, 2.5360], device='cuda:3'), covar=tensor([0.0210, 0.0336, 0.1656, 0.0122, 0.0850, 0.0451, 0.1567, 0.0812], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0161, 0.0186, 0.0118, 0.0165, 0.0202, 0.0191, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 06:45:54,164 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:46:01,761 INFO [optim.py:368] (3/8) 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 06:46:27,408 INFO [train.py:904] (3/8) Epoch 10, batch 7000, loss[loss=0.2414, simple_loss=0.331, pruned_loss=0.07585, over 16389.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3108, pruned_loss=0.07306, over 3114000.04 frames. ], batch size: 146, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:08,046 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:47:43,136 INFO [train.py:904] (3/8) Epoch 10, batch 7050, loss[loss=0.2031, simple_loss=0.292, pruned_loss=0.05709, over 17021.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3111, pruned_loss=0.07286, over 3107183.12 frames. ], batch size: 50, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:57,086 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:48:34,408 INFO [optim.py:368] (3/8) 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 06:48:59,554 INFO [train.py:904] (3/8) Epoch 10, batch 7100, loss[loss=0.211, simple_loss=0.2914, pruned_loss=0.0653, over 17044.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3106, pruned_loss=0.07367, over 3084196.21 frames. ], batch size: 41, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:49:14,097 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:49:27,959 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:50:12,804 INFO [train.py:904] (3/8) Epoch 10, batch 7150, loss[loss=0.2245, simple_loss=0.3056, pruned_loss=0.07171, over 16547.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3089, pruned_loss=0.07359, over 3069669.59 frames. ], batch size: 35, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:50:58,268 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0473, 2.3133, 1.9010, 2.1022, 2.6749, 2.3593, 2.8659, 2.9277], device='cuda:3'), covar=tensor([0.0086, 0.0281, 0.0365, 0.0336, 0.0185, 0.0306, 0.0157, 0.0167], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0195, 0.0194, 0.0193, 0.0194, 0.0198, 0.0198, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:51:03,784 INFO [optim.py:368] (3/8) 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 06:51:15,724 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:16,744 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:29,369 INFO [train.py:904] (3/8) Epoch 10, batch 7200, loss[loss=0.184, simple_loss=0.2806, pruned_loss=0.04369, over 16781.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3064, pruned_loss=0.07116, over 3085803.36 frames. ], batch size: 83, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:52:10,904 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 06:52:32,648 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:52:48,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6869, 3.6999, 1.8976, 4.1686, 2.6863, 4.1062, 2.1757, 2.8860], device='cuda:3'), covar=tensor([0.0184, 0.0297, 0.1696, 0.0101, 0.0768, 0.0385, 0.1395, 0.0622], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0161, 0.0186, 0.0117, 0.0165, 0.0201, 0.0190, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 06:52:49,208 INFO [train.py:904] (3/8) Epoch 10, batch 7250, loss[loss=0.1889, simple_loss=0.2709, pruned_loss=0.05341, over 16748.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3031, pruned_loss=0.0694, over 3093102.06 frames. ], batch size: 83, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:52:53,574 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:53:45,146 INFO [optim.py:368] (3/8) 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 06:54:00,575 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 06:54:05,880 INFO [train.py:904] (3/8) Epoch 10, batch 7300, loss[loss=0.2155, simple_loss=0.3042, pruned_loss=0.06339, over 16898.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3022, pruned_loss=0.06855, over 3110857.77 frames. ], batch size: 109, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:54:43,199 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8810, 2.1075, 2.3556, 3.1045, 2.1149, 2.3736, 2.3003, 2.1826], device='cuda:3'), covar=tensor([0.0882, 0.2549, 0.1685, 0.0551, 0.3278, 0.1720, 0.2305, 0.2729], device='cuda:3'), in_proj_covar=tensor([0.0356, 0.0384, 0.0321, 0.0317, 0.0409, 0.0435, 0.0346, 0.0450], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:54:51,863 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8246, 4.8872, 5.2915, 5.2491, 5.2615, 4.8612, 4.9081, 4.5717], device='cuda:3'), covar=tensor([0.0245, 0.0331, 0.0249, 0.0292, 0.0346, 0.0251, 0.0696, 0.0414], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0318, 0.0320, 0.0306, 0.0369, 0.0341, 0.0441, 0.0275], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 06:55:23,864 INFO [train.py:904] (3/8) Epoch 10, batch 7350, loss[loss=0.2092, simple_loss=0.2967, pruned_loss=0.06082, over 16852.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3032, pruned_loss=0.0695, over 3090213.96 frames. ], batch size: 90, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:55:42,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8448, 2.6982, 2.1316, 2.4605, 3.1714, 2.7624, 3.5786, 3.3927], device='cuda:3'), covar=tensor([0.0041, 0.0242, 0.0354, 0.0299, 0.0159, 0.0279, 0.0130, 0.0139], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0193, 0.0193, 0.0191, 0.0192, 0.0197, 0.0196, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 06:56:19,126 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.271e+02 3.294e+02 3.832e+02 4.622e+02 6.263e+02, threshold=7.664e+02, percent-clipped=0.0 2023-04-29 06:56:23,873 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3129, 4.2625, 4.7646, 4.7310, 4.6791, 4.3838, 4.3820, 4.1831], device='cuda:3'), covar=tensor([0.0302, 0.0551, 0.0363, 0.0378, 0.0447, 0.0378, 0.0895, 0.0532], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0319, 0.0321, 0.0307, 0.0369, 0.0341, 0.0442, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 06:56:41,003 INFO [train.py:904] (3/8) Epoch 10, batch 7400, loss[loss=0.3169, simple_loss=0.3632, pruned_loss=0.1353, over 11280.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3052, pruned_loss=0.07119, over 3078716.08 frames. ], batch size: 250, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:57,597 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:03,106 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 7450, loss[loss=0.2132, simple_loss=0.306, pruned_loss=0.06021, over 16480.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3065, pruned_loss=0.07222, over 3072133.15 frames. ], batch size: 68, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:58:14,252 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:58:33,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 06:58:57,288 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.303e+02 3.942e+02 4.934e+02 1.157e+03, threshold=7.883e+02, percent-clipped=3.0 2023-04-29 06:59:03,834 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 06:59:15,870 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-04-29 06:59:16,001 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-29 06:59:19,898 INFO [train.py:904] (3/8) Epoch 10, batch 7500, loss[loss=0.2294, simple_loss=0.3064, pruned_loss=0.07621, over 16670.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3069, pruned_loss=0.07201, over 3070929.79 frames. ], batch size: 76, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:59:26,732 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 07:00:20,554 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:27,941 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:34,222 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:38,279 INFO [train.py:904] (3/8) Epoch 10, batch 7550, loss[loss=0.1856, simple_loss=0.2647, pruned_loss=0.05327, over 16640.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3061, pruned_loss=0.07275, over 3043930.03 frames. ], batch size: 62, lr: 6.76e-03, grad_scale: 2.0 2023-04-29 07:01:32,281 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.425e+02 3.280e+02 3.927e+02 5.001e+02 7.830e+02, threshold=7.854e+02, percent-clipped=0.0 2023-04-29 07:01:41,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4602, 2.0060, 1.6180, 1.8337, 2.3180, 2.1003, 2.4092, 2.5639], device='cuda:3'), covar=tensor([0.0100, 0.0318, 0.0419, 0.0353, 0.0182, 0.0283, 0.0159, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0194, 0.0192, 0.0191, 0.0193, 0.0196, 0.0196, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:01:53,291 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:01:53,973 INFO [train.py:904] (3/8) Epoch 10, batch 7600, loss[loss=0.2045, simple_loss=0.2891, pruned_loss=0.0599, over 16790.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3048, pruned_loss=0.0725, over 3047792.39 frames. ], batch size: 102, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:02:01,615 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:02:24,335 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 7650, loss[loss=0.2306, simple_loss=0.3123, pruned_loss=0.07438, over 16775.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.305, pruned_loss=0.07276, over 3069189.69 frames. ], batch size: 124, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:03:12,365 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7388, 6.0185, 5.6460, 5.7888, 5.3129, 5.1899, 5.5316, 6.0924], device='cuda:3'), covar=tensor([0.0931, 0.0690, 0.1030, 0.0680, 0.0766, 0.0603, 0.0867, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0513, 0.0638, 0.0535, 0.0443, 0.0399, 0.0423, 0.0534, 0.0485], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:03:26,732 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6390, 3.6174, 3.9741, 1.7855, 4.1901, 4.2162, 2.9965, 3.1873], device='cuda:3'), covar=tensor([0.0697, 0.0184, 0.0148, 0.1201, 0.0040, 0.0098, 0.0370, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0099, 0.0085, 0.0137, 0.0068, 0.0096, 0.0120, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 07:03:44,999 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6049, 2.5623, 2.2988, 3.7703, 2.7646, 3.8945, 1.3123, 2.6861], device='cuda:3'), covar=tensor([0.1351, 0.0687, 0.1162, 0.0142, 0.0206, 0.0361, 0.1607, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0136, 0.0202, 0.0207, 0.0179, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 07:03:59,246 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:08,572 INFO [optim.py:368] (3/8) 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 07:04:28,262 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:29,012 INFO [train.py:904] (3/8) Epoch 10, batch 7700, loss[loss=0.2269, simple_loss=0.3169, pruned_loss=0.06849, over 16315.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3053, pruned_loss=0.07304, over 3081856.74 frames. ], batch size: 146, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:04:50,180 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:05:32,229 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4535, 4.4469, 4.3295, 3.6080, 4.3951, 1.5247, 4.0938, 4.0366], device='cuda:3'), covar=tensor([0.0091, 0.0065, 0.0136, 0.0358, 0.0078, 0.2583, 0.0118, 0.0204], device='cuda:3'), in_proj_covar=tensor([0.0122, 0.0107, 0.0155, 0.0150, 0.0126, 0.0173, 0.0142, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:05:43,694 INFO [train.py:904] (3/8) Epoch 10, batch 7750, loss[loss=0.2316, simple_loss=0.3126, pruned_loss=0.07531, over 16942.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3062, pruned_loss=0.07361, over 3074562.26 frames. ], batch size: 109, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:05:51,236 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0438, 1.9512, 2.1919, 3.4862, 1.9309, 2.3135, 2.1059, 2.1361], device='cuda:3'), covar=tensor([0.1027, 0.3233, 0.2066, 0.0482, 0.3759, 0.2179, 0.2919, 0.2888], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0385, 0.0321, 0.0319, 0.0413, 0.0436, 0.0346, 0.0450], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:05:52,391 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4319, 2.0729, 2.1424, 4.0998, 1.9627, 2.4980, 2.1798, 2.2955], device='cuda:3'), covar=tensor([0.0920, 0.3216, 0.2155, 0.0367, 0.3780, 0.2192, 0.2879, 0.2843], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0385, 0.0321, 0.0319, 0.0413, 0.0436, 0.0346, 0.0450], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:06:00,253 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:06,075 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 07:06:38,058 INFO [optim.py:368] (3/8) 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 07:06:59,328 INFO [train.py:904] (3/8) Epoch 10, batch 7800, loss[loss=0.2265, simple_loss=0.316, pruned_loss=0.06846, over 16264.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3069, pruned_loss=0.07401, over 3078420.05 frames. ], batch size: 35, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:07:05,275 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0373, 3.9339, 2.3189, 4.7482, 2.9343, 4.6584, 2.2031, 3.0439], device='cuda:3'), covar=tensor([0.0188, 0.0356, 0.1591, 0.0113, 0.0762, 0.0365, 0.1583, 0.0734], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0163, 0.0187, 0.0120, 0.0165, 0.0204, 0.0192, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 07:08:06,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2296, 3.4392, 3.6451, 1.7356, 3.8856, 3.9252, 2.7244, 2.9642], device='cuda:3'), covar=tensor([0.0832, 0.0205, 0.0193, 0.1213, 0.0050, 0.0105, 0.0444, 0.0402], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0100, 0.0085, 0.0138, 0.0069, 0.0097, 0.0121, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 07:08:12,877 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 7850, loss[loss=0.2274, simple_loss=0.3104, pruned_loss=0.07221, over 16166.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.307, pruned_loss=0.07311, over 3082734.25 frames. ], batch size: 165, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:08:52,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4667, 3.5236, 3.1645, 3.0320, 3.1190, 3.3902, 3.3141, 3.1295], device='cuda:3'), covar=tensor([0.0593, 0.0508, 0.0249, 0.0252, 0.0593, 0.0459, 0.0964, 0.0533], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0295, 0.0272, 0.0252, 0.0294, 0.0289, 0.0188, 0.0318], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:08:57,356 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:10,220 INFO [optim.py:368] (3/8) 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 07:09:22,396 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:23,679 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:29,663 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:30,435 INFO [train.py:904] (3/8) Epoch 10, batch 7900, loss[loss=0.2647, simple_loss=0.3205, pruned_loss=0.1044, over 11833.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3056, pruned_loss=0.07232, over 3087708.03 frames. ], batch size: 247, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:09:55,564 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:10:31,505 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 7950, loss[loss=0.2268, simple_loss=0.3037, pruned_loss=0.07499, over 16876.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3055, pruned_loss=0.07232, over 3088393.07 frames. ], batch size: 116, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:11:22,432 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:27,242 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:30,513 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:12:06,027 INFO [train.py:904] (3/8) Epoch 10, batch 8000, loss[loss=0.2572, simple_loss=0.3205, pruned_loss=0.09698, over 11505.00 frames. ], tot_loss[loss=0.2254, simple_loss=0.3059, pruned_loss=0.07248, over 3091812.14 frames. ], batch size: 247, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:12:57,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2717, 5.1473, 4.9691, 4.3809, 5.0727, 1.8423, 4.8457, 4.9302], device='cuda:3'), covar=tensor([0.0053, 0.0047, 0.0124, 0.0312, 0.0058, 0.2216, 0.0082, 0.0127], device='cuda:3'), in_proj_covar=tensor([0.0123, 0.0108, 0.0156, 0.0151, 0.0127, 0.0174, 0.0142, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:12:57,279 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 8050, loss[loss=0.2045, simple_loss=0.2986, pruned_loss=0.05522, over 16481.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3066, pruned_loss=0.07263, over 3086145.31 frames. ], batch size: 75, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:13:29,868 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:50,457 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:14:01,615 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6652, 3.6928, 2.1207, 4.1163, 2.7406, 4.1395, 2.2438, 2.7864], device='cuda:3'), covar=tensor([0.0201, 0.0323, 0.1536, 0.0153, 0.0783, 0.0437, 0.1482, 0.0727], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0162, 0.0187, 0.0120, 0.0166, 0.0205, 0.0193, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 07:14:18,153 INFO [optim.py:368] (3/8) 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 07:14:39,576 INFO [train.py:904] (3/8) Epoch 10, batch 8100, loss[loss=0.1962, simple_loss=0.2781, pruned_loss=0.05718, over 16661.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3055, pruned_loss=0.07144, over 3091383.53 frames. ], batch size: 62, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:15:03,921 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:15:23,111 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 8150, loss[loss=0.2015, simple_loss=0.2816, pruned_loss=0.06069, over 16701.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3036, pruned_loss=0.071, over 3087021.53 frames. ], batch size: 124, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:16:35,971 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:16:49,692 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 2.958e+02 3.724e+02 4.411e+02 7.985e+02, threshold=7.447e+02, percent-clipped=3.0 2023-04-29 07:17:01,359 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:08,937 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:09,686 INFO [train.py:904] (3/8) Epoch 10, batch 8200, loss[loss=0.2199, simple_loss=0.2977, pruned_loss=0.07106, over 16826.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3018, pruned_loss=0.07066, over 3082650.68 frames. ], batch size: 42, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:17:18,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4355, 4.5353, 4.7018, 4.5339, 4.5989, 5.1043, 4.6017, 4.3490], device='cuda:3'), covar=tensor([0.1123, 0.1732, 0.1680, 0.2055, 0.2442, 0.0987, 0.1464, 0.2472], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0477, 0.0512, 0.0414, 0.0544, 0.0538, 0.0416, 0.0565], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 07:17:35,526 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:04,664 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:11,173 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:28,356 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 8250, loss[loss=0.2161, simple_loss=0.3085, pruned_loss=0.06185, over 15159.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3018, pruned_loss=0.06873, over 3079468.53 frames. ], batch size: 190, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:19:10,182 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:16,606 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:19,864 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:32,259 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2572, 3.4513, 3.4703, 2.4455, 3.2744, 3.4888, 3.3862, 2.1931], device='cuda:3'), covar=tensor([0.0352, 0.0033, 0.0036, 0.0271, 0.0057, 0.0073, 0.0051, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0064, 0.0067, 0.0122, 0.0074, 0.0086, 0.0074, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 07:19:36,140 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.844e+02 3.459e+02 4.111e+02 7.773e+02, threshold=6.919e+02, percent-clipped=2.0 2023-04-29 07:19:55,512 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:57,962 INFO [train.py:904] (3/8) Epoch 10, batch 8300, loss[loss=0.2179, simple_loss=0.3231, pruned_loss=0.05633, over 16812.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2991, pruned_loss=0.06576, over 3082545.44 frames. ], batch size: 124, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:20:37,233 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,555 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,644 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:21:21,652 INFO [train.py:904] (3/8) Epoch 10, batch 8350, loss[loss=0.1917, simple_loss=0.2886, pruned_loss=0.04738, over 16437.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2976, pruned_loss=0.06345, over 3067797.11 frames. ], batch size: 146, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:21:29,871 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:21:39,598 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3758, 4.2624, 4.7229, 4.7095, 4.6856, 4.4325, 4.3819, 4.2590], device='cuda:3'), covar=tensor([0.0260, 0.0675, 0.0366, 0.0387, 0.0413, 0.0308, 0.0848, 0.0411], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0319, 0.0318, 0.0305, 0.0368, 0.0339, 0.0437, 0.0275], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 07:22:03,551 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 07:22:21,849 INFO [optim.py:368] (3/8) 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 07:22:26,020 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:43,905 INFO [train.py:904] (3/8) Epoch 10, batch 8400, loss[loss=0.2037, simple_loss=0.3033, pruned_loss=0.05202, over 16217.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2945, pruned_loss=0.06095, over 3059928.84 frames. ], batch size: 165, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:22:49,993 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:23:25,452 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:24:06,252 INFO [train.py:904] (3/8) Epoch 10, batch 8450, loss[loss=0.2058, simple_loss=0.2907, pruned_loss=0.06043, over 16427.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2925, pruned_loss=0.05884, over 3077025.41 frames. ], batch size: 146, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:24:28,570 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3869, 4.3815, 4.6410, 4.4718, 4.5323, 5.0090, 4.6191, 4.3524], device='cuda:3'), covar=tensor([0.1112, 0.1840, 0.1584, 0.1724, 0.2200, 0.0915, 0.1242, 0.2159], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0453, 0.0488, 0.0396, 0.0521, 0.0517, 0.0398, 0.0539], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 07:24:42,070 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:25:06,072 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.551e+02 3.005e+02 3.693e+02 6.066e+02, threshold=6.011e+02, percent-clipped=2.0 2023-04-29 07:25:25,941 INFO [train.py:904] (3/8) Epoch 10, batch 8500, loss[loss=0.1782, simple_loss=0.2692, pruned_loss=0.04361, over 16328.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2881, pruned_loss=0.05634, over 3057833.94 frames. ], batch size: 146, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:26:09,254 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 07:26:21,571 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 8550, loss[loss=0.2028, simple_loss=0.2987, pruned_loss=0.05339, over 16301.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2853, pruned_loss=0.05493, over 3064935.61 frames. ], batch size: 165, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:27:34,852 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:34,907 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:50,547 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:02,240 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:18,841 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8964, 5.1871, 4.9130, 4.9225, 4.6257, 4.6434, 4.6208, 5.2738], device='cuda:3'), covar=tensor([0.0917, 0.0884, 0.1009, 0.0712, 0.0818, 0.0815, 0.1044, 0.0824], device='cuda:3'), in_proj_covar=tensor([0.0492, 0.0620, 0.0515, 0.0431, 0.0389, 0.0413, 0.0520, 0.0471], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:28:29,552 INFO [train.py:904] (3/8) Epoch 10, batch 8600, loss[loss=0.1679, simple_loss=0.2685, pruned_loss=0.03363, over 16846.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2847, pruned_loss=0.05359, over 3048799.85 frames. ], batch size: 90, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:29:10,991 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:29:26,462 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:29:54,919 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3644, 2.9532, 2.6773, 2.1369, 2.1711, 2.1384, 3.0505, 2.8018], device='cuda:3'), covar=tensor([0.2406, 0.0694, 0.1315, 0.2170, 0.2181, 0.1808, 0.0452, 0.1039], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0246, 0.0272, 0.0267, 0.0267, 0.0213, 0.0255, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:30:11,017 INFO [train.py:904] (3/8) Epoch 10, batch 8650, loss[loss=0.1905, simple_loss=0.2928, pruned_loss=0.04411, over 16570.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2825, pruned_loss=0.05219, over 3036569.78 frames. ], batch size: 62, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:30:24,813 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:30:52,936 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 07:31:10,430 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 07:31:56,449 INFO [train.py:904] (3/8) Epoch 10, batch 8700, loss[loss=0.1958, simple_loss=0.2799, pruned_loss=0.05589, over 16845.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2797, pruned_loss=0.05078, over 3048481.39 frames. ], batch size: 124, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:32:28,536 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1891, 4.2480, 4.0518, 3.8438, 3.7337, 4.1555, 3.8721, 3.8636], device='cuda:3'), covar=tensor([0.0514, 0.0402, 0.0236, 0.0251, 0.0729, 0.0398, 0.0575, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0227, 0.0286, 0.0265, 0.0246, 0.0286, 0.0281, 0.0184, 0.0310], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:32:28,574 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:32:42,795 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:33:35,355 INFO [train.py:904] (3/8) Epoch 10, batch 8750, loss[loss=0.1803, simple_loss=0.2784, pruned_loss=0.04109, over 16546.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2796, pruned_loss=0.05013, over 3040066.38 frames. ], batch size: 68, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:34:32,194 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:34:32,325 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:34:33,991 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 07:35:02,704 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.530e+02 3.072e+02 4.161e+02 6.852e+02, threshold=6.144e+02, percent-clipped=8.0 2023-04-29 07:35:29,749 INFO [train.py:904] (3/8) Epoch 10, batch 8800, loss[loss=0.1916, simple_loss=0.2849, pruned_loss=0.04918, over 16333.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2786, pruned_loss=0.04905, over 3060297.31 frames. ], batch size: 146, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:36:12,282 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:36:23,675 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-04-29 07:36:33,782 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 07:37:15,877 INFO [train.py:904] (3/8) Epoch 10, batch 8850, loss[loss=0.1768, simple_loss=0.2692, pruned_loss=0.04222, over 12728.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2802, pruned_loss=0.04839, over 3039035.17 frames. ], batch size: 248, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:37:40,766 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 07:38:03,389 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:24,186 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4143, 2.8945, 2.6142, 2.1975, 2.1760, 2.1605, 2.9624, 2.7853], device='cuda:3'), covar=tensor([0.2120, 0.0776, 0.1340, 0.2054, 0.2067, 0.1724, 0.0413, 0.1037], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0244, 0.0270, 0.0263, 0.0260, 0.0210, 0.0253, 0.0273], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:38:34,779 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:36,934 INFO [optim.py:368] (3/8) 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 07:38:48,405 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 8900, loss[loss=0.2012, simple_loss=0.2924, pruned_loss=0.05502, over 16928.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2799, pruned_loss=0.04726, over 3051107.52 frames. ], batch size: 116, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:39:43,459 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:40:45,778 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:01,128 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 8950, loss[loss=0.1631, simple_loss=0.2583, pruned_loss=0.03398, over 15292.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2796, pruned_loss=0.04753, over 3063367.10 frames. ], batch size: 190, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:42:22,239 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.569e+02 2.987e+02 3.833e+02 6.749e+02, threshold=5.974e+02, percent-clipped=2.0 2023-04-29 07:42:55,064 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5065, 4.5605, 4.3819, 4.0942, 4.0074, 4.4424, 4.2726, 4.1064], device='cuda:3'), covar=tensor([0.0540, 0.0458, 0.0256, 0.0256, 0.0863, 0.0460, 0.0394, 0.0643], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0282, 0.0263, 0.0244, 0.0285, 0.0279, 0.0182, 0.0306], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:42:55,809 INFO [train.py:904] (3/8) Epoch 10, batch 9000, loss[loss=0.1759, simple_loss=0.2622, pruned_loss=0.04477, over 16961.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2757, pruned_loss=0.04581, over 3052467.64 frames. ], batch size: 116, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:42:55,809 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 07:43:05,470 INFO [train.py:938] (3/8) Epoch 10, validation: loss=0.1565, simple_loss=0.2604, pruned_loss=0.02634, over 944034.00 frames. 2023-04-29 07:43:05,471 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 07:43:18,848 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3711, 3.6044, 3.9700, 1.8624, 4.1172, 4.1989, 3.0739, 2.9843], device='cuda:3'), covar=tensor([0.0772, 0.0183, 0.0121, 0.1153, 0.0036, 0.0071, 0.0326, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0097, 0.0081, 0.0135, 0.0066, 0.0094, 0.0116, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:3') 2023-04-29 07:43:30,274 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:44:12,989 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:44:26,229 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0464, 2.6241, 2.7047, 1.7854, 2.9107, 2.9983, 2.4840, 2.4982], device='cuda:3'), covar=tensor([0.0653, 0.0209, 0.0157, 0.1057, 0.0079, 0.0134, 0.0404, 0.0396], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0097, 0.0081, 0.0136, 0.0066, 0.0094, 0.0116, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 07:44:48,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2234, 3.2588, 1.7790, 3.4792, 2.3588, 3.4744, 1.9863, 2.6926], device='cuda:3'), covar=tensor([0.0222, 0.0323, 0.1683, 0.0156, 0.0885, 0.0484, 0.1562, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0151, 0.0178, 0.0113, 0.0157, 0.0188, 0.0186, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 07:44:50,487 INFO [train.py:904] (3/8) Epoch 10, batch 9050, loss[loss=0.1803, simple_loss=0.269, pruned_loss=0.04578, over 16720.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2771, pruned_loss=0.04633, over 3067089.78 frames. ], batch size: 134, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:44:51,442 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:45:15,201 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 07:45:25,589 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5312, 4.5402, 4.3764, 4.0875, 4.0023, 4.4652, 4.2751, 4.1520], device='cuda:3'), covar=tensor([0.0460, 0.0473, 0.0247, 0.0255, 0.0893, 0.0357, 0.0437, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0280, 0.0261, 0.0243, 0.0283, 0.0277, 0.0182, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:46:08,563 INFO [optim.py:368] (3/8) 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 07:46:37,386 INFO [train.py:904] (3/8) Epoch 10, batch 9100, loss[loss=0.1942, simple_loss=0.2885, pruned_loss=0.05002, over 16209.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.277, pruned_loss=0.04695, over 3061738.53 frames. ], batch size: 165, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:46:58,833 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:47:00,892 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 07:47:08,312 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-04-29 07:48:02,237 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 07:48:34,506 INFO [train.py:904] (3/8) Epoch 10, batch 9150, loss[loss=0.1855, simple_loss=0.2654, pruned_loss=0.05284, over 11921.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2773, pruned_loss=0.04668, over 3050636.34 frames. ], batch size: 247, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:49:12,767 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 07:49:49,086 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3896, 4.4463, 4.7921, 4.7725, 4.7575, 4.5006, 4.4066, 4.2755], device='cuda:3'), covar=tensor([0.0245, 0.0423, 0.0363, 0.0354, 0.0325, 0.0312, 0.0692, 0.0396], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0303, 0.0309, 0.0296, 0.0354, 0.0326, 0.0417, 0.0268], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-29 07:49:54,593 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.281e+02 2.663e+02 3.521e+02 5.779e+02, threshold=5.326e+02, percent-clipped=0.0 2023-04-29 07:50:13,980 INFO [train.py:904] (3/8) Epoch 10, batch 9200, loss[loss=0.1878, simple_loss=0.2755, pruned_loss=0.05011, over 16360.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2735, pruned_loss=0.0458, over 3079254.02 frames. ], batch size: 146, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:50:18,450 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7408, 3.8922, 2.1768, 4.3187, 2.6869, 4.2336, 2.3738, 3.0867], device='cuda:3'), covar=tensor([0.0198, 0.0261, 0.1515, 0.0162, 0.0764, 0.0381, 0.1357, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0153, 0.0180, 0.0115, 0.0159, 0.0190, 0.0188, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 07:51:35,109 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:51:50,966 INFO [train.py:904] (3/8) Epoch 10, batch 9250, loss[loss=0.2151, simple_loss=0.3094, pruned_loss=0.06038, over 15308.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2727, pruned_loss=0.0455, over 3059422.18 frames. ], batch size: 190, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:53:14,008 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.424e+02 2.858e+02 3.550e+02 5.723e+02, threshold=5.716e+02, percent-clipped=3.0 2023-04-29 07:53:42,535 INFO [train.py:904] (3/8) Epoch 10, batch 9300, loss[loss=0.171, simple_loss=0.2614, pruned_loss=0.04025, over 15451.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2707, pruned_loss=0.04459, over 3056500.28 frames. ], batch size: 191, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:53:46,088 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7398, 4.7064, 4.5342, 4.1182, 4.5591, 1.9771, 4.4195, 4.4761], device='cuda:3'), covar=tensor([0.0064, 0.0079, 0.0115, 0.0228, 0.0079, 0.2047, 0.0084, 0.0153], device='cuda:3'), in_proj_covar=tensor([0.0120, 0.0107, 0.0149, 0.0140, 0.0123, 0.0171, 0.0137, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-29 07:54:09,123 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:55:29,731 INFO [train.py:904] (3/8) Epoch 10, batch 9350, loss[loss=0.1826, simple_loss=0.262, pruned_loss=0.05157, over 12163.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2707, pruned_loss=0.04458, over 3072213.09 frames. ], batch size: 246, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:55:50,100 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:56:29,097 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:56:39,944 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 07:56:47,393 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 07:56:48,376 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 9400, loss[loss=0.1571, simple_loss=0.242, pruned_loss=0.03605, over 12292.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2701, pruned_loss=0.04411, over 3065310.77 frames. ], batch size: 248, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:57:25,166 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:58:33,102 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:58:52,911 INFO [train.py:904] (3/8) Epoch 10, batch 9450, loss[loss=0.1803, simple_loss=0.2697, pruned_loss=0.04539, over 16960.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2722, pruned_loss=0.04451, over 3072206.51 frames. ], batch size: 109, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:59:14,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7626, 4.0628, 3.2527, 2.4143, 2.7517, 2.5178, 4.2931, 3.6867], device='cuda:3'), covar=tensor([0.2331, 0.0570, 0.1243, 0.1976, 0.2185, 0.1623, 0.0295, 0.0824], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0245, 0.0271, 0.0265, 0.0256, 0.0212, 0.0254, 0.0274], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 07:59:25,803 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9774, 2.7242, 2.8135, 1.9397, 2.6618, 2.0678, 2.6892, 2.7797], device='cuda:3'), covar=tensor([0.0271, 0.0710, 0.0466, 0.1651, 0.0665, 0.0976, 0.0620, 0.0771], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0134, 0.0154, 0.0140, 0.0133, 0.0123, 0.0133, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 08:00:10,871 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 9500, loss[loss=0.1695, simple_loss=0.2658, pruned_loss=0.03662, over 15421.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2714, pruned_loss=0.04413, over 3084672.84 frames. ], batch size: 191, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:02:03,498 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:02:20,733 INFO [train.py:904] (3/8) Epoch 10, batch 9550, loss[loss=0.1809, simple_loss=0.2647, pruned_loss=0.04855, over 12709.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2709, pruned_loss=0.04427, over 3097887.36 frames. ], batch size: 248, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:03:40,250 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 10, batch 9600, loss[loss=0.194, simple_loss=0.2805, pruned_loss=0.0538, over 12419.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2735, pruned_loss=0.04562, over 3082620.49 frames. ], batch size: 248, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:05:52,887 INFO [train.py:904] (3/8) Epoch 10, batch 9650, loss[loss=0.1996, simple_loss=0.3004, pruned_loss=0.04946, over 16382.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2755, pruned_loss=0.04608, over 3075792.34 frames. ], batch size: 146, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:06:51,841 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4279, 4.4937, 4.3203, 4.0805, 3.9608, 4.3867, 4.1912, 4.1019], device='cuda:3'), covar=tensor([0.0546, 0.0443, 0.0247, 0.0243, 0.0812, 0.0381, 0.0473, 0.0656], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0276, 0.0260, 0.0239, 0.0279, 0.0274, 0.0181, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:07:15,512 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 10, batch 9700, loss[loss=0.1778, simple_loss=0.267, pruned_loss=0.04429, over 16626.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2743, pruned_loss=0.04577, over 3069255.20 frames. ], batch size: 134, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:07:52,606 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:08:09,661 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 08:08:54,362 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 08:09:25,195 INFO [train.py:904] (3/8) Epoch 10, batch 9750, loss[loss=0.1754, simple_loss=0.275, pruned_loss=0.03791, over 15340.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2732, pruned_loss=0.04591, over 3075242.39 frames. ], batch size: 190, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:09:32,960 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:09:34,809 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.458e+02 3.043e+02 3.868e+02 6.520e+02, threshold=6.087e+02, percent-clipped=0.0 2023-04-29 08:11:05,030 INFO [train.py:904] (3/8) Epoch 10, batch 9800, loss[loss=0.1738, simple_loss=0.256, pruned_loss=0.04577, over 11991.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2731, pruned_loss=0.04552, over 3049034.36 frames. ], batch size: 248, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:11:29,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 08:11:36,086 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 10, batch 9850, loss[loss=0.172, simple_loss=0.2647, pruned_loss=0.03971, over 12603.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.274, pruned_loss=0.04519, over 3051253.12 frames. ], batch size: 250, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:12:56,406 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6711, 4.0090, 3.0882, 2.2930, 2.7473, 2.5221, 4.2940, 3.5737], device='cuda:3'), covar=tensor([0.2565, 0.0560, 0.1394, 0.2188, 0.2064, 0.1547, 0.0334, 0.0905], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0246, 0.0273, 0.0266, 0.0256, 0.0212, 0.0254, 0.0276], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:14:17,926 INFO [optim.py:368] (3/8) 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 08:14:41,617 INFO [train.py:904] (3/8) Epoch 10, batch 9900, loss[loss=0.1779, simple_loss=0.2758, pruned_loss=0.03997, over 15349.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2742, pruned_loss=0.04499, over 3053361.33 frames. ], batch size: 192, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:14:42,379 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9602, 4.0441, 4.3073, 4.2740, 4.3628, 4.1326, 4.0388, 4.0099], device='cuda:3'), covar=tensor([0.0369, 0.0594, 0.0502, 0.0598, 0.0481, 0.0379, 0.0906, 0.0433], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0299, 0.0303, 0.0289, 0.0347, 0.0319, 0.0404, 0.0260], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-29 08:16:23,601 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7144, 3.6982, 4.0596, 4.0553, 4.0471, 3.8230, 3.8305, 3.8180], device='cuda:3'), covar=tensor([0.0287, 0.0541, 0.0416, 0.0392, 0.0449, 0.0339, 0.0728, 0.0381], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0300, 0.0304, 0.0289, 0.0348, 0.0320, 0.0406, 0.0261], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-29 08:16:26,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2151, 4.0693, 4.2618, 4.4080, 4.5343, 4.1158, 4.5216, 4.5519], device='cuda:3'), covar=tensor([0.1329, 0.0879, 0.1244, 0.0593, 0.0515, 0.0829, 0.0460, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0466, 0.0583, 0.0700, 0.0597, 0.0455, 0.0452, 0.0468, 0.0532], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:16:40,044 INFO [train.py:904] (3/8) Epoch 10, batch 9950, loss[loss=0.1713, simple_loss=0.2689, pruned_loss=0.03689, over 16782.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2762, pruned_loss=0.04502, over 3052407.78 frames. ], batch size: 76, lr: 6.68e-03, grad_scale: 4.0 2023-04-29 08:18:13,665 INFO [optim.py:368] (3/8) 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 08:18:42,239 INFO [train.py:904] (3/8) Epoch 10, batch 10000, loss[loss=0.1801, simple_loss=0.2766, pruned_loss=0.04175, over 16838.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2744, pruned_loss=0.04432, over 3066478.35 frames. ], batch size: 124, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:19:16,830 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8014, 2.7274, 2.2397, 4.1552, 3.0144, 3.9067, 1.4628, 2.8632], device='cuda:3'), covar=tensor([0.1119, 0.0571, 0.1142, 0.0094, 0.0147, 0.0422, 0.1372, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0152, 0.0176, 0.0129, 0.0180, 0.0198, 0.0176, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 08:19:54,025 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 08:20:00,742 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5650, 3.6223, 3.3807, 3.1541, 3.2002, 3.5269, 3.2955, 3.3201], device='cuda:3'), covar=tensor([0.0470, 0.0357, 0.0248, 0.0232, 0.0545, 0.0333, 0.1030, 0.0471], device='cuda:3'), in_proj_covar=tensor([0.0220, 0.0272, 0.0258, 0.0237, 0.0273, 0.0270, 0.0177, 0.0299], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-29 08:20:23,603 INFO [train.py:904] (3/8) Epoch 10, batch 10050, loss[loss=0.2013, simple_loss=0.2915, pruned_loss=0.0555, over 16925.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2747, pruned_loss=0.04443, over 3069062.10 frames. ], batch size: 116, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:21:00,574 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4423, 3.4079, 2.7921, 2.0936, 2.2787, 2.2112, 3.6820, 3.1162], device='cuda:3'), covar=tensor([0.2646, 0.0687, 0.1423, 0.2259, 0.2420, 0.1799, 0.0391, 0.1202], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0241, 0.0269, 0.0261, 0.0250, 0.0208, 0.0251, 0.0271], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:21:25,329 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 10, batch 10100, loss[loss=0.1796, simple_loss=0.2658, pruned_loss=0.04665, over 16941.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2751, pruned_loss=0.0452, over 3056074.37 frames. ], batch size: 109, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:22:16,002 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:23:07,594 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2907, 4.0632, 4.3305, 4.4869, 4.6055, 4.1091, 4.6066, 4.5942], device='cuda:3'), covar=tensor([0.1328, 0.0994, 0.1330, 0.0592, 0.0439, 0.1034, 0.0424, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0468, 0.0583, 0.0704, 0.0600, 0.0453, 0.0454, 0.0469, 0.0532], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:23:38,425 INFO [train.py:904] (3/8) Epoch 11, batch 0, loss[loss=0.2421, simple_loss=0.3276, pruned_loss=0.0783, over 16726.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3276, pruned_loss=0.0783, over 16726.00 frames. ], batch size: 57, lr: 6.37e-03, grad_scale: 8.0 2023-04-29 08:23:38,425 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 08:23:45,829 INFO [train.py:938] (3/8) Epoch 11, validation: loss=0.1557, simple_loss=0.2595, pruned_loss=0.02591, over 944034.00 frames. 2023-04-29 08:23:45,829 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 08:24:32,799 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7397, 5.0853, 4.8270, 4.8162, 4.5765, 4.6245, 4.6195, 5.1692], device='cuda:3'), covar=tensor([0.1194, 0.0982, 0.1321, 0.0669, 0.0855, 0.0986, 0.1043, 0.0868], device='cuda:3'), in_proj_covar=tensor([0.0495, 0.0619, 0.0514, 0.0430, 0.0392, 0.0411, 0.0523, 0.0471], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:24:43,125 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.573e+02 2.991e+02 3.932e+02 8.479e+02, threshold=5.981e+02, percent-clipped=4.0 2023-04-29 08:24:55,133 INFO [train.py:904] (3/8) Epoch 11, batch 50, loss[loss=0.2, simple_loss=0.2922, pruned_loss=0.05386, over 17112.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2871, pruned_loss=0.06525, over 746444.60 frames. ], batch size: 48, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:25:21,524 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 08:26:05,598 INFO [train.py:904] (3/8) Epoch 11, batch 100, loss[loss=0.2069, simple_loss=0.2782, pruned_loss=0.06778, over 16790.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2813, pruned_loss=0.0609, over 1316306.41 frames. ], batch size: 96, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:26:49,113 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-29 08:27:03,337 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.523e+02 3.127e+02 3.796e+02 9.987e+02, threshold=6.254e+02, percent-clipped=2.0 2023-04-29 08:27:06,061 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5312, 1.5983, 2.0746, 2.3408, 2.4821, 2.3337, 1.6216, 2.5928], device='cuda:3'), covar=tensor([0.0146, 0.0321, 0.0215, 0.0192, 0.0161, 0.0181, 0.0337, 0.0085], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0168, 0.0152, 0.0153, 0.0164, 0.0117, 0.0169, 0.0107], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 08:27:12,713 INFO [train.py:904] (3/8) Epoch 11, batch 150, loss[loss=0.1926, simple_loss=0.2734, pruned_loss=0.05585, over 16527.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2787, pruned_loss=0.05783, over 1768229.89 frames. ], batch size: 68, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:27:39,334 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 08:27:58,560 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:28:01,460 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 08:28:03,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1171, 4.2841, 4.5186, 3.1906, 3.9479, 4.4598, 4.0342, 2.5807], device='cuda:3'), covar=tensor([0.0316, 0.0065, 0.0028, 0.0268, 0.0086, 0.0053, 0.0059, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0068, 0.0069, 0.0126, 0.0077, 0.0086, 0.0076, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:28:21,561 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5040, 5.8902, 5.6150, 5.6388, 5.1966, 5.1250, 5.3300, 5.9579], device='cuda:3'), covar=tensor([0.1108, 0.0816, 0.0992, 0.0671, 0.0868, 0.0678, 0.0945, 0.0892], device='cuda:3'), in_proj_covar=tensor([0.0519, 0.0645, 0.0538, 0.0450, 0.0409, 0.0429, 0.0548, 0.0493], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:28:23,293 INFO [train.py:904] (3/8) Epoch 11, batch 200, loss[loss=0.1813, simple_loss=0.2745, pruned_loss=0.04402, over 17050.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2796, pruned_loss=0.05846, over 2110424.17 frames. ], batch size: 50, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:21,756 INFO [optim.py:368] (3/8) 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 08:29:22,141 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6880, 3.9180, 4.2404, 2.8419, 3.7815, 4.2112, 3.8973, 2.4415], device='cuda:3'), covar=tensor([0.0405, 0.0065, 0.0032, 0.0320, 0.0081, 0.0067, 0.0065, 0.0386], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0069, 0.0070, 0.0127, 0.0078, 0.0086, 0.0077, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:29:23,293 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 250, loss[loss=0.195, simple_loss=0.2668, pruned_loss=0.06158, over 16688.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2768, pruned_loss=0.05792, over 2378720.96 frames. ], batch size: 134, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:46,107 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 300, loss[loss=0.2171, simple_loss=0.2938, pruned_loss=0.07024, over 16456.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2746, pruned_loss=0.05661, over 2587436.08 frames. ], batch size: 146, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:30:51,058 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:31:35,542 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 350, loss[loss=0.1974, simple_loss=0.2938, pruned_loss=0.05047, over 16681.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2719, pruned_loss=0.05516, over 2758393.34 frames. ], batch size: 62, lr: 6.36e-03, grad_scale: 1.0 2023-04-29 08:32:42,851 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9259, 4.9459, 5.4086, 5.4043, 5.4501, 5.0468, 5.0045, 4.7551], device='cuda:3'), covar=tensor([0.0277, 0.0433, 0.0424, 0.0459, 0.0386, 0.0340, 0.0817, 0.0390], device='cuda:3'), in_proj_covar=tensor([0.0316, 0.0328, 0.0332, 0.0311, 0.0373, 0.0346, 0.0444, 0.0281], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 08:32:54,920 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 08:32:56,331 INFO [train.py:904] (3/8) Epoch 11, batch 400, loss[loss=0.2247, simple_loss=0.2986, pruned_loss=0.07542, over 16458.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2705, pruned_loss=0.05453, over 2875263.04 frames. ], batch size: 146, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:33:00,758 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3192, 4.0592, 4.3252, 4.5414, 4.6355, 4.1710, 4.3905, 4.5709], device='cuda:3'), covar=tensor([0.1264, 0.1079, 0.1301, 0.0618, 0.0520, 0.1094, 0.2176, 0.0634], device='cuda:3'), in_proj_covar=tensor([0.0517, 0.0646, 0.0785, 0.0662, 0.0502, 0.0497, 0.0518, 0.0582], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:33:22,144 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:33:45,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5139, 3.2735, 2.5953, 2.1633, 2.3270, 2.1508, 3.3676, 3.0615], device='cuda:3'), covar=tensor([0.2338, 0.0686, 0.1568, 0.2201, 0.2144, 0.1856, 0.0507, 0.1173], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0253, 0.0281, 0.0272, 0.0270, 0.0220, 0.0265, 0.0290], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:33:54,608 INFO [optim.py:368] (3/8) 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 08:34:05,256 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 08:34:06,160 INFO [train.py:904] (3/8) Epoch 11, batch 450, loss[loss=0.1706, simple_loss=0.2594, pruned_loss=0.04084, over 16644.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.269, pruned_loss=0.0536, over 2975282.97 frames. ], batch size: 62, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:34:46,968 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:34:54,621 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:35:19,216 INFO [train.py:904] (3/8) Epoch 11, batch 500, loss[loss=0.1668, simple_loss=0.257, pruned_loss=0.03827, over 17225.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2671, pruned_loss=0.05242, over 3056822.30 frames. ], batch size: 45, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:35:48,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8908, 4.1846, 3.9894, 4.0251, 3.6545, 3.7966, 3.8370, 4.1639], device='cuda:3'), covar=tensor([0.1136, 0.1048, 0.0986, 0.0698, 0.0796, 0.1615, 0.0949, 0.1128], device='cuda:3'), in_proj_covar=tensor([0.0546, 0.0677, 0.0563, 0.0472, 0.0429, 0.0446, 0.0572, 0.0516], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:36:06,463 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0533, 3.1260, 1.8386, 3.2337, 2.3422, 3.2957, 2.0498, 2.5889], device='cuda:3'), covar=tensor([0.0257, 0.0340, 0.1468, 0.0280, 0.0767, 0.0481, 0.1322, 0.0635], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0162, 0.0187, 0.0127, 0.0167, 0.0204, 0.0196, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 08:36:13,543 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:36:24,789 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:29,128 INFO [train.py:904] (3/8) Epoch 11, batch 550, loss[loss=0.2205, simple_loss=0.2968, pruned_loss=0.07208, over 16730.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2665, pruned_loss=0.05199, over 3107230.55 frames. ], batch size: 134, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:36:37,720 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 08:37:40,170 INFO [train.py:904] (3/8) Epoch 11, batch 600, loss[loss=0.2132, simple_loss=0.2738, pruned_loss=0.07629, over 16648.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2656, pruned_loss=0.05187, over 3155055.68 frames. ], batch size: 124, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:37:43,061 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9094, 3.8658, 4.2352, 1.9839, 4.4210, 4.5110, 3.1120, 3.3627], device='cuda:3'), covar=tensor([0.0630, 0.0189, 0.0180, 0.1089, 0.0046, 0.0094, 0.0401, 0.0350], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0098, 0.0085, 0.0138, 0.0067, 0.0100, 0.0118, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 08:38:38,906 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 650, loss[loss=0.1775, simple_loss=0.2551, pruned_loss=0.04997, over 16936.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.264, pruned_loss=0.05152, over 3182703.95 frames. ], batch size: 109, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:39:03,722 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7073, 2.6425, 2.2009, 2.4843, 2.9784, 2.7553, 3.5170, 3.2564], device='cuda:3'), covar=tensor([0.0061, 0.0282, 0.0353, 0.0320, 0.0185, 0.0264, 0.0157, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0204, 0.0201, 0.0199, 0.0203, 0.0203, 0.0205, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:39:58,906 INFO [train.py:904] (3/8) Epoch 11, batch 700, loss[loss=0.1588, simple_loss=0.236, pruned_loss=0.04076, over 16970.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2636, pruned_loss=0.05084, over 3205632.59 frames. ], batch size: 41, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:40:41,780 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1228, 3.9614, 4.1670, 4.3344, 4.4316, 3.9989, 4.2394, 4.4096], device='cuda:3'), covar=tensor([0.1470, 0.0958, 0.1295, 0.0646, 0.0590, 0.1287, 0.1887, 0.0589], device='cuda:3'), in_proj_covar=tensor([0.0531, 0.0662, 0.0809, 0.0681, 0.0515, 0.0511, 0.0528, 0.0600], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:40:51,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1186, 5.8132, 5.9710, 5.6480, 5.7057, 6.2265, 5.8886, 5.6557], device='cuda:3'), covar=tensor([0.0959, 0.1699, 0.1836, 0.2085, 0.3045, 0.1166, 0.1280, 0.2284], device='cuda:3'), in_proj_covar=tensor([0.0344, 0.0485, 0.0525, 0.0420, 0.0561, 0.0552, 0.0418, 0.0566], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:40:57,198 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 750, loss[loss=0.2142, simple_loss=0.2811, pruned_loss=0.07363, over 16810.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2642, pruned_loss=0.05066, over 3233363.34 frames. ], batch size: 116, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:41:42,360 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:42:18,047 INFO [train.py:904] (3/8) Epoch 11, batch 800, loss[loss=0.1544, simple_loss=0.2343, pruned_loss=0.0372, over 16761.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2645, pruned_loss=0.05069, over 3252076.81 frames. ], batch size: 83, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:42:57,918 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0530, 4.7648, 5.0570, 5.2886, 5.4544, 4.7219, 5.4150, 5.4224], device='cuda:3'), covar=tensor([0.1407, 0.1141, 0.1517, 0.0640, 0.0487, 0.0853, 0.0493, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0525, 0.0658, 0.0805, 0.0678, 0.0512, 0.0511, 0.0526, 0.0597], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:43:11,965 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:15,324 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.524e+02 3.019e+02 3.517e+02 6.392e+02, threshold=6.037e+02, percent-clipped=1.0 2023-04-29 08:43:25,138 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5686, 2.4034, 1.9414, 2.2366, 2.8131, 2.6550, 2.9340, 2.9457], device='cuda:3'), covar=tensor([0.0152, 0.0266, 0.0374, 0.0332, 0.0152, 0.0219, 0.0179, 0.0184], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0206, 0.0202, 0.0201, 0.0205, 0.0204, 0.0208, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:43:27,549 INFO [train.py:904] (3/8) Epoch 11, batch 850, loss[loss=0.1925, simple_loss=0.2642, pruned_loss=0.06035, over 16893.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2633, pruned_loss=0.05016, over 3275994.86 frames. ], batch size: 116, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:17,573 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:44:31,795 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5134, 3.5012, 3.8596, 2.7484, 3.5706, 3.9011, 3.6132, 2.0539], device='cuda:3'), covar=tensor([0.0409, 0.0191, 0.0044, 0.0305, 0.0072, 0.0086, 0.0084, 0.0425], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0071, 0.0071, 0.0128, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:44:37,484 INFO [train.py:904] (3/8) Epoch 11, batch 900, loss[loss=0.1737, simple_loss=0.2529, pruned_loss=0.04728, over 16805.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2628, pruned_loss=0.04953, over 3283821.55 frames. ], batch size: 102, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:45:35,217 INFO [optim.py:368] (3/8) 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 08:45:45,385 INFO [train.py:904] (3/8) Epoch 11, batch 950, loss[loss=0.1611, simple_loss=0.2366, pruned_loss=0.04276, over 15463.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2637, pruned_loss=0.05031, over 3285259.56 frames. ], batch size: 190, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:46:54,282 INFO [train.py:904] (3/8) Epoch 11, batch 1000, loss[loss=0.1874, simple_loss=0.2637, pruned_loss=0.05556, over 12228.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2626, pruned_loss=0.0507, over 3289889.90 frames. ], batch size: 247, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:47:00,252 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8351, 2.8931, 2.4244, 4.1100, 3.1957, 4.0770, 1.7378, 2.9367], device='cuda:3'), covar=tensor([0.1284, 0.0633, 0.1155, 0.0152, 0.0166, 0.0365, 0.1340, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0157, 0.0180, 0.0139, 0.0193, 0.0209, 0.0180, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 08:47:52,027 INFO [optim.py:368] (3/8) 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 08:48:02,251 INFO [train.py:904] (3/8) Epoch 11, batch 1050, loss[loss=0.2166, simple_loss=0.2783, pruned_loss=0.07748, over 16883.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2627, pruned_loss=0.04982, over 3307641.10 frames. ], batch size: 109, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:48:36,414 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:49:12,606 INFO [train.py:904] (3/8) Epoch 11, batch 1100, loss[loss=0.1749, simple_loss=0.266, pruned_loss=0.04191, over 17293.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2618, pruned_loss=0.0497, over 3306377.87 frames. ], batch size: 52, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:49:14,951 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7772, 4.2672, 4.4458, 2.9495, 3.8002, 4.3138, 3.9180, 2.6193], device='cuda:3'), covar=tensor([0.0377, 0.0048, 0.0027, 0.0299, 0.0069, 0.0070, 0.0053, 0.0325], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0070, 0.0070, 0.0126, 0.0078, 0.0088, 0.0077, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:49:43,776 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:50:07,618 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:50:20,434 INFO [train.py:904] (3/8) Epoch 11, batch 1150, loss[loss=0.1691, simple_loss=0.2575, pruned_loss=0.0404, over 17166.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2612, pruned_loss=0.0494, over 3303731.99 frames. ], batch size: 46, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:51:14,266 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 1200, loss[loss=0.165, simple_loss=0.26, pruned_loss=0.03497, over 17259.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2611, pruned_loss=0.04884, over 3318253.67 frames. ], batch size: 52, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:52:02,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9649, 5.0005, 5.4103, 5.4436, 5.4483, 5.1032, 5.0496, 4.7279], device='cuda:3'), covar=tensor([0.0266, 0.0390, 0.0408, 0.0341, 0.0354, 0.0302, 0.0729, 0.0398], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0349, 0.0352, 0.0330, 0.0397, 0.0369, 0.0466, 0.0298], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 08:52:27,631 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.248e+02 2.693e+02 3.149e+02 5.178e+02, threshold=5.386e+02, percent-clipped=0.0 2023-04-29 08:52:39,081 INFO [train.py:904] (3/8) Epoch 11, batch 1250, loss[loss=0.1769, simple_loss=0.2709, pruned_loss=0.04141, over 17057.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2608, pruned_loss=0.04891, over 3323690.94 frames. ], batch size: 55, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:53:06,645 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 08:53:08,810 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 08:53:13,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8370, 2.6865, 2.6591, 1.8748, 2.5586, 2.6972, 2.6494, 1.8606], device='cuda:3'), covar=tensor([0.0345, 0.0081, 0.0056, 0.0300, 0.0094, 0.0084, 0.0080, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0070, 0.0070, 0.0125, 0.0078, 0.0087, 0.0076, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:53:46,791 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5435, 3.5880, 3.3050, 3.0780, 3.1303, 3.4697, 3.2398, 3.2996], device='cuda:3'), covar=tensor([0.0510, 0.0421, 0.0271, 0.0253, 0.0566, 0.0344, 0.1356, 0.0433], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0318, 0.0298, 0.0277, 0.0320, 0.0316, 0.0204, 0.0348], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 08:53:49,408 INFO [train.py:904] (3/8) Epoch 11, batch 1300, loss[loss=0.1478, simple_loss=0.2312, pruned_loss=0.03223, over 17017.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2612, pruned_loss=0.04921, over 3322189.40 frames. ], batch size: 41, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:54:17,671 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-29 08:54:23,681 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5089, 3.9981, 3.9643, 2.1037, 3.3280, 2.7050, 3.9474, 3.9837], device='cuda:3'), covar=tensor([0.0264, 0.0665, 0.0505, 0.1720, 0.0712, 0.0874, 0.0548, 0.0868], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0142, 0.0156, 0.0142, 0.0135, 0.0125, 0.0137, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 08:54:27,470 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5226, 4.8567, 5.1409, 5.1564, 5.1960, 4.8771, 4.4780, 4.4780], device='cuda:3'), covar=tensor([0.0628, 0.0680, 0.0737, 0.0723, 0.0633, 0.0602, 0.1300, 0.0572], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0352, 0.0353, 0.0331, 0.0400, 0.0369, 0.0470, 0.0299], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 08:54:46,485 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.481e+02 2.988e+02 3.715e+02 8.832e+02, threshold=5.975e+02, percent-clipped=5.0 2023-04-29 08:54:57,180 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 1350, loss[loss=0.1887, simple_loss=0.2596, pruned_loss=0.05892, over 16856.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2607, pruned_loss=0.04853, over 3331670.76 frames. ], batch size: 109, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:55:45,171 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8302, 3.8794, 2.0825, 4.1765, 2.7922, 4.1454, 2.4334, 3.0153], device='cuda:3'), covar=tensor([0.0215, 0.0318, 0.1576, 0.0254, 0.0792, 0.0483, 0.1278, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0166, 0.0187, 0.0131, 0.0170, 0.0209, 0.0196, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 08:55:45,175 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:56:05,631 INFO [train.py:904] (3/8) Epoch 11, batch 1400, loss[loss=0.181, simple_loss=0.2554, pruned_loss=0.05333, over 12071.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2612, pruned_loss=0.04854, over 3331191.52 frames. ], batch size: 247, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:56:19,974 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:57:05,116 INFO [optim.py:368] (3/8) 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 08:57:10,130 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 1450, loss[loss=0.1853, simple_loss=0.2535, pruned_loss=0.05857, over 12331.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2605, pruned_loss=0.04844, over 3331800.31 frames. ], batch size: 246, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:16,270 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6643, 4.6448, 4.5964, 4.0982, 4.6075, 1.9108, 4.4021, 4.4541], device='cuda:3'), covar=tensor([0.0086, 0.0066, 0.0124, 0.0280, 0.0077, 0.2111, 0.0105, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0119, 0.0169, 0.0157, 0.0137, 0.0182, 0.0155, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 08:58:25,000 INFO [train.py:904] (3/8) Epoch 11, batch 1500, loss[loss=0.2041, simple_loss=0.3035, pruned_loss=0.05238, over 17055.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2604, pruned_loss=0.04862, over 3320221.20 frames. ], batch size: 50, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:59:24,564 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 1550, loss[loss=0.2056, simple_loss=0.2672, pruned_loss=0.07199, over 16863.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2612, pruned_loss=0.04964, over 3315759.44 frames. ], batch size: 96, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:59:43,338 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 09:00:40,012 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 1600, loss[loss=0.1673, simple_loss=0.2567, pruned_loss=0.03897, over 17117.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2639, pruned_loss=0.05106, over 3308752.19 frames. ], batch size: 47, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:00:57,381 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5334, 2.5784, 1.8529, 2.1155, 2.8950, 2.6519, 3.5362, 3.2481], device='cuda:3'), covar=tensor([0.0107, 0.0355, 0.0536, 0.0446, 0.0246, 0.0330, 0.0163, 0.0194], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0208, 0.0205, 0.0205, 0.0208, 0.0208, 0.0213, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:01:25,950 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:01:42,624 INFO [optim.py:368] (3/8) 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 09:01:53,519 INFO [train.py:904] (3/8) Epoch 11, batch 1650, loss[loss=0.1767, simple_loss=0.2646, pruned_loss=0.04443, over 17105.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2654, pruned_loss=0.05179, over 3310785.63 frames. ], batch size: 47, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:01:57,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6032, 4.5534, 4.4865, 4.2600, 3.9375, 4.5907, 4.4599, 4.1916], device='cuda:3'), covar=tensor([0.0792, 0.0705, 0.0381, 0.0351, 0.1321, 0.0502, 0.0483, 0.0790], device='cuda:3'), in_proj_covar=tensor([0.0260, 0.0325, 0.0303, 0.0283, 0.0326, 0.0322, 0.0208, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:02:03,329 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:50,061 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:02,605 INFO [train.py:904] (3/8) Epoch 11, batch 1700, loss[loss=0.1586, simple_loss=0.2582, pruned_loss=0.02946, over 17154.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2662, pruned_loss=0.0519, over 3312100.83 frames. ], batch size: 48, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:03:10,657 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:41,601 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6696, 3.6237, 3.9205, 1.9721, 4.0975, 4.0679, 3.1393, 2.9921], device='cuda:3'), covar=tensor([0.0749, 0.0172, 0.0158, 0.1164, 0.0054, 0.0130, 0.0358, 0.0418], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0099, 0.0091, 0.0139, 0.0070, 0.0104, 0.0122, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 09:04:01,647 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 2.478e+02 3.058e+02 3.640e+02 7.869e+02, threshold=6.116e+02, percent-clipped=2.0 2023-04-29 09:04:13,825 INFO [train.py:904] (3/8) Epoch 11, batch 1750, loss[loss=0.2031, simple_loss=0.2886, pruned_loss=0.05885, over 17092.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.268, pruned_loss=0.05224, over 3308295.83 frames. ], batch size: 55, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:04:14,277 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0243, 5.0373, 4.8770, 4.5240, 4.2236, 4.9870, 5.0329, 4.4414], device='cuda:3'), covar=tensor([0.0712, 0.0472, 0.0334, 0.0359, 0.1352, 0.0443, 0.0281, 0.0771], device='cuda:3'), in_proj_covar=tensor([0.0259, 0.0324, 0.0303, 0.0282, 0.0325, 0.0321, 0.0207, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:05:22,483 INFO [train.py:904] (3/8) Epoch 11, batch 1800, loss[loss=0.1981, simple_loss=0.2919, pruned_loss=0.05212, over 17042.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2687, pruned_loss=0.05186, over 3312802.55 frames. ], batch size: 53, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:21,366 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.500e+02 2.964e+02 3.680e+02 1.127e+03, threshold=5.929e+02, percent-clipped=5.0 2023-04-29 09:06:31,968 INFO [train.py:904] (3/8) Epoch 11, batch 1850, loss[loss=0.1641, simple_loss=0.2511, pruned_loss=0.03852, over 17199.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2691, pruned_loss=0.05163, over 3312557.93 frames. ], batch size: 46, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:50,810 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:07:24,696 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-04-29 09:07:39,138 INFO [train.py:904] (3/8) Epoch 11, batch 1900, loss[loss=0.194, simple_loss=0.2875, pruned_loss=0.05025, over 16712.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2682, pruned_loss=0.0508, over 3318332.31 frames. ], batch size: 57, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:08:16,140 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:08:40,923 INFO [optim.py:368] (3/8) 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 09:08:51,896 INFO [train.py:904] (3/8) Epoch 11, batch 1950, loss[loss=0.1456, simple_loss=0.2256, pruned_loss=0.03277, over 16772.00 frames. ], tot_loss[loss=0.184, simple_loss=0.268, pruned_loss=0.05005, over 3314908.53 frames. ], batch size: 39, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:08:55,093 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:09:41,328 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:00,140 INFO [train.py:904] (3/8) Epoch 11, batch 2000, loss[loss=0.2109, simple_loss=0.2734, pruned_loss=0.07418, over 16877.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2678, pruned_loss=0.05022, over 3319370.31 frames. ], batch size: 109, lr: 6.31e-03, grad_scale: 8.0 2023-04-29 09:10:07,701 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:30,356 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2129, 2.0027, 2.2379, 3.9053, 2.0664, 2.4246, 2.1447, 2.2555], device='cuda:3'), covar=tensor([0.1086, 0.3318, 0.2108, 0.0413, 0.3392, 0.2199, 0.3057, 0.2666], device='cuda:3'), in_proj_covar=tensor([0.0366, 0.0391, 0.0330, 0.0322, 0.0411, 0.0447, 0.0353, 0.0461], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:10:58,944 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:11:00,876 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.374e+02 2.796e+02 3.607e+02 7.066e+02, threshold=5.592e+02, percent-clipped=4.0 2023-04-29 09:11:10,863 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 09:11:11,321 INFO [train.py:904] (3/8) Epoch 11, batch 2050, loss[loss=0.1886, simple_loss=0.2825, pruned_loss=0.04734, over 17115.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.268, pruned_loss=0.05014, over 3308340.53 frames. ], batch size: 49, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:11:16,434 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:05,786 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:21,597 INFO [train.py:904] (3/8) Epoch 11, batch 2100, loss[loss=0.1952, simple_loss=0.2881, pruned_loss=0.05116, over 17047.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2698, pruned_loss=0.05142, over 3288996.09 frames. ], batch size: 55, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:15,835 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 09:13:22,846 INFO [optim.py:368] (3/8) 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 09:13:31,361 INFO [train.py:904] (3/8) Epoch 11, batch 2150, loss[loss=0.1622, simple_loss=0.2442, pruned_loss=0.04007, over 16994.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2703, pruned_loss=0.05224, over 3292714.64 frames. ], batch size: 41, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:13:39,833 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 09:13:52,859 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 09:14:42,107 INFO [train.py:904] (3/8) Epoch 11, batch 2200, loss[loss=0.1967, simple_loss=0.2678, pruned_loss=0.06278, over 16790.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2706, pruned_loss=0.05213, over 3303621.26 frames. ], batch size: 124, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:10,360 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:15:35,187 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 09:15:51,098 INFO [train.py:904] (3/8) Epoch 11, batch 2250, loss[loss=0.1797, simple_loss=0.2594, pruned_loss=0.04998, over 15955.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2724, pruned_loss=0.05273, over 3303424.42 frames. ], batch size: 35, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:54,349 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:12,398 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:44,193 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:01,622 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 2300, loss[loss=0.1953, simple_loss=0.2682, pruned_loss=0.06122, over 16861.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2721, pruned_loss=0.053, over 3296394.66 frames. ], batch size: 96, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:17:02,657 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:08,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6456, 4.9986, 4.7661, 4.7590, 4.5029, 4.4334, 4.4743, 5.0409], device='cuda:3'), covar=tensor([0.1169, 0.0849, 0.0915, 0.0650, 0.0744, 0.1036, 0.0994, 0.0863], device='cuda:3'), in_proj_covar=tensor([0.0544, 0.0687, 0.0565, 0.0479, 0.0429, 0.0442, 0.0572, 0.0521], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:17:35,790 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:39,025 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 09:17:45,609 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 09:17:48,570 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.457e+02 2.837e+02 3.610e+02 7.865e+02, threshold=5.674e+02, percent-clipped=5.0 2023-04-29 09:18:11,650 INFO [train.py:904] (3/8) Epoch 11, batch 2350, loss[loss=0.2023, simple_loss=0.272, pruned_loss=0.06631, over 16927.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2736, pruned_loss=0.05392, over 3303973.84 frames. ], batch size: 90, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:18:45,263 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:19:19,657 INFO [train.py:904] (3/8) Epoch 11, batch 2400, loss[loss=0.2151, simple_loss=0.2936, pruned_loss=0.06828, over 16387.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2744, pruned_loss=0.05406, over 3305019.52 frames. ], batch size: 165, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:09,892 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:20:20,789 INFO [optim.py:368] (3/8) 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 09:20:28,624 INFO [train.py:904] (3/8) Epoch 11, batch 2450, loss[loss=0.1995, simple_loss=0.2881, pruned_loss=0.05542, over 16609.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2734, pruned_loss=0.0527, over 3310468.78 frames. ], batch size: 62, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:33,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8946, 5.3053, 5.4783, 5.2359, 5.2645, 5.8787, 5.4516, 5.2317], device='cuda:3'), covar=tensor([0.1018, 0.1809, 0.1828, 0.1970, 0.2755, 0.0989, 0.1166, 0.2085], device='cuda:3'), in_proj_covar=tensor([0.0358, 0.0501, 0.0545, 0.0439, 0.0576, 0.0571, 0.0428, 0.0588], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:21:22,262 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:31,627 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:42,814 INFO [train.py:904] (3/8) Epoch 11, batch 2500, loss[loss=0.2067, simple_loss=0.3001, pruned_loss=0.05667, over 17048.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2734, pruned_loss=0.05237, over 3314386.44 frames. ], batch size: 53, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:11,628 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:22:38,470 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9346, 4.1482, 2.3976, 4.6512, 2.9926, 4.5573, 2.5203, 3.3271], device='cuda:3'), covar=tensor([0.0187, 0.0276, 0.1378, 0.0194, 0.0723, 0.0420, 0.1372, 0.0586], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0167, 0.0187, 0.0134, 0.0167, 0.0211, 0.0193, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 09:22:43,745 INFO [optim.py:368] (3/8) 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 09:22:47,569 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:49,942 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:50,606 INFO [train.py:904] (3/8) Epoch 11, batch 2550, loss[loss=0.1652, simple_loss=0.2518, pruned_loss=0.03926, over 15791.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2734, pruned_loss=0.05288, over 3318030.27 frames. ], batch size: 35, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:59,816 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:00,166 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 09:23:08,317 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 09:23:15,995 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:23:38,136 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 09:23:47,227 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6718, 2.5504, 2.1785, 2.3996, 2.8930, 2.6784, 3.4583, 3.1739], device='cuda:3'), covar=tensor([0.0077, 0.0301, 0.0367, 0.0321, 0.0187, 0.0269, 0.0180, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0208, 0.0201, 0.0202, 0.0206, 0.0206, 0.0215, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:23:51,813 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 2600, loss[loss=0.1973, simple_loss=0.2807, pruned_loss=0.05692, over 16542.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2726, pruned_loss=0.0524, over 3325677.46 frames. ], batch size: 68, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:24:12,083 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:24:26,686 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:25:01,155 INFO [optim.py:368] (3/8) 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 09:25:10,377 INFO [train.py:904] (3/8) Epoch 11, batch 2650, loss[loss=0.1901, simple_loss=0.2731, pruned_loss=0.05352, over 15786.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2736, pruned_loss=0.05257, over 3325402.77 frames. ], batch size: 35, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:26:07,025 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2626, 4.5369, 4.5200, 3.3881, 3.8097, 4.4581, 4.0936, 2.6641], device='cuda:3'), covar=tensor([0.0288, 0.0045, 0.0030, 0.0226, 0.0079, 0.0064, 0.0047, 0.0320], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0125, 0.0079, 0.0089, 0.0077, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:26:18,840 INFO [train.py:904] (3/8) Epoch 11, batch 2700, loss[loss=0.1684, simple_loss=0.2576, pruned_loss=0.03965, over 16552.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2729, pruned_loss=0.05171, over 3321069.56 frames. ], batch size: 75, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:27:00,692 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:27:19,005 INFO [optim.py:368] (3/8) 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 09:27:27,258 INFO [train.py:904] (3/8) Epoch 11, batch 2750, loss[loss=0.1673, simple_loss=0.2554, pruned_loss=0.03956, over 17200.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2734, pruned_loss=0.05127, over 3328331.89 frames. ], batch size: 46, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:28:16,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1397, 4.2742, 3.2442, 2.4833, 3.2428, 2.8497, 4.7323, 4.0153], device='cuda:3'), covar=tensor([0.2315, 0.0716, 0.1526, 0.2096, 0.2189, 0.1602, 0.0367, 0.0967], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0257, 0.0282, 0.0277, 0.0285, 0.0222, 0.0268, 0.0301], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:28:36,571 INFO [train.py:904] (3/8) Epoch 11, batch 2800, loss[loss=0.1853, simple_loss=0.2768, pruned_loss=0.04695, over 17063.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2734, pruned_loss=0.05141, over 3325921.71 frames. ], batch size: 50, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:37,339 INFO [optim.py:368] (3/8) 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 09:29:37,600 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:29:44,351 INFO [train.py:904] (3/8) Epoch 11, batch 2850, loss[loss=0.1679, simple_loss=0.2521, pruned_loss=0.0419, over 17029.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2721, pruned_loss=0.05141, over 3320198.84 frames. ], batch size: 41, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:45,809 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:17,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3083, 5.2900, 5.1363, 4.5429, 5.1122, 2.2060, 4.8621, 5.1150], device='cuda:3'), covar=tensor([0.0067, 0.0061, 0.0132, 0.0343, 0.0076, 0.1962, 0.0107, 0.0144], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0121, 0.0170, 0.0160, 0.0139, 0.0180, 0.0157, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:30:41,697 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6295, 4.6990, 4.8763, 4.8270, 4.7383, 5.3151, 4.8841, 4.6468], device='cuda:3'), covar=tensor([0.1470, 0.2204, 0.2040, 0.2093, 0.2984, 0.1132, 0.1357, 0.2427], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0499, 0.0542, 0.0434, 0.0576, 0.0568, 0.0427, 0.0584], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:30:44,695 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:52,870 INFO [train.py:904] (3/8) Epoch 11, batch 2900, loss[loss=0.1882, simple_loss=0.2644, pruned_loss=0.05601, over 16029.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2712, pruned_loss=0.0527, over 3309138.78 frames. ], batch size: 35, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:30:57,321 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:19,952 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:52,197 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 09:32:04,092 INFO [train.py:904] (3/8) Epoch 11, batch 2950, loss[loss=0.1907, simple_loss=0.256, pruned_loss=0.06274, over 16864.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2708, pruned_loss=0.05362, over 3306383.04 frames. ], batch size: 90, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:32:27,833 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:33:12,380 INFO [train.py:904] (3/8) Epoch 11, batch 3000, loss[loss=0.1828, simple_loss=0.2685, pruned_loss=0.04854, over 15946.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2714, pruned_loss=0.05446, over 3298117.86 frames. ], batch size: 35, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:33:12,381 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 09:33:22,063 INFO [train.py:938] (3/8) Epoch 11, validation: loss=0.1413, simple_loss=0.2475, pruned_loss=0.01754, over 944034.00 frames. 2023-04-29 09:33:22,064 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 09:33:57,260 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8736, 4.6189, 4.6853, 5.0878, 5.2404, 4.6437, 5.3231, 5.2491], device='cuda:3'), covar=tensor([0.1768, 0.1340, 0.2129, 0.0881, 0.0780, 0.0891, 0.0653, 0.0757], device='cuda:3'), in_proj_covar=tensor([0.0556, 0.0696, 0.0854, 0.0713, 0.0537, 0.0547, 0.0551, 0.0631], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:34:04,215 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:34:20,848 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.544e+02 3.030e+02 3.628e+02 6.112e+02, threshold=6.060e+02, percent-clipped=1.0 2023-04-29 09:34:30,328 INFO [train.py:904] (3/8) Epoch 11, batch 3050, loss[loss=0.1987, simple_loss=0.2699, pruned_loss=0.06381, over 16420.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2707, pruned_loss=0.05373, over 3302481.70 frames. ], batch size: 146, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:35:07,353 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:35:37,117 INFO [train.py:904] (3/8) Epoch 11, batch 3100, loss[loss=0.1942, simple_loss=0.2581, pruned_loss=0.0651, over 16849.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2702, pruned_loss=0.05371, over 3298467.48 frames. ], batch size: 109, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:03,713 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-29 09:36:26,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3531, 4.2072, 4.3916, 4.5746, 4.7075, 4.3106, 4.4562, 4.6657], device='cuda:3'), covar=tensor([0.1415, 0.0972, 0.1391, 0.0677, 0.0556, 0.0891, 0.1447, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0551, 0.0689, 0.0850, 0.0704, 0.0530, 0.0541, 0.0544, 0.0626], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:36:39,246 INFO [optim.py:368] (3/8) 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 09:36:39,612 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:36:43,538 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 09:36:47,518 INFO [train.py:904] (3/8) Epoch 11, batch 3150, loss[loss=0.1941, simple_loss=0.2666, pruned_loss=0.06084, over 16529.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2694, pruned_loss=0.05317, over 3304633.89 frames. ], batch size: 146, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:49,128 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:18,550 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2277, 2.0713, 1.6286, 1.9266, 2.3884, 2.2814, 2.4660, 2.5597], device='cuda:3'), covar=tensor([0.0154, 0.0282, 0.0384, 0.0335, 0.0161, 0.0242, 0.0176, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0208, 0.0202, 0.0203, 0.0208, 0.0205, 0.0217, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:37:46,270 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:56,593 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:57,453 INFO [train.py:904] (3/8) Epoch 11, batch 3200, loss[loss=0.1835, simple_loss=0.2689, pruned_loss=0.04906, over 17096.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2684, pruned_loss=0.05211, over 3296474.10 frames. ], batch size: 47, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:38:01,354 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:38:16,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7757, 4.3184, 3.0707, 2.1819, 2.9474, 2.4262, 4.5700, 3.8574], device='cuda:3'), covar=tensor([0.2571, 0.0552, 0.1718, 0.2544, 0.2356, 0.1757, 0.0392, 0.1024], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0259, 0.0284, 0.0277, 0.0287, 0.0224, 0.0270, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:38:56,072 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6146, 3.6935, 3.9281, 1.9299, 3.9996, 4.0768, 3.1331, 2.9466], device='cuda:3'), covar=tensor([0.0683, 0.0164, 0.0169, 0.1111, 0.0070, 0.0125, 0.0380, 0.0403], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0098, 0.0090, 0.0137, 0.0070, 0.0104, 0.0121, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 09:38:57,881 INFO [optim.py:368] (3/8) 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 09:39:02,777 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8802, 5.1985, 4.8905, 4.9362, 4.6913, 4.6453, 4.6122, 5.2285], device='cuda:3'), covar=tensor([0.1048, 0.0738, 0.1006, 0.0692, 0.0763, 0.0994, 0.1101, 0.0850], device='cuda:3'), in_proj_covar=tensor([0.0567, 0.0706, 0.0588, 0.0495, 0.0444, 0.0453, 0.0592, 0.0542], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:39:06,600 INFO [train.py:904] (3/8) Epoch 11, batch 3250, loss[loss=0.1541, simple_loss=0.2473, pruned_loss=0.03051, over 17092.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2682, pruned_loss=0.05156, over 3304455.87 frames. ], batch size: 47, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:39:08,055 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:39:09,793 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8069, 3.5445, 3.8398, 3.5774, 3.7064, 4.1633, 3.8533, 3.5441], device='cuda:3'), covar=tensor([0.2175, 0.2575, 0.2158, 0.2591, 0.3223, 0.2173, 0.1603, 0.2826], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0510, 0.0550, 0.0440, 0.0588, 0.0578, 0.0434, 0.0597], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:40:15,729 INFO [train.py:904] (3/8) Epoch 11, batch 3300, loss[loss=0.196, simple_loss=0.281, pruned_loss=0.05553, over 16490.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2691, pruned_loss=0.0515, over 3296334.28 frames. ], batch size: 68, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:41:16,318 INFO [optim.py:368] (3/8) 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 09:41:16,688 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0129, 5.0768, 5.6079, 5.6046, 5.5536, 5.1396, 5.1389, 4.8462], device='cuda:3'), covar=tensor([0.0317, 0.0481, 0.0368, 0.0451, 0.0455, 0.0350, 0.0881, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0342, 0.0357, 0.0357, 0.0335, 0.0404, 0.0373, 0.0478, 0.0301], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 09:41:24,650 INFO [train.py:904] (3/8) Epoch 11, batch 3350, loss[loss=0.1742, simple_loss=0.2595, pruned_loss=0.04451, over 17193.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2698, pruned_loss=0.05193, over 3302192.45 frames. ], batch size: 44, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:42:05,002 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1597, 4.2746, 4.1295, 3.9520, 3.5448, 4.3665, 4.0796, 3.9118], device='cuda:3'), covar=tensor([0.0892, 0.0636, 0.0405, 0.0356, 0.1372, 0.0412, 0.0769, 0.0626], device='cuda:3'), in_proj_covar=tensor([0.0265, 0.0335, 0.0314, 0.0292, 0.0335, 0.0332, 0.0214, 0.0363], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:42:15,954 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4675, 1.9172, 2.8379, 3.2320, 3.0829, 3.8239, 2.2373, 3.6848], device='cuda:3'), covar=tensor([0.0103, 0.0351, 0.0180, 0.0164, 0.0157, 0.0094, 0.0371, 0.0075], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0158, 0.0161, 0.0170, 0.0126, 0.0170, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 09:42:20,174 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-29 09:42:33,960 INFO [train.py:904] (3/8) Epoch 11, batch 3400, loss[loss=0.2113, simple_loss=0.2959, pruned_loss=0.06333, over 16643.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2693, pruned_loss=0.05173, over 3309667.31 frames. ], batch size: 62, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:10,161 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4077, 5.2712, 5.1910, 4.7848, 4.7662, 5.2519, 5.2246, 4.8267], device='cuda:3'), covar=tensor([0.0501, 0.0416, 0.0263, 0.0287, 0.1164, 0.0351, 0.0252, 0.0720], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0339, 0.0316, 0.0294, 0.0339, 0.0335, 0.0216, 0.0367], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:43:33,846 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 3450, loss[loss=0.1715, simple_loss=0.26, pruned_loss=0.0415, over 17101.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2672, pruned_loss=0.0505, over 3316184.63 frames. ], batch size: 47, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:46,676 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 09:44:52,762 INFO [train.py:904] (3/8) Epoch 11, batch 3500, loss[loss=0.2209, simple_loss=0.2933, pruned_loss=0.07425, over 16224.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2671, pruned_loss=0.05055, over 3312317.76 frames. ], batch size: 165, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:45:10,654 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2407, 3.4619, 3.3400, 2.0777, 2.9423, 2.4760, 3.6136, 3.6421], device='cuda:3'), covar=tensor([0.0193, 0.0710, 0.0559, 0.1601, 0.0704, 0.0880, 0.0479, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0147, 0.0156, 0.0142, 0.0135, 0.0124, 0.0136, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 09:45:27,110 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-29 09:45:55,145 INFO [optim.py:368] (3/8) 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 09:46:03,266 INFO [train.py:904] (3/8) Epoch 11, batch 3550, loss[loss=0.1813, simple_loss=0.2611, pruned_loss=0.05078, over 15408.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2651, pruned_loss=0.04955, over 3304702.87 frames. ], batch size: 191, lr: 6.27e-03, grad_scale: 4.0 2023-04-29 09:46:54,486 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 09:47:12,589 INFO [train.py:904] (3/8) Epoch 11, batch 3600, loss[loss=0.2168, simple_loss=0.2744, pruned_loss=0.07959, over 16735.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2647, pruned_loss=0.05007, over 3294632.62 frames. ], batch size: 124, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:47:58,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8763, 5.0867, 5.2900, 5.1079, 5.0947, 5.7172, 5.2924, 4.9544], device='cuda:3'), covar=tensor([0.1057, 0.1788, 0.2020, 0.1826, 0.2633, 0.1021, 0.1439, 0.2331], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0499, 0.0540, 0.0431, 0.0572, 0.0568, 0.0428, 0.0585], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:48:17,992 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 3650, loss[loss=0.1706, simple_loss=0.2381, pruned_loss=0.05154, over 16909.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.264, pruned_loss=0.05077, over 3300052.37 frames. ], batch size: 96, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:49:22,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5254, 4.5834, 4.9403, 4.9571, 4.9528, 4.6299, 4.6299, 4.4075], device='cuda:3'), covar=tensor([0.0279, 0.0438, 0.0316, 0.0312, 0.0335, 0.0295, 0.0709, 0.0428], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0355, 0.0355, 0.0333, 0.0400, 0.0371, 0.0475, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 09:49:32,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5996, 4.6400, 4.9826, 5.0081, 5.0111, 4.6736, 4.6664, 4.4417], device='cuda:3'), covar=tensor([0.0291, 0.0508, 0.0380, 0.0353, 0.0368, 0.0326, 0.0705, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0355, 0.0356, 0.0332, 0.0400, 0.0371, 0.0474, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 09:49:37,393 INFO [train.py:904] (3/8) Epoch 11, batch 3700, loss[loss=0.1957, simple_loss=0.2772, pruned_loss=0.0571, over 16714.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2631, pruned_loss=0.05198, over 3281646.90 frames. ], batch size: 57, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:50:41,009 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 3750, loss[loss=0.1726, simple_loss=0.2411, pruned_loss=0.05201, over 16569.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2641, pruned_loss=0.05382, over 3279445.54 frames. ], batch size: 75, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:51:57,076 INFO [train.py:904] (3/8) Epoch 11, batch 3800, loss[loss=0.189, simple_loss=0.2638, pruned_loss=0.05715, over 16880.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2654, pruned_loss=0.05536, over 3278165.03 frames. ], batch size: 102, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:53:02,333 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 3850, loss[loss=0.204, simple_loss=0.2768, pruned_loss=0.06563, over 17036.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2657, pruned_loss=0.05602, over 3282230.04 frames. ], batch size: 53, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:54:00,665 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:54:11,032 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9296, 3.1463, 3.0624, 1.9775, 2.7204, 2.2305, 3.4510, 3.3152], device='cuda:3'), covar=tensor([0.0238, 0.0765, 0.0638, 0.1682, 0.0810, 0.0913, 0.0506, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0148, 0.0157, 0.0144, 0.0135, 0.0125, 0.0137, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 09:54:19,876 INFO [train.py:904] (3/8) Epoch 11, batch 3900, loss[loss=0.1858, simple_loss=0.2613, pruned_loss=0.05517, over 16806.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2647, pruned_loss=0.05622, over 3288702.76 frames. ], batch size: 83, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:54:24,032 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1126, 3.2239, 3.3892, 2.2088, 3.0277, 3.4395, 3.1072, 1.9886], device='cuda:3'), covar=tensor([0.0396, 0.0075, 0.0038, 0.0308, 0.0086, 0.0062, 0.0071, 0.0354], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0125, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:55:25,282 INFO [optim.py:368] (3/8) 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 09:55:28,924 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:55:31,848 INFO [train.py:904] (3/8) Epoch 11, batch 3950, loss[loss=0.1889, simple_loss=0.2643, pruned_loss=0.05672, over 16493.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2644, pruned_loss=0.05712, over 3290290.45 frames. ], batch size: 68, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:55:51,647 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7741, 4.7139, 4.6876, 4.1949, 4.7096, 1.9018, 4.5301, 4.4742], device='cuda:3'), covar=tensor([0.0079, 0.0061, 0.0125, 0.0257, 0.0075, 0.2159, 0.0101, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0120, 0.0169, 0.0161, 0.0140, 0.0180, 0.0158, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 09:56:21,757 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 4000, loss[loss=0.1972, simple_loss=0.2722, pruned_loss=0.06107, over 16838.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2645, pruned_loss=0.05748, over 3292356.16 frames. ], batch size: 90, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:44,492 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:35,444 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 09:57:39,618 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0893, 3.7173, 3.8000, 2.2898, 3.4268, 3.6651, 3.4850, 1.7214], device='cuda:3'), covar=tensor([0.0477, 0.0055, 0.0051, 0.0383, 0.0077, 0.0125, 0.0078, 0.0486], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0124, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:57:48,231 INFO [optim.py:368] (3/8) 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 09:57:48,827 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:55,346 INFO [train.py:904] (3/8) Epoch 11, batch 4050, loss[loss=0.1972, simple_loss=0.2768, pruned_loss=0.05876, over 12609.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2646, pruned_loss=0.05629, over 3285081.34 frames. ], batch size: 247, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:57:58,226 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8146, 2.7792, 2.6622, 1.9098, 2.5803, 2.7751, 2.6414, 1.8169], device='cuda:3'), covar=tensor([0.0361, 0.0056, 0.0045, 0.0295, 0.0077, 0.0071, 0.0072, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0124, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 09:58:11,041 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:59:08,742 INFO [train.py:904] (3/8) Epoch 11, batch 4100, loss[loss=0.1939, simple_loss=0.2712, pruned_loss=0.05827, over 17003.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2662, pruned_loss=0.05556, over 3278549.45 frames. ], batch size: 50, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:00:18,931 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 2.197e+02 2.575e+02 3.230e+02 7.189e+02, threshold=5.150e+02, percent-clipped=4.0 2023-04-29 10:00:26,723 INFO [train.py:904] (3/8) Epoch 11, batch 4150, loss[loss=0.2008, simple_loss=0.2874, pruned_loss=0.05715, over 16473.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2741, pruned_loss=0.05824, over 3244772.90 frames. ], batch size: 68, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:00:40,653 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9031, 4.1776, 3.9331, 4.0202, 3.6264, 3.7454, 3.8383, 4.1280], device='cuda:3'), covar=tensor([0.1141, 0.0933, 0.1001, 0.0664, 0.0835, 0.1669, 0.0881, 0.1066], device='cuda:3'), in_proj_covar=tensor([0.0552, 0.0684, 0.0569, 0.0478, 0.0434, 0.0444, 0.0573, 0.0524], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:01:31,382 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8588, 4.1611, 3.9605, 4.0237, 3.6118, 3.7196, 3.8010, 4.1192], device='cuda:3'), covar=tensor([0.1060, 0.0883, 0.0905, 0.0647, 0.0823, 0.1798, 0.0829, 0.1060], device='cuda:3'), in_proj_covar=tensor([0.0547, 0.0676, 0.0563, 0.0473, 0.0429, 0.0440, 0.0567, 0.0519], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:01:43,951 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5972, 4.9046, 4.6982, 4.7126, 4.3968, 4.3706, 4.3376, 4.9594], device='cuda:3'), covar=tensor([0.1010, 0.0793, 0.0821, 0.0665, 0.0784, 0.1061, 0.0962, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0546, 0.0676, 0.0563, 0.0473, 0.0429, 0.0440, 0.0566, 0.0519], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:01:44,702 INFO [train.py:904] (3/8) Epoch 11, batch 4200, loss[loss=0.2141, simple_loss=0.3024, pruned_loss=0.0629, over 16426.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2812, pruned_loss=0.0602, over 3207292.56 frames. ], batch size: 146, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:02:14,473 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4904, 3.6762, 2.0248, 4.0168, 2.6722, 4.0406, 2.2436, 2.9037], device='cuda:3'), covar=tensor([0.0260, 0.0328, 0.1524, 0.0246, 0.0731, 0.0495, 0.1465, 0.0677], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0165, 0.0186, 0.0131, 0.0166, 0.0210, 0.0194, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 10:02:27,641 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4905, 2.0357, 1.6836, 1.8584, 2.3577, 2.1210, 2.3807, 2.5260], device='cuda:3'), covar=tensor([0.0095, 0.0288, 0.0376, 0.0361, 0.0177, 0.0291, 0.0141, 0.0194], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0206, 0.0201, 0.0202, 0.0206, 0.0204, 0.0211, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:02:49,762 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 11, batch 4250, loss[loss=0.2093, simple_loss=0.2949, pruned_loss=0.06189, over 15368.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2842, pruned_loss=0.06018, over 3178403.70 frames. ], batch size: 190, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:03:59,172 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:03:59,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2342, 2.0742, 2.0737, 3.9786, 1.9655, 2.5339, 2.2328, 2.2701], device='cuda:3'), covar=tensor([0.1006, 0.3273, 0.2344, 0.0406, 0.3741, 0.2147, 0.2784, 0.3331], device='cuda:3'), in_proj_covar=tensor([0.0363, 0.0392, 0.0326, 0.0321, 0.0409, 0.0452, 0.0355, 0.0463], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:04:08,523 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 4300, loss[loss=0.2215, simple_loss=0.3107, pruned_loss=0.06618, over 15250.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2845, pruned_loss=0.05905, over 3166303.06 frames. ], batch size: 190, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:05:11,563 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:17,304 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 4350, loss[loss=0.2215, simple_loss=0.307, pruned_loss=0.06803, over 16721.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2885, pruned_loss=0.06036, over 3173497.82 frames. ], batch size: 124, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:05:27,930 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:34,724 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:37,394 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:06:38,330 INFO [train.py:904] (3/8) Epoch 11, batch 4400, loss[loss=0.2296, simple_loss=0.3121, pruned_loss=0.0735, over 17028.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2913, pruned_loss=0.06184, over 3168327.68 frames. ], batch size: 55, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:07:32,180 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 10:07:40,705 INFO [optim.py:368] (3/8) 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 10:07:48,893 INFO [train.py:904] (3/8) Epoch 11, batch 4450, loss[loss=0.2068, simple_loss=0.2945, pruned_loss=0.05953, over 17018.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2952, pruned_loss=0.06326, over 3172520.97 frames. ], batch size: 41, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:09:04,900 INFO [train.py:904] (3/8) Epoch 11, batch 4500, loss[loss=0.2063, simple_loss=0.2838, pruned_loss=0.06436, over 17122.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2955, pruned_loss=0.06367, over 3175192.53 frames. ], batch size: 47, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:09:51,336 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-29 10:10:04,913 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:10:05,267 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-29 10:10:09,391 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 1.945e+02 2.292e+02 2.586e+02 4.524e+02, threshold=4.584e+02, percent-clipped=0.0 2023-04-29 10:10:17,383 INFO [train.py:904] (3/8) Epoch 11, batch 4550, loss[loss=0.1999, simple_loss=0.284, pruned_loss=0.05791, over 16649.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2962, pruned_loss=0.06421, over 3189654.83 frames. ], batch size: 57, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:10:53,578 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:15,423 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:29,157 INFO [train.py:904] (3/8) Epoch 11, batch 4600, loss[loss=0.188, simple_loss=0.279, pruned_loss=0.0485, over 17126.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.296, pruned_loss=0.06331, over 3217371.89 frames. ], batch size: 49, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:11:46,331 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5506, 4.5071, 4.2679, 3.5294, 4.4024, 1.5050, 4.1632, 3.9778], device='cuda:3'), covar=tensor([0.0061, 0.0056, 0.0130, 0.0351, 0.0063, 0.2679, 0.0096, 0.0201], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0116, 0.0163, 0.0156, 0.0134, 0.0175, 0.0150, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:12:15,339 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 10:12:22,258 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:28,004 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:31,851 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2915, 2.4591, 1.9438, 2.3645, 2.8979, 2.5012, 3.1066, 3.0899], device='cuda:3'), covar=tensor([0.0059, 0.0287, 0.0407, 0.0295, 0.0163, 0.0262, 0.0147, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0204, 0.0199, 0.0198, 0.0202, 0.0203, 0.0208, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:12:35,552 INFO [optim.py:368] (3/8) 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 10:12:37,645 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:38,726 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8783, 4.1513, 3.9289, 3.9964, 3.6172, 3.7623, 3.8114, 4.1201], device='cuda:3'), covar=tensor([0.1113, 0.0840, 0.0999, 0.0662, 0.0765, 0.1593, 0.0848, 0.1068], device='cuda:3'), in_proj_covar=tensor([0.0526, 0.0655, 0.0541, 0.0454, 0.0416, 0.0422, 0.0544, 0.0500], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:12:41,945 INFO [train.py:904] (3/8) Epoch 11, batch 4650, loss[loss=0.1975, simple_loss=0.282, pruned_loss=0.05654, over 16460.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2945, pruned_loss=0.0628, over 3214301.23 frames. ], batch size: 75, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:12:47,341 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:51,746 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:37,561 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:55,735 INFO [train.py:904] (3/8) Epoch 11, batch 4700, loss[loss=0.1991, simple_loss=0.2782, pruned_loss=0.06003, over 11650.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.292, pruned_loss=0.0617, over 3214120.14 frames. ], batch size: 247, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:14:01,835 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:14:49,653 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8189, 4.7739, 4.7226, 3.7870, 4.7037, 1.5953, 4.4558, 4.4567], device='cuda:3'), covar=tensor([0.0086, 0.0081, 0.0117, 0.0485, 0.0090, 0.2411, 0.0117, 0.0208], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0116, 0.0163, 0.0157, 0.0133, 0.0176, 0.0150, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:15:01,675 INFO [optim.py:368] (3/8) 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 10:15:04,001 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6693, 4.8998, 4.5946, 4.3400, 3.7096, 4.7995, 4.7349, 4.3108], device='cuda:3'), covar=tensor([0.0873, 0.0510, 0.0377, 0.0332, 0.1869, 0.0434, 0.0319, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0297, 0.0278, 0.0259, 0.0300, 0.0295, 0.0192, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:15:09,055 INFO [train.py:904] (3/8) Epoch 11, batch 4750, loss[loss=0.1893, simple_loss=0.2687, pruned_loss=0.05493, over 16161.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.288, pruned_loss=0.05968, over 3201197.46 frames. ], batch size: 35, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:15:32,290 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:15:45,217 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0634, 3.3989, 3.5683, 1.7261, 3.7140, 3.8421, 2.7433, 2.8510], device='cuda:3'), covar=tensor([0.0895, 0.0200, 0.0213, 0.1216, 0.0079, 0.0094, 0.0432, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0100, 0.0090, 0.0138, 0.0070, 0.0102, 0.0121, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 10:16:07,504 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4521, 5.7292, 5.4468, 5.5772, 5.1816, 5.0363, 5.2064, 5.8625], device='cuda:3'), covar=tensor([0.0996, 0.0744, 0.0896, 0.0634, 0.0647, 0.0556, 0.0886, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0528, 0.0655, 0.0543, 0.0453, 0.0415, 0.0423, 0.0542, 0.0501], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:16:14,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5566, 4.1538, 4.0952, 2.6365, 3.5113, 4.0299, 3.7092, 2.1926], device='cuda:3'), covar=tensor([0.0374, 0.0024, 0.0025, 0.0307, 0.0081, 0.0075, 0.0067, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0069, 0.0070, 0.0125, 0.0078, 0.0089, 0.0076, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 10:16:22,051 INFO [train.py:904] (3/8) Epoch 11, batch 4800, loss[loss=0.1756, simple_loss=0.272, pruned_loss=0.03963, over 16663.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2841, pruned_loss=0.0576, over 3216682.09 frames. ], batch size: 89, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:16:45,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3791, 1.5001, 1.9637, 2.3654, 2.3668, 2.6503, 1.6455, 2.5930], device='cuda:3'), covar=tensor([0.0157, 0.0379, 0.0236, 0.0214, 0.0215, 0.0139, 0.0360, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0155, 0.0161, 0.0166, 0.0123, 0.0169, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 10:17:02,228 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:12,304 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:27,378 INFO [optim.py:368] (3/8) 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 10:17:35,182 INFO [train.py:904] (3/8) Epoch 11, batch 4850, loss[loss=0.1983, simple_loss=0.2942, pruned_loss=0.05119, over 15407.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2846, pruned_loss=0.05694, over 3202517.12 frames. ], batch size: 191, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:17:46,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9261, 3.7589, 3.9856, 4.1424, 4.2407, 3.8006, 4.1721, 4.2438], device='cuda:3'), covar=tensor([0.1172, 0.1029, 0.1168, 0.0530, 0.0445, 0.1412, 0.0566, 0.0550], device='cuda:3'), in_proj_covar=tensor([0.0503, 0.0635, 0.0775, 0.0651, 0.0494, 0.0500, 0.0504, 0.0578], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:17:49,041 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6639, 4.7313, 4.5113, 4.2551, 4.1059, 4.6023, 4.4342, 4.2368], device='cuda:3'), covar=tensor([0.0514, 0.0288, 0.0227, 0.0238, 0.0912, 0.0351, 0.0345, 0.0591], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0299, 0.0279, 0.0260, 0.0302, 0.0297, 0.0192, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:18:16,208 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 10:18:38,668 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 4900, loss[loss=0.1761, simple_loss=0.274, pruned_loss=0.03906, over 16792.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2835, pruned_loss=0.05533, over 3194035.81 frames. ], batch size: 96, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:18:49,665 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 10:19:16,832 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1798, 5.4551, 5.1333, 5.2178, 4.8829, 4.8242, 4.8025, 5.5346], device='cuda:3'), covar=tensor([0.1015, 0.0709, 0.1030, 0.0716, 0.0778, 0.0715, 0.1198, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0524, 0.0649, 0.0538, 0.0449, 0.0412, 0.0420, 0.0538, 0.0499], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:19:18,354 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 10:19:29,911 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:19:51,817 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 4950, loss[loss=0.1972, simple_loss=0.294, pruned_loss=0.05024, over 16717.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.283, pruned_loss=0.05442, over 3201431.22 frames. ], batch size: 134, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:20:01,963 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:02,974 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:04,220 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6942, 2.7275, 1.7004, 2.7971, 2.1465, 2.8262, 2.0068, 2.4001], device='cuda:3'), covar=tensor([0.0219, 0.0334, 0.1252, 0.0151, 0.0639, 0.0434, 0.1135, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0163, 0.0186, 0.0126, 0.0166, 0.0205, 0.0192, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 10:21:00,651 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:08,428 INFO [train.py:904] (3/8) Epoch 11, batch 5000, loss[loss=0.1937, simple_loss=0.2735, pruned_loss=0.05697, over 16993.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2846, pruned_loss=0.0549, over 3199975.57 frames. ], batch size: 53, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:21:10,911 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:28,646 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:54,788 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 10:22:14,238 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 5050, loss[loss=0.1927, simple_loss=0.2895, pruned_loss=0.04794, over 16857.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2855, pruned_loss=0.05484, over 3200693.75 frames. ], batch size: 96, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:22:58,359 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 10:23:32,362 INFO [train.py:904] (3/8) Epoch 11, batch 5100, loss[loss=0.1695, simple_loss=0.2609, pruned_loss=0.0391, over 16852.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2834, pruned_loss=0.05364, over 3219848.02 frames. ], batch size: 96, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:23:52,952 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7129, 3.9132, 2.1564, 4.4491, 2.7662, 4.3843, 2.3153, 2.9184], device='cuda:3'), covar=tensor([0.0193, 0.0247, 0.1505, 0.0084, 0.0780, 0.0285, 0.1407, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0162, 0.0185, 0.0125, 0.0166, 0.0203, 0.0192, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 10:23:57,027 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9698, 5.3973, 5.6908, 5.3879, 5.4863, 5.9985, 5.5693, 5.2740], device='cuda:3'), covar=tensor([0.0764, 0.1539, 0.1245, 0.1671, 0.2229, 0.0858, 0.1180, 0.2241], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0467, 0.0505, 0.0406, 0.0537, 0.0536, 0.0403, 0.0554], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 10:24:03,608 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:24:38,775 INFO [optim.py:368] (3/8) 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 10:24:44,358 INFO [train.py:904] (3/8) Epoch 11, batch 5150, loss[loss=0.1965, simple_loss=0.2743, pruned_loss=0.05931, over 16320.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2835, pruned_loss=0.05326, over 3219405.97 frames. ], batch size: 35, lr: 6.22e-03, grad_scale: 4.0 2023-04-29 10:25:13,989 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8743, 3.9562, 2.3091, 4.4961, 2.9082, 4.4434, 2.3569, 3.1045], device='cuda:3'), covar=tensor([0.0174, 0.0276, 0.1424, 0.0128, 0.0724, 0.0326, 0.1311, 0.0599], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0162, 0.0185, 0.0124, 0.0165, 0.0203, 0.0191, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 10:25:43,745 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 5200, loss[loss=0.1599, simple_loss=0.2437, pruned_loss=0.03805, over 16818.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2817, pruned_loss=0.05244, over 3215751.44 frames. ], batch size: 39, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:26:42,841 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:27:04,335 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 5250, loss[loss=0.1846, simple_loss=0.2784, pruned_loss=0.04534, over 16900.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2796, pruned_loss=0.05235, over 3211440.24 frames. ], batch size: 109, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:27:39,927 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 10:27:52,545 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:22,266 INFO [train.py:904] (3/8) Epoch 11, batch 5300, loss[loss=0.1581, simple_loss=0.2467, pruned_loss=0.03477, over 16854.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2766, pruned_loss=0.05097, over 3217138.81 frames. ], batch size: 96, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:28:29,633 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9500, 4.7703, 4.9655, 5.2004, 5.3733, 4.7398, 5.3651, 5.3154], device='cuda:3'), covar=tensor([0.1310, 0.0951, 0.1449, 0.0558, 0.0369, 0.0712, 0.0343, 0.0459], device='cuda:3'), in_proj_covar=tensor([0.0512, 0.0644, 0.0788, 0.0659, 0.0500, 0.0506, 0.0509, 0.0584], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:28:32,981 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:47,196 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7336, 4.9340, 5.1685, 4.9454, 4.9768, 5.5526, 5.0417, 4.6792], device='cuda:3'), covar=tensor([0.0933, 0.1569, 0.1308, 0.1523, 0.2131, 0.0759, 0.1177, 0.2401], device='cuda:3'), in_proj_covar=tensor([0.0345, 0.0472, 0.0513, 0.0414, 0.0548, 0.0542, 0.0410, 0.0565], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 10:29:27,200 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 5350, loss[loss=0.1943, simple_loss=0.2856, pruned_loss=0.0515, over 16899.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2746, pruned_loss=0.05009, over 3218171.55 frames. ], batch size: 83, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:30:45,866 INFO [train.py:904] (3/8) Epoch 11, batch 5400, loss[loss=0.2068, simple_loss=0.3008, pruned_loss=0.05637, over 16926.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2776, pruned_loss=0.05132, over 3195790.98 frames. ], batch size: 96, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:30:51,519 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2787, 4.0993, 4.2854, 4.4422, 4.5863, 4.1622, 4.5329, 4.5603], device='cuda:3'), covar=tensor([0.1351, 0.0987, 0.1377, 0.0618, 0.0466, 0.1010, 0.0504, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0518, 0.0650, 0.0793, 0.0666, 0.0507, 0.0513, 0.0512, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:31:18,215 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:31:54,542 INFO [optim.py:368] (3/8) 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 10:32:02,042 INFO [train.py:904] (3/8) Epoch 11, batch 5450, loss[loss=0.2165, simple_loss=0.2988, pruned_loss=0.06709, over 16343.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2806, pruned_loss=0.05296, over 3192231.94 frames. ], batch size: 35, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:32:11,895 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8175, 1.3256, 1.6020, 1.7118, 1.7713, 1.8767, 1.5092, 1.8332], device='cuda:3'), covar=tensor([0.0165, 0.0261, 0.0141, 0.0200, 0.0166, 0.0116, 0.0267, 0.0079], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0173, 0.0155, 0.0162, 0.0168, 0.0124, 0.0170, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 10:32:34,434 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:03,884 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:18,584 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4860, 3.4652, 3.4638, 2.8880, 3.3671, 2.0478, 3.2184, 2.9355], device='cuda:3'), covar=tensor([0.0124, 0.0118, 0.0142, 0.0233, 0.0090, 0.1924, 0.0109, 0.0195], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0116, 0.0164, 0.0159, 0.0135, 0.0177, 0.0150, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:33:19,270 INFO [train.py:904] (3/8) Epoch 11, batch 5500, loss[loss=0.2613, simple_loss=0.3332, pruned_loss=0.09466, over 11555.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2891, pruned_loss=0.0587, over 3149144.83 frames. ], batch size: 246, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:34:18,908 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:34:37,906 INFO [train.py:904] (3/8) Epoch 11, batch 5550, loss[loss=0.2211, simple_loss=0.3023, pruned_loss=0.06994, over 16589.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2969, pruned_loss=0.06458, over 3136754.02 frames. ], batch size: 57, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:35:57,900 INFO [train.py:904] (3/8) Epoch 11, batch 5600, loss[loss=0.2986, simple_loss=0.3601, pruned_loss=0.1185, over 11203.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3031, pruned_loss=0.06975, over 3105954.64 frames. ], batch size: 246, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:36:12,115 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:02,617 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:37:21,714 INFO [train.py:904] (3/8) Epoch 11, batch 5650, loss[loss=0.2274, simple_loss=0.2971, pruned_loss=0.07881, over 17022.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3091, pruned_loss=0.07522, over 3064241.17 frames. ], batch size: 53, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:37:22,251 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:32,232 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:38:29,837 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5866, 3.6210, 2.1463, 4.0887, 2.5944, 4.0675, 2.1279, 2.8129], device='cuda:3'), covar=tensor([0.0209, 0.0333, 0.1548, 0.0134, 0.0895, 0.0483, 0.1554, 0.0715], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0163, 0.0186, 0.0125, 0.0166, 0.0203, 0.0192, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 10:38:42,735 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:38:43,536 INFO [train.py:904] (3/8) Epoch 11, batch 5700, loss[loss=0.2996, simple_loss=0.3503, pruned_loss=0.1245, over 11278.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3108, pruned_loss=0.07669, over 3062691.63 frames. ], batch size: 247, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:38:47,181 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3326, 2.8932, 2.5861, 2.2686, 2.3070, 2.2158, 2.8451, 2.9066], device='cuda:3'), covar=tensor([0.2066, 0.0693, 0.1327, 0.1812, 0.1961, 0.1705, 0.0422, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0255, 0.0281, 0.0275, 0.0280, 0.0219, 0.0265, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:39:02,866 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:39:34,266 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 10:39:59,317 INFO [optim.py:368] (3/8) 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 10:40:04,426 INFO [train.py:904] (3/8) Epoch 11, batch 5750, loss[loss=0.276, simple_loss=0.3302, pruned_loss=0.1109, over 11197.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.3138, pruned_loss=0.07864, over 3041903.22 frames. ], batch size: 248, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:40:35,481 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6776, 2.6231, 1.8935, 2.7163, 2.2183, 2.7883, 2.0923, 2.4596], device='cuda:3'), covar=tensor([0.0271, 0.0406, 0.1186, 0.0224, 0.0663, 0.0537, 0.1223, 0.0542], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0164, 0.0187, 0.0126, 0.0168, 0.0205, 0.0195, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 10:41:25,722 INFO [train.py:904] (3/8) Epoch 11, batch 5800, loss[loss=0.2579, simple_loss=0.3386, pruned_loss=0.08858, over 17017.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3125, pruned_loss=0.07635, over 3057132.07 frames. ], batch size: 53, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:41:59,321 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 10:42:39,026 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 2.956e+02 3.560e+02 4.660e+02 9.709e+02, threshold=7.120e+02, percent-clipped=1.0 2023-04-29 10:42:43,689 INFO [train.py:904] (3/8) Epoch 11, batch 5850, loss[loss=0.2112, simple_loss=0.2954, pruned_loss=0.06354, over 16365.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3101, pruned_loss=0.07446, over 3053142.68 frames. ], batch size: 146, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:42:52,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9121, 5.2072, 4.9613, 4.9584, 4.6988, 4.6380, 4.6738, 5.3274], device='cuda:3'), covar=tensor([0.0980, 0.0769, 0.0924, 0.0723, 0.0711, 0.0837, 0.0962, 0.0738], device='cuda:3'), in_proj_covar=tensor([0.0539, 0.0664, 0.0552, 0.0457, 0.0421, 0.0432, 0.0553, 0.0507], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:43:19,698 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-29 10:44:05,219 INFO [train.py:904] (3/8) Epoch 11, batch 5900, loss[loss=0.2553, simple_loss=0.3174, pruned_loss=0.09662, over 11736.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3096, pruned_loss=0.07431, over 3065473.01 frames. ], batch size: 246, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:45:21,995 INFO [optim.py:368] (3/8) 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 10:45:26,030 INFO [train.py:904] (3/8) Epoch 11, batch 5950, loss[loss=0.2398, simple_loss=0.3324, pruned_loss=0.07359, over 17199.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.3098, pruned_loss=0.07259, over 3064385.26 frames. ], batch size: 44, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:45:39,029 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9815, 4.0465, 4.5019, 2.0969, 4.7203, 4.7277, 3.1979, 3.5604], device='cuda:3'), covar=tensor([0.0616, 0.0158, 0.0108, 0.1051, 0.0035, 0.0065, 0.0297, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0098, 0.0087, 0.0137, 0.0069, 0.0101, 0.0118, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 10:46:40,590 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:46:48,947 INFO [train.py:904] (3/8) Epoch 11, batch 6000, loss[loss=0.1959, simple_loss=0.2825, pruned_loss=0.05467, over 16895.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.309, pruned_loss=0.07196, over 3062439.64 frames. ], batch size: 109, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:46:48,948 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 10:46:59,888 INFO [train.py:938] (3/8) Epoch 11, validation: loss=0.163, simple_loss=0.2761, pruned_loss=0.02492, over 944034.00 frames. 2023-04-29 10:46:59,889 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 10:47:06,066 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8455, 4.1338, 3.3127, 2.3809, 2.9422, 2.4868, 4.5665, 3.8676], device='cuda:3'), covar=tensor([0.2496, 0.0631, 0.1281, 0.1855, 0.2230, 0.1662, 0.0306, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0256, 0.0281, 0.0276, 0.0281, 0.0219, 0.0266, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:47:10,234 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:47:13,553 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:47:13,734 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7732, 2.3075, 2.3525, 4.5961, 2.2646, 2.8137, 2.4488, 2.6161], device='cuda:3'), covar=tensor([0.0823, 0.3009, 0.2113, 0.0344, 0.3418, 0.1989, 0.2610, 0.2779], device='cuda:3'), in_proj_covar=tensor([0.0358, 0.0385, 0.0323, 0.0315, 0.0406, 0.0443, 0.0350, 0.0452], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:48:12,720 INFO [optim.py:368] (3/8) 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 10:48:18,426 INFO [train.py:904] (3/8) Epoch 11, batch 6050, loss[loss=0.2203, simple_loss=0.324, pruned_loss=0.05823, over 16942.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3076, pruned_loss=0.07152, over 3070111.35 frames. ], batch size: 96, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:48:44,797 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6122, 4.4233, 4.6582, 4.8582, 5.0081, 4.4442, 4.9525, 4.9593], device='cuda:3'), covar=tensor([0.1614, 0.1154, 0.1508, 0.0642, 0.0549, 0.0892, 0.0676, 0.0596], device='cuda:3'), in_proj_covar=tensor([0.0516, 0.0644, 0.0780, 0.0655, 0.0503, 0.0508, 0.0518, 0.0585], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:48:48,366 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:49:28,896 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 10:49:34,932 INFO [train.py:904] (3/8) Epoch 11, batch 6100, loss[loss=0.2258, simple_loss=0.3151, pruned_loss=0.06827, over 16503.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3068, pruned_loss=0.07001, over 3083665.73 frames. ], batch size: 68, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:50:51,420 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 6150, loss[loss=0.2207, simple_loss=0.306, pruned_loss=0.06773, over 15379.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3049, pruned_loss=0.06946, over 3074367.29 frames. ], batch size: 190, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:51:07,618 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:51:11,039 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0011, 5.0884, 5.5106, 5.4877, 5.4680, 5.1255, 5.0445, 4.6992], device='cuda:3'), covar=tensor([0.0259, 0.0426, 0.0286, 0.0320, 0.0458, 0.0271, 0.0905, 0.0396], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0341, 0.0343, 0.0325, 0.0388, 0.0361, 0.0463, 0.0294], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 10:51:36,457 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9544, 3.3893, 3.3549, 2.0191, 2.8822, 2.3444, 3.4077, 3.5194], device='cuda:3'), covar=tensor([0.0258, 0.0612, 0.0540, 0.1717, 0.0736, 0.0858, 0.0579, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0142, 0.0156, 0.0143, 0.0135, 0.0124, 0.0136, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 10:52:14,168 INFO [train.py:904] (3/8) Epoch 11, batch 6200, loss[loss=0.2198, simple_loss=0.3027, pruned_loss=0.06848, over 15298.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3031, pruned_loss=0.06875, over 3093920.14 frames. ], batch size: 190, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:52:17,831 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1661, 3.4997, 3.5302, 2.0293, 2.9901, 2.4364, 3.5096, 3.6571], device='cuda:3'), covar=tensor([0.0244, 0.0646, 0.0517, 0.1811, 0.0756, 0.0837, 0.0645, 0.0809], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0142, 0.0156, 0.0143, 0.0135, 0.0124, 0.0136, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 10:52:20,383 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-29 10:52:39,386 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8596, 4.1632, 3.9378, 3.9920, 3.6294, 3.7849, 3.8483, 4.1328], device='cuda:3'), covar=tensor([0.1106, 0.0827, 0.1009, 0.0740, 0.0816, 0.1380, 0.0920, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0547, 0.0674, 0.0564, 0.0466, 0.0426, 0.0438, 0.0562, 0.0514], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:52:42,537 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:52:57,591 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4512, 4.3715, 4.3308, 2.8677, 3.7536, 4.3339, 3.9563, 2.3180], device='cuda:3'), covar=tensor([0.0425, 0.0022, 0.0028, 0.0291, 0.0069, 0.0065, 0.0047, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0068, 0.0069, 0.0125, 0.0077, 0.0090, 0.0076, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 10:53:07,924 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:27,429 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.232e+02 3.321e+02 3.959e+02 5.175e+02 1.024e+03, threshold=7.919e+02, percent-clipped=5.0 2023-04-29 10:53:29,995 INFO [train.py:904] (3/8) Epoch 11, batch 6250, loss[loss=0.2632, simple_loss=0.3341, pruned_loss=0.09614, over 11416.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3022, pruned_loss=0.06813, over 3106325.23 frames. ], batch size: 247, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:53:55,923 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6087, 2.8239, 2.5939, 4.3504, 3.1524, 4.1569, 1.5011, 2.9129], device='cuda:3'), covar=tensor([0.1392, 0.0651, 0.1078, 0.0115, 0.0272, 0.0378, 0.1493, 0.0796], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0157, 0.0180, 0.0143, 0.0197, 0.0207, 0.0179, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 10:54:36,284 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:54:37,537 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9808, 4.0718, 3.8545, 3.6401, 3.5666, 3.9588, 3.6677, 3.7110], device='cuda:3'), covar=tensor([0.0615, 0.0468, 0.0271, 0.0257, 0.0786, 0.0398, 0.0928, 0.0578], device='cuda:3'), in_proj_covar=tensor([0.0237, 0.0303, 0.0279, 0.0258, 0.0301, 0.0298, 0.0193, 0.0326], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:54:39,030 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:54:42,639 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5902, 2.6124, 2.1412, 2.4553, 3.0081, 2.7493, 3.2512, 3.2929], device='cuda:3'), covar=tensor([0.0063, 0.0278, 0.0367, 0.0322, 0.0180, 0.0262, 0.0172, 0.0162], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0200, 0.0197, 0.0197, 0.0200, 0.0201, 0.0205, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:54:45,102 INFO [train.py:904] (3/8) Epoch 11, batch 6300, loss[loss=0.1975, simple_loss=0.286, pruned_loss=0.05453, over 16831.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3022, pruned_loss=0.06782, over 3102367.35 frames. ], batch size: 96, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:54,182 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:55:52,343 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:56:00,515 INFO [optim.py:368] (3/8) 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 10:56:03,040 INFO [train.py:904] (3/8) Epoch 11, batch 6350, loss[loss=0.2314, simple_loss=0.3093, pruned_loss=0.07674, over 16558.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3033, pruned_loss=0.06919, over 3095694.31 frames. ], batch size: 57, lr: 6.18e-03, grad_scale: 4.0 2023-04-29 10:56:10,568 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:25,827 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:35,701 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0233, 2.7273, 2.7099, 2.0494, 2.5719, 2.1760, 2.8302, 2.9745], device='cuda:3'), covar=tensor([0.0265, 0.0668, 0.0490, 0.1590, 0.0726, 0.0858, 0.0515, 0.0555], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0143, 0.0157, 0.0144, 0.0136, 0.0125, 0.0137, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 10:57:20,953 INFO [train.py:904] (3/8) Epoch 11, batch 6400, loss[loss=0.2889, simple_loss=0.3502, pruned_loss=0.1138, over 11066.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3039, pruned_loss=0.07058, over 3090314.41 frames. ], batch size: 247, lr: 6.18e-03, grad_scale: 8.0 2023-04-29 10:57:26,297 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 10:57:32,739 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8646, 3.9089, 4.2693, 4.2311, 4.2465, 3.9605, 4.0187, 3.9177], device='cuda:3'), covar=tensor([0.0322, 0.0575, 0.0373, 0.0446, 0.0481, 0.0399, 0.0791, 0.0478], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0341, 0.0344, 0.0322, 0.0387, 0.0359, 0.0460, 0.0292], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 10:58:23,896 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-04-29 10:58:35,865 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 6450, loss[loss=0.2494, simple_loss=0.3171, pruned_loss=0.09086, over 11535.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3033, pruned_loss=0.06956, over 3091037.82 frames. ], batch size: 247, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 10:58:52,476 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5729, 2.1161, 1.7062, 1.9721, 2.4866, 2.1780, 2.4823, 2.7009], device='cuda:3'), covar=tensor([0.0130, 0.0299, 0.0426, 0.0360, 0.0187, 0.0290, 0.0188, 0.0199], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0198, 0.0196, 0.0196, 0.0198, 0.0200, 0.0203, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 10:59:54,944 INFO [train.py:904] (3/8) Epoch 11, batch 6500, loss[loss=0.1793, simple_loss=0.2578, pruned_loss=0.05043, over 16983.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.301, pruned_loss=0.06886, over 3085801.64 frames. ], batch size: 41, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:00:14,833 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:00:48,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0477, 3.2604, 2.9416, 5.1537, 3.9881, 4.5483, 1.5468, 3.2696], device='cuda:3'), covar=tensor([0.1221, 0.0626, 0.1088, 0.0127, 0.0419, 0.0337, 0.1548, 0.0775], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0158, 0.0182, 0.0143, 0.0199, 0.0208, 0.0180, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 11:01:12,919 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 6550, loss[loss=0.24, simple_loss=0.3324, pruned_loss=0.07385, over 15325.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.304, pruned_loss=0.07027, over 3075200.02 frames. ], batch size: 190, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:01:26,266 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7051, 1.6712, 2.2225, 2.6693, 2.5857, 2.9378, 1.7809, 2.9660], device='cuda:3'), covar=tensor([0.0145, 0.0397, 0.0208, 0.0208, 0.0220, 0.0139, 0.0379, 0.0090], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0171, 0.0150, 0.0158, 0.0168, 0.0124, 0.0168, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 11:02:13,591 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:02:25,852 INFO [train.py:904] (3/8) Epoch 11, batch 6600, loss[loss=0.2062, simple_loss=0.2874, pruned_loss=0.06247, over 17002.00 frames. ], tot_loss[loss=0.224, simple_loss=0.306, pruned_loss=0.07098, over 3054092.49 frames. ], batch size: 55, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:02:41,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6066, 4.5814, 5.0523, 4.9675, 4.9902, 4.6469, 4.6807, 4.4312], device='cuda:3'), covar=tensor([0.0263, 0.0430, 0.0354, 0.0420, 0.0409, 0.0348, 0.0831, 0.0448], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0341, 0.0343, 0.0322, 0.0389, 0.0360, 0.0461, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 11:03:41,607 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 6650, loss[loss=0.1859, simple_loss=0.2631, pruned_loss=0.05434, over 16848.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3058, pruned_loss=0.07105, over 3068498.18 frames. ], batch size: 42, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:03:51,288 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 11:04:03,292 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:04:56,964 INFO [train.py:904] (3/8) Epoch 11, batch 6700, loss[loss=0.2279, simple_loss=0.3228, pruned_loss=0.06652, over 16744.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3038, pruned_loss=0.07043, over 3088232.79 frames. ], batch size: 89, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:05:14,876 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:05:28,605 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2647, 4.2434, 4.7028, 4.6762, 4.6630, 4.3417, 4.3616, 4.1989], device='cuda:3'), covar=tensor([0.0309, 0.0565, 0.0348, 0.0392, 0.0480, 0.0322, 0.0846, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0341, 0.0346, 0.0323, 0.0392, 0.0360, 0.0462, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 11:06:13,521 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 6750, loss[loss=0.2351, simple_loss=0.3159, pruned_loss=0.07718, over 16887.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3032, pruned_loss=0.07065, over 3104432.43 frames. ], batch size: 116, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:07:28,508 INFO [train.py:904] (3/8) Epoch 11, batch 6800, loss[loss=0.2281, simple_loss=0.3192, pruned_loss=0.06853, over 16917.00 frames. ], tot_loss[loss=0.223, simple_loss=0.304, pruned_loss=0.07098, over 3101774.86 frames. ], batch size: 109, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:07:48,638 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:05,616 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:21,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4457, 5.7184, 5.3852, 5.4282, 5.1313, 4.9227, 5.2020, 5.7905], device='cuda:3'), covar=tensor([0.1003, 0.0669, 0.0910, 0.0698, 0.0803, 0.0742, 0.0922, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0541, 0.0660, 0.0551, 0.0457, 0.0419, 0.0432, 0.0554, 0.0509], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:08:21,469 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7881, 3.9327, 4.2283, 1.7249, 4.4919, 4.5007, 3.0597, 3.1920], device='cuda:3'), covar=tensor([0.0688, 0.0166, 0.0155, 0.1268, 0.0044, 0.0080, 0.0378, 0.0405], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0099, 0.0088, 0.0138, 0.0069, 0.0102, 0.0120, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 11:08:45,531 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 11, batch 6850, loss[loss=0.2202, simple_loss=0.3225, pruned_loss=0.05898, over 16898.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3059, pruned_loss=0.0718, over 3098413.47 frames. ], batch size: 90, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:09:01,710 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:07,967 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9428, 4.0000, 4.3635, 1.7255, 4.6159, 4.6809, 3.3761, 3.2452], device='cuda:3'), covar=tensor([0.0679, 0.0180, 0.0119, 0.1341, 0.0042, 0.0075, 0.0286, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0099, 0.0088, 0.0139, 0.0069, 0.0102, 0.0120, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 11:09:35,687 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:46,459 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 6900, loss[loss=0.2397, simple_loss=0.3162, pruned_loss=0.08162, over 16954.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3085, pruned_loss=0.07199, over 3102090.75 frames. ], batch size: 109, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:10:12,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5094, 2.7384, 2.3689, 3.8112, 2.7007, 3.8381, 1.5072, 2.6999], device='cuda:3'), covar=tensor([0.1465, 0.0666, 0.1199, 0.0144, 0.0282, 0.0463, 0.1554, 0.0873], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0159, 0.0182, 0.0144, 0.0201, 0.0209, 0.0182, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 11:10:45,877 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:10:59,097 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:11:15,789 INFO [train.py:904] (3/8) Epoch 11, batch 6950, loss[loss=0.2044, simple_loss=0.2918, pruned_loss=0.0585, over 16465.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.3093, pruned_loss=0.07312, over 3095962.14 frames. ], batch size: 146, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:11:17,884 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.933e+02 3.744e+02 4.621e+02 9.342e+02, threshold=7.489e+02, percent-clipped=9.0 2023-04-29 11:11:31,138 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3929, 4.3855, 4.2433, 3.6034, 4.2709, 1.6839, 4.0403, 4.0484], device='cuda:3'), covar=tensor([0.0078, 0.0066, 0.0141, 0.0312, 0.0079, 0.2273, 0.0111, 0.0162], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0113, 0.0160, 0.0151, 0.0131, 0.0175, 0.0146, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:12:20,831 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:12:33,547 INFO [train.py:904] (3/8) Epoch 11, batch 7000, loss[loss=0.217, simple_loss=0.3077, pruned_loss=0.06315, over 16778.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3093, pruned_loss=0.07236, over 3097904.60 frames. ], batch size: 124, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:12:56,360 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-29 11:13:19,959 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 11:13:52,262 INFO [train.py:904] (3/8) Epoch 11, batch 7050, loss[loss=0.2489, simple_loss=0.3257, pruned_loss=0.08606, over 15342.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3098, pruned_loss=0.07207, over 3090062.43 frames. ], batch size: 190, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:13:53,481 INFO [optim.py:368] (3/8) 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 11:13:59,950 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 11:14:20,354 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6588, 4.4578, 4.6972, 4.8892, 5.0584, 4.4933, 5.0122, 4.9947], device='cuda:3'), covar=tensor([0.1486, 0.1253, 0.1534, 0.0653, 0.0501, 0.0928, 0.0468, 0.0543], device='cuda:3'), in_proj_covar=tensor([0.0509, 0.0632, 0.0766, 0.0644, 0.0495, 0.0497, 0.0511, 0.0577], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:14:32,904 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:14:39,215 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4784, 3.4639, 3.4210, 2.8766, 3.3149, 2.0812, 3.1346, 2.8637], device='cuda:3'), covar=tensor([0.0122, 0.0110, 0.0149, 0.0218, 0.0092, 0.1870, 0.0116, 0.0164], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0113, 0.0161, 0.0152, 0.0132, 0.0177, 0.0148, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:15:11,200 INFO [train.py:904] (3/8) Epoch 11, batch 7100, loss[loss=0.2267, simple_loss=0.3139, pruned_loss=0.06975, over 16481.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3081, pruned_loss=0.07196, over 3081194.04 frames. ], batch size: 146, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:04,528 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1638, 1.4363, 1.8760, 2.0723, 2.2382, 2.2961, 1.6361, 2.3310], device='cuda:3'), covar=tensor([0.0168, 0.0329, 0.0198, 0.0213, 0.0175, 0.0129, 0.0319, 0.0081], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0167, 0.0148, 0.0156, 0.0164, 0.0121, 0.0167, 0.0111], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 11:16:07,297 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:16:12,990 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4903, 3.5539, 3.2941, 3.1344, 3.1616, 3.4155, 3.2810, 3.2203], device='cuda:3'), covar=tensor([0.0531, 0.0459, 0.0237, 0.0215, 0.0526, 0.0384, 0.1201, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0300, 0.0275, 0.0252, 0.0296, 0.0293, 0.0193, 0.0320], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:16:17,917 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9905, 3.3948, 3.4955, 1.5079, 3.6743, 3.8132, 2.8149, 2.4726], device='cuda:3'), covar=tensor([0.1059, 0.0171, 0.0182, 0.1340, 0.0073, 0.0117, 0.0404, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0099, 0.0088, 0.0138, 0.0069, 0.0102, 0.0120, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 11:16:21,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4348, 3.8251, 3.8926, 2.2764, 3.1568, 2.4929, 3.8165, 4.0383], device='cuda:3'), covar=tensor([0.0235, 0.0619, 0.0485, 0.1641, 0.0723, 0.0834, 0.0588, 0.0809], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0144, 0.0158, 0.0144, 0.0137, 0.0126, 0.0137, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 11:16:27,131 INFO [train.py:904] (3/8) Epoch 11, batch 7150, loss[loss=0.2723, simple_loss=0.3243, pruned_loss=0.1102, over 11566.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3061, pruned_loss=0.07176, over 3086013.67 frames. ], batch size: 248, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:28,930 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.054e+02 3.516e+02 4.486e+02 8.068e+02, threshold=7.031e+02, percent-clipped=2.0 2023-04-29 11:17:10,531 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 7200, loss[loss=0.1952, simple_loss=0.2754, pruned_loss=0.05748, over 11430.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.304, pruned_loss=0.06973, over 3089584.37 frames. ], batch size: 246, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:18:06,084 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4434, 3.2073, 2.6656, 2.0811, 2.2811, 2.1067, 3.3433, 3.0133], device='cuda:3'), covar=tensor([0.2494, 0.0776, 0.1564, 0.2266, 0.2238, 0.1899, 0.0498, 0.1099], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0257, 0.0283, 0.0276, 0.0281, 0.0221, 0.0266, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:18:48,127 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 11:19:02,056 INFO [train.py:904] (3/8) Epoch 11, batch 7250, loss[loss=0.2576, simple_loss=0.3168, pruned_loss=0.09922, over 11687.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3017, pruned_loss=0.06865, over 3081909.85 frames. ], batch size: 248, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:19:03,145 INFO [optim.py:368] (3/8) 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 11:19:12,227 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2982, 5.6513, 5.3839, 5.3866, 5.0588, 4.9234, 5.1630, 5.7468], device='cuda:3'), covar=tensor([0.0990, 0.0767, 0.0859, 0.0632, 0.0727, 0.0742, 0.0870, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0534, 0.0657, 0.0544, 0.0451, 0.0416, 0.0430, 0.0549, 0.0502], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:19:26,184 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 11:19:55,676 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:20:01,024 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:20:09,489 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9864, 4.0121, 3.8190, 3.6784, 3.5526, 3.9416, 3.6033, 3.7109], device='cuda:3'), covar=tensor([0.0535, 0.0400, 0.0266, 0.0236, 0.0745, 0.0400, 0.1031, 0.0499], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0300, 0.0275, 0.0254, 0.0297, 0.0294, 0.0193, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:20:16,090 INFO [train.py:904] (3/8) Epoch 11, batch 7300, loss[loss=0.2356, simple_loss=0.2977, pruned_loss=0.08677, over 11224.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3013, pruned_loss=0.06858, over 3080822.23 frames. ], batch size: 246, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:20:45,530 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4900, 3.5434, 3.1636, 3.0174, 3.1050, 3.3963, 3.2700, 3.1530], device='cuda:3'), covar=tensor([0.0504, 0.0361, 0.0241, 0.0212, 0.0489, 0.0329, 0.1152, 0.0386], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0299, 0.0274, 0.0253, 0.0297, 0.0294, 0.0192, 0.0321], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:20:46,869 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8190, 2.1500, 1.6708, 1.9443, 2.4738, 2.0215, 2.6963, 2.7351], device='cuda:3'), covar=tensor([0.0085, 0.0289, 0.0428, 0.0351, 0.0192, 0.0332, 0.0136, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0198, 0.0194, 0.0195, 0.0198, 0.0199, 0.0202, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:21:34,169 INFO [train.py:904] (3/8) Epoch 11, batch 7350, loss[loss=0.2172, simple_loss=0.2969, pruned_loss=0.06878, over 15385.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3017, pruned_loss=0.06859, over 3089911.05 frames. ], batch size: 192, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,663 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:21:35,285 INFO [optim.py:368] (3/8) 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 11:21:39,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1060, 4.0962, 4.5591, 4.5202, 4.5258, 4.1801, 4.2180, 4.0493], device='cuda:3'), covar=tensor([0.0301, 0.0517, 0.0337, 0.0408, 0.0489, 0.0352, 0.0867, 0.0499], device='cuda:3'), in_proj_covar=tensor([0.0316, 0.0329, 0.0334, 0.0314, 0.0380, 0.0349, 0.0450, 0.0284], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 11:22:54,331 INFO [train.py:904] (3/8) Epoch 11, batch 7400, loss[loss=0.2121, simple_loss=0.3052, pruned_loss=0.05955, over 16664.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3033, pruned_loss=0.06967, over 3074972.09 frames. ], batch size: 134, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:23:05,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4875, 4.2466, 4.3171, 2.6550, 3.7913, 4.2115, 3.8492, 2.4183], device='cuda:3'), covar=tensor([0.0431, 0.0027, 0.0026, 0.0340, 0.0061, 0.0078, 0.0062, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0068, 0.0069, 0.0127, 0.0078, 0.0090, 0.0077, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 11:23:25,663 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-29 11:23:42,500 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 7450, loss[loss=0.2086, simple_loss=0.306, pruned_loss=0.05563, over 16246.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3039, pruned_loss=0.0702, over 3095536.52 frames. ], batch size: 165, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:24:13,636 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 3.048e+02 3.705e+02 5.111e+02 8.676e+02, threshold=7.410e+02, percent-clipped=2.0 2023-04-29 11:24:59,156 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:25:24,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5228, 3.4825, 1.7972, 3.9643, 2.5284, 3.9293, 2.0653, 2.6884], device='cuda:3'), covar=tensor([0.0194, 0.0363, 0.1734, 0.0147, 0.0791, 0.0434, 0.1540, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0163, 0.0187, 0.0125, 0.0167, 0.0203, 0.0196, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 11:25:31,627 INFO [train.py:904] (3/8) Epoch 11, batch 7500, loss[loss=0.2128, simple_loss=0.2947, pruned_loss=0.06544, over 16919.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3039, pruned_loss=0.0693, over 3085488.43 frames. ], batch size: 116, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:00,596 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-04-29 11:26:16,354 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:26:27,929 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 7550, loss[loss=0.2121, simple_loss=0.2898, pruned_loss=0.0672, over 15380.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3029, pruned_loss=0.06968, over 3075779.58 frames. ], batch size: 191, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:52,322 INFO [optim.py:368] (3/8) 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 11:27:46,043 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:28:02,005 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:28:06,702 INFO [train.py:904] (3/8) Epoch 11, batch 7600, loss[loss=0.2276, simple_loss=0.3061, pruned_loss=0.07452, over 15266.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3021, pruned_loss=0.06999, over 3061194.59 frames. ], batch size: 190, lr: 6.15e-03, grad_scale: 8.0 2023-04-29 11:28:45,338 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8758, 4.0637, 3.0716, 2.2558, 3.0032, 2.4783, 4.3818, 3.7522], device='cuda:3'), covar=tensor([0.2576, 0.0666, 0.1587, 0.2279, 0.2208, 0.1739, 0.0447, 0.0990], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0257, 0.0284, 0.0277, 0.0282, 0.0223, 0.0267, 0.0294], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:28:56,882 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:12,236 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:20,692 INFO [train.py:904] (3/8) Epoch 11, batch 7650, loss[loss=0.2263, simple_loss=0.3084, pruned_loss=0.07214, over 17189.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3031, pruned_loss=0.07078, over 3074037.44 frames. ], batch size: 46, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:29:23,629 INFO [optim.py:368] (3/8) 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 11:29:46,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4814, 4.2939, 4.4985, 4.6779, 4.8233, 4.3476, 4.7636, 4.7945], device='cuda:3'), covar=tensor([0.1491, 0.1116, 0.1390, 0.0659, 0.0475, 0.0971, 0.0604, 0.0562], device='cuda:3'), in_proj_covar=tensor([0.0516, 0.0638, 0.0771, 0.0650, 0.0502, 0.0501, 0.0512, 0.0581], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:30:25,796 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:30:36,527 INFO [train.py:904] (3/8) Epoch 11, batch 7700, loss[loss=0.2289, simple_loss=0.3102, pruned_loss=0.0738, over 16606.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3033, pruned_loss=0.07166, over 3065767.28 frames. ], batch size: 57, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:31:25,282 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:31:53,770 INFO [train.py:904] (3/8) Epoch 11, batch 7750, loss[loss=0.2358, simple_loss=0.3118, pruned_loss=0.07991, over 16450.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3035, pruned_loss=0.07156, over 3072164.24 frames. ], batch size: 68, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:31:56,716 INFO [optim.py:368] (3/8) 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 11:31:59,738 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:09,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 11:32:19,734 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2135, 4.2344, 4.7029, 4.6543, 4.6278, 4.3165, 4.3168, 4.1349], device='cuda:3'), covar=tensor([0.0297, 0.0580, 0.0330, 0.0395, 0.0493, 0.0372, 0.0946, 0.0508], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0335, 0.0340, 0.0317, 0.0386, 0.0354, 0.0460, 0.0289], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 11:32:39,671 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:58,021 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:33:10,715 INFO [train.py:904] (3/8) Epoch 11, batch 7800, loss[loss=0.2153, simple_loss=0.3061, pruned_loss=0.06228, over 16937.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3039, pruned_loss=0.07133, over 3076410.30 frames. ], batch size: 96, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:33:35,206 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:34:25,572 INFO [train.py:904] (3/8) Epoch 11, batch 7850, loss[loss=0.2406, simple_loss=0.3127, pruned_loss=0.08423, over 11792.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3052, pruned_loss=0.07182, over 3053348.24 frames. ], batch size: 246, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:34:30,495 INFO [optim.py:368] (3/8) 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 11:34:30,964 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:35:07,095 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:35:08,854 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 11:35:15,524 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 11:35:27,279 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:35:41,207 INFO [train.py:904] (3/8) Epoch 11, batch 7900, loss[loss=0.2372, simple_loss=0.3143, pruned_loss=0.08002, over 16255.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3043, pruned_loss=0.07112, over 3069222.89 frames. ], batch size: 35, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:36:02,565 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1666, 3.3899, 3.6194, 3.5780, 3.5686, 3.3644, 3.3984, 3.4731], device='cuda:3'), covar=tensor([0.0369, 0.0592, 0.0380, 0.0414, 0.0475, 0.0460, 0.0868, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0336, 0.0339, 0.0318, 0.0387, 0.0355, 0.0461, 0.0288], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 11:36:05,649 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 11:36:52,245 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 7950, loss[loss=0.23, simple_loss=0.3108, pruned_loss=0.07459, over 16662.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3057, pruned_loss=0.07219, over 3069679.87 frames. ], batch size: 134, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:37:04,708 INFO [optim.py:368] (3/8) 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 11:37:24,431 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3265, 3.2764, 3.3661, 3.4497, 3.5089, 3.2660, 3.4557, 3.5437], device='cuda:3'), covar=tensor([0.1006, 0.0816, 0.0926, 0.0553, 0.0623, 0.1927, 0.0866, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0516, 0.0642, 0.0770, 0.0651, 0.0504, 0.0499, 0.0515, 0.0584], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:37:52,566 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-29 11:38:04,154 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:38:14,154 INFO [train.py:904] (3/8) Epoch 11, batch 8000, loss[loss=0.2062, simple_loss=0.293, pruned_loss=0.05963, over 15240.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3064, pruned_loss=0.07262, over 3071276.84 frames. ], batch size: 190, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:38:49,422 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0082, 3.8409, 4.1021, 4.2125, 4.3419, 3.9386, 4.2832, 4.3308], device='cuda:3'), covar=tensor([0.1441, 0.1151, 0.1284, 0.0670, 0.0540, 0.1289, 0.0668, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0513, 0.0637, 0.0768, 0.0648, 0.0502, 0.0495, 0.0511, 0.0581], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:39:24,939 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:39:27,474 INFO [train.py:904] (3/8) Epoch 11, batch 8050, loss[loss=0.2635, simple_loss=0.3419, pruned_loss=0.09251, over 15320.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3055, pruned_loss=0.07144, over 3093184.15 frames. ], batch size: 190, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:31,003 INFO [optim.py:368] (3/8) 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 11:39:37,727 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9586, 1.9774, 2.1998, 3.4832, 1.9067, 2.3116, 2.0992, 2.1238], device='cuda:3'), covar=tensor([0.1146, 0.3406, 0.2220, 0.0525, 0.3979, 0.2350, 0.3212, 0.3079], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0392, 0.0327, 0.0319, 0.0415, 0.0447, 0.0354, 0.0459], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:40:06,532 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:40:40,895 INFO [train.py:904] (3/8) Epoch 11, batch 8100, loss[loss=0.2483, simple_loss=0.3142, pruned_loss=0.09114, over 11538.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3053, pruned_loss=0.07146, over 3075005.21 frames. ], batch size: 247, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:41:01,490 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.44 vs. limit=5.0 2023-04-29 11:41:39,243 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 11, batch 8150, loss[loss=0.1936, simple_loss=0.2752, pruned_loss=0.05599, over 16862.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3024, pruned_loss=0.06996, over 3088461.82 frames. ], batch size: 116, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:42:01,362 INFO [optim.py:368] (3/8) 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 11:42:08,199 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:42:59,323 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:43:12,266 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-29 11:43:12,476 INFO [train.py:904] (3/8) Epoch 11, batch 8200, loss[loss=0.2418, simple_loss=0.325, pruned_loss=0.07929, over 15204.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2994, pruned_loss=0.0685, over 3108679.56 frames. ], batch size: 190, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:43:41,772 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:44:04,816 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2958, 4.2251, 4.6672, 4.6368, 4.6074, 4.3610, 4.3141, 4.1722], device='cuda:3'), covar=tensor([0.0285, 0.0481, 0.0360, 0.0362, 0.0460, 0.0337, 0.0984, 0.0480], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0338, 0.0342, 0.0319, 0.0391, 0.0356, 0.0464, 0.0291], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 11:44:16,310 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:44:33,540 INFO [train.py:904] (3/8) Epoch 11, batch 8250, loss[loss=0.2084, simple_loss=0.302, pruned_loss=0.05735, over 16207.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2991, pruned_loss=0.06684, over 3092676.73 frames. ], batch size: 165, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:44:38,004 INFO [optim.py:368] (3/8) 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 11:45:52,519 INFO [train.py:904] (3/8) Epoch 11, batch 8300, loss[loss=0.1958, simple_loss=0.2863, pruned_loss=0.05266, over 16655.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2964, pruned_loss=0.06358, over 3096838.64 frames. ], batch size: 57, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:46:12,682 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8295, 4.8540, 4.6891, 4.3762, 4.3436, 4.7912, 4.6468, 4.4348], device='cuda:3'), covar=tensor([0.0557, 0.0497, 0.0227, 0.0273, 0.0905, 0.0414, 0.0301, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0305, 0.0276, 0.0256, 0.0294, 0.0295, 0.0192, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:46:13,276 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 11:46:39,456 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-29 11:47:09,115 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8350, loss[loss=0.2304, simple_loss=0.3007, pruned_loss=0.08006, over 11865.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.295, pruned_loss=0.06097, over 3099533.40 frames. ], batch size: 248, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:16,942 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.495e+02 2.844e+02 3.376e+02 6.294e+02, threshold=5.687e+02, percent-clipped=0.0 2023-04-29 11:48:25,292 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8400, loss[loss=0.1791, simple_loss=0.266, pruned_loss=0.04616, over 15222.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2916, pruned_loss=0.05879, over 3082975.39 frames. ], batch size: 191, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:21,800 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:45,300 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8450, loss[loss=0.1949, simple_loss=0.2827, pruned_loss=0.05359, over 16670.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2895, pruned_loss=0.05689, over 3078514.30 frames. ], batch size: 124, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:52,355 INFO [optim.py:368] (3/8) 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 11:50:19,006 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7233, 2.1919, 1.8420, 1.9352, 2.4690, 2.2044, 2.5627, 2.7107], device='cuda:3'), covar=tensor([0.0117, 0.0285, 0.0365, 0.0358, 0.0195, 0.0273, 0.0168, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0197, 0.0193, 0.0192, 0.0195, 0.0195, 0.0197, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:50:23,528 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:50:55,272 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 11:51:00,422 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:09,671 INFO [train.py:904] (3/8) Epoch 11, batch 8500, loss[loss=0.1724, simple_loss=0.2708, pruned_loss=0.037, over 16508.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.285, pruned_loss=0.05414, over 3064232.21 frames. ], batch size: 75, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:51:31,231 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:33,963 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:41,404 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:13,237 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1972, 4.0005, 4.2219, 4.3906, 4.5319, 4.0390, 4.4968, 4.5320], device='cuda:3'), covar=tensor([0.1448, 0.1112, 0.1464, 0.0667, 0.0546, 0.1220, 0.0563, 0.0600], device='cuda:3'), in_proj_covar=tensor([0.0505, 0.0634, 0.0757, 0.0641, 0.0495, 0.0493, 0.0505, 0.0579], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:52:17,051 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8550, loss[loss=0.1709, simple_loss=0.2556, pruned_loss=0.0431, over 12008.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.282, pruned_loss=0.05281, over 3046209.75 frames. ], batch size: 248, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:52:37,011 INFO [optim.py:368] (3/8) 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 11:53:13,819 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5726, 2.6791, 2.3239, 3.9180, 2.4783, 4.0195, 1.3553, 2.9019], device='cuda:3'), covar=tensor([0.1432, 0.0668, 0.1210, 0.0150, 0.0133, 0.0350, 0.1649, 0.0715], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0158, 0.0180, 0.0142, 0.0197, 0.0207, 0.0181, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 11:53:19,618 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:53:51,373 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3139, 4.0131, 3.9994, 4.4325, 4.5763, 4.1726, 4.5376, 4.5889], device='cuda:3'), covar=tensor([0.1489, 0.1357, 0.2556, 0.1076, 0.0888, 0.1378, 0.0916, 0.0957], device='cuda:3'), in_proj_covar=tensor([0.0503, 0.0632, 0.0752, 0.0636, 0.0493, 0.0490, 0.0503, 0.0575], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 11:54:07,134 INFO [train.py:904] (3/8) Epoch 11, batch 8600, loss[loss=0.1759, simple_loss=0.2603, pruned_loss=0.04575, over 12595.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2823, pruned_loss=0.05193, over 3045979.59 frames. ], batch size: 248, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:54:14,052 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:22,971 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:35,027 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:55:43,444 INFO [train.py:904] (3/8) Epoch 11, batch 8650, loss[loss=0.1641, simple_loss=0.2647, pruned_loss=0.03169, over 16788.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.281, pruned_loss=0.05068, over 3040795.41 frames. ], batch size: 83, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:55:50,533 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3972, 3.2618, 3.4861, 1.7193, 3.6425, 3.7260, 2.9013, 2.7813], device='cuda:3'), covar=tensor([0.0611, 0.0220, 0.0147, 0.1069, 0.0057, 0.0100, 0.0342, 0.0381], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0097, 0.0084, 0.0134, 0.0066, 0.0098, 0.0117, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 11:55:53,865 INFO [optim.py:368] (3/8) 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 11:56:09,786 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 11:56:28,199 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:56:40,279 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2428, 3.7817, 3.7104, 2.4906, 3.3678, 3.7123, 3.4822, 2.0667], device='cuda:3'), covar=tensor([0.0438, 0.0032, 0.0028, 0.0299, 0.0068, 0.0052, 0.0049, 0.0417], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0066, 0.0067, 0.0122, 0.0076, 0.0086, 0.0075, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 11:56:40,308 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8700, loss[loss=0.193, simple_loss=0.2902, pruned_loss=0.04789, over 16883.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2782, pruned_loss=0.04937, over 3036740.51 frames. ], batch size: 116, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:58:34,008 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:58:41,801 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 11:59:05,959 INFO [train.py:904] (3/8) Epoch 11, batch 8750, loss[loss=0.2011, simple_loss=0.2947, pruned_loss=0.05371, over 16735.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2775, pruned_loss=0.04845, over 3046819.02 frames. ], batch size: 134, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:59:15,714 INFO [optim.py:368] (3/8) 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 12:00:22,508 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:00:22,702 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8800, loss[loss=0.1904, simple_loss=0.2817, pruned_loss=0.04951, over 15297.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2759, pruned_loss=0.04718, over 3055730.54 frames. ], batch size: 191, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 12:01:03,749 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:27,785 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:34,267 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:02:29,341 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8850, loss[loss=0.1948, simple_loss=0.2925, pruned_loss=0.04854, over 15242.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2785, pruned_loss=0.04666, over 3050326.01 frames. ], batch size: 190, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:02:52,480 INFO [optim.py:368] (3/8) 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 12:03:09,150 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:11,309 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:17,143 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0146, 1.9375, 2.2031, 3.5134, 1.8894, 2.2416, 2.0793, 2.0731], device='cuda:3'), covar=tensor([0.1030, 0.3514, 0.2304, 0.0475, 0.4191, 0.2447, 0.3224, 0.3417], device='cuda:3'), in_proj_covar=tensor([0.0349, 0.0382, 0.0322, 0.0310, 0.0404, 0.0432, 0.0343, 0.0445], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:03:29,778 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:42,530 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:04:21,273 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8608, 3.6757, 3.9230, 3.7722, 3.8850, 4.2928, 3.9959, 3.7558], device='cuda:3'), covar=tensor([0.1623, 0.2140, 0.1992, 0.1932, 0.2625, 0.1441, 0.1427, 0.2354], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0460, 0.0504, 0.0396, 0.0525, 0.0531, 0.0401, 0.0540], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 12:04:21,512 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8786, 1.7981, 2.2184, 3.1797, 1.9457, 1.9925, 2.0417, 1.8804], device='cuda:3'), covar=tensor([0.1133, 0.4196, 0.2297, 0.0617, 0.4792, 0.3081, 0.3510, 0.4509], device='cuda:3'), in_proj_covar=tensor([0.0350, 0.0382, 0.0323, 0.0310, 0.0405, 0.0433, 0.0344, 0.0446], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:04:23,095 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2979, 4.1143, 4.4242, 2.0568, 4.6439, 4.6893, 3.3956, 3.5313], device='cuda:3'), covar=tensor([0.0466, 0.0163, 0.0165, 0.1008, 0.0032, 0.0063, 0.0271, 0.0328], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0097, 0.0084, 0.0136, 0.0066, 0.0098, 0.0117, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 12:04:27,923 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 8900, loss[loss=0.185, simple_loss=0.2685, pruned_loss=0.05073, over 12557.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2791, pruned_loss=0.04592, over 3057944.94 frames. ], batch size: 247, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:05:40,337 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7049, 4.6893, 4.4575, 4.0535, 4.5179, 1.7721, 4.2822, 4.4843], device='cuda:3'), covar=tensor([0.0064, 0.0052, 0.0148, 0.0241, 0.0069, 0.2202, 0.0100, 0.0129], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0110, 0.0158, 0.0146, 0.0130, 0.0177, 0.0145, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:06:36,590 INFO [train.py:904] (3/8) Epoch 11, batch 8950, loss[loss=0.2093, simple_loss=0.2891, pruned_loss=0.06473, over 12445.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2789, pruned_loss=0.04661, over 3076239.71 frames. ], batch size: 246, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:06:37,830 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1824, 2.0088, 2.2675, 3.7330, 1.9957, 2.3842, 2.1234, 2.1896], device='cuda:3'), covar=tensor([0.0948, 0.3360, 0.2272, 0.0426, 0.3981, 0.2357, 0.3337, 0.3235], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0379, 0.0321, 0.0308, 0.0401, 0.0430, 0.0342, 0.0442], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:06:45,507 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.155e+02 2.674e+02 3.353e+02 7.252e+02, threshold=5.349e+02, percent-clipped=1.0 2023-04-29 12:07:08,189 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:07:17,858 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 9000, loss[loss=0.1515, simple_loss=0.2411, pruned_loss=0.03094, over 16650.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.275, pruned_loss=0.04483, over 3087764.36 frames. ], batch size: 62, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:08:26,525 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 12:08:36,935 INFO [train.py:938] (3/8) Epoch 11, validation: loss=0.1545, simple_loss=0.2586, pruned_loss=0.02523, over 944034.00 frames. 2023-04-29 12:08:36,936 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 12:09:08,812 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0619, 3.1500, 1.5772, 3.3499, 2.3331, 3.2996, 1.7318, 2.6019], device='cuda:3'), covar=tensor([0.0231, 0.0302, 0.1933, 0.0186, 0.0773, 0.0482, 0.1899, 0.0688], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0156, 0.0182, 0.0120, 0.0161, 0.0194, 0.0191, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 12:10:21,915 INFO [train.py:904] (3/8) Epoch 11, batch 9050, loss[loss=0.178, simple_loss=0.2632, pruned_loss=0.04642, over 17043.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2764, pruned_loss=0.04562, over 3085859.45 frames. ], batch size: 53, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:10:28,904 INFO [optim.py:368] (3/8) 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 12:10:45,883 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9146, 2.7569, 2.6160, 1.9126, 2.5552, 2.7571, 2.6255, 1.7440], device='cuda:3'), covar=tensor([0.0401, 0.0044, 0.0055, 0.0326, 0.0078, 0.0071, 0.0073, 0.0417], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0067, 0.0068, 0.0125, 0.0078, 0.0087, 0.0076, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 12:11:06,867 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 12:11:11,540 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9635, 3.4109, 3.5836, 1.8728, 2.9799, 2.3952, 3.4928, 3.5292], device='cuda:3'), covar=tensor([0.0245, 0.0658, 0.0416, 0.1754, 0.0688, 0.0846, 0.0604, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0137, 0.0155, 0.0141, 0.0135, 0.0123, 0.0134, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 12:12:06,405 INFO [train.py:904] (3/8) Epoch 11, batch 9100, loss[loss=0.1956, simple_loss=0.2811, pruned_loss=0.05509, over 16943.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2759, pruned_loss=0.04594, over 3108197.50 frames. ], batch size: 109, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:12:25,086 INFO [zipformer.py:625] (3/8) 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] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 12:13:28,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4063, 3.4453, 2.7279, 1.9613, 2.2198, 2.1032, 3.6175, 3.0723], device='cuda:3'), covar=tensor([0.2736, 0.0726, 0.1562, 0.2386, 0.2307, 0.1870, 0.0484, 0.1099], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0250, 0.0275, 0.0269, 0.0262, 0.0217, 0.0259, 0.0281], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:13:34,781 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7015, 4.5262, 4.7031, 4.8559, 5.0263, 4.3963, 5.0297, 5.0475], device='cuda:3'), covar=tensor([0.1322, 0.0902, 0.1273, 0.0596, 0.0426, 0.0859, 0.0357, 0.0457], device='cuda:3'), in_proj_covar=tensor([0.0495, 0.0618, 0.0734, 0.0629, 0.0481, 0.0486, 0.0495, 0.0564], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:13:36,692 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:06,380 INFO [train.py:904] (3/8) Epoch 11, batch 9150, loss[loss=0.1753, simple_loss=0.2645, pruned_loss=0.04308, over 16808.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2766, pruned_loss=0.04545, over 3124248.42 frames. ], batch size: 124, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:14:15,982 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:52,240 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:53,800 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:56,538 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:46,177 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 9200, loss[loss=0.1638, simple_loss=0.2466, pruned_loss=0.04049, over 12340.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2722, pruned_loss=0.04443, over 3122547.12 frames. ], batch size: 248, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:16:24,194 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:16:53,011 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:18,046 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 9250, loss[loss=0.161, simple_loss=0.2567, pruned_loss=0.03263, over 16656.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.272, pruned_loss=0.04475, over 3112378.37 frames. ], batch size: 134, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:17:32,955 INFO [optim.py:368] (3/8) 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 12:17:53,025 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:18:05,046 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:06,500 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 12:19:16,916 INFO [train.py:904] (3/8) Epoch 11, batch 9300, loss[loss=0.1586, simple_loss=0.2531, pruned_loss=0.03199, over 16789.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2695, pruned_loss=0.0438, over 3062992.39 frames. ], batch size: 134, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:19:17,603 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2291, 4.1340, 4.3733, 4.4754, 4.6231, 4.1147, 4.5770, 4.6243], device='cuda:3'), covar=tensor([0.1680, 0.1077, 0.1245, 0.0607, 0.0453, 0.1077, 0.0511, 0.0469], device='cuda:3'), in_proj_covar=tensor([0.0490, 0.0610, 0.0729, 0.0619, 0.0474, 0.0479, 0.0490, 0.0555], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:19:45,571 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:59,651 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 9350, loss[loss=0.2045, simple_loss=0.291, pruned_loss=0.05898, over 16327.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2697, pruned_loss=0.04415, over 3052335.81 frames. ], batch size: 146, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:21:10,062 INFO [optim.py:368] (3/8) 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 12:21:35,931 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:21:55,377 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0159, 4.0657, 4.4313, 4.4085, 4.3940, 4.1461, 4.0928, 4.0034], device='cuda:3'), covar=tensor([0.0290, 0.0558, 0.0332, 0.0366, 0.0451, 0.0374, 0.0808, 0.0438], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0312, 0.0318, 0.0296, 0.0357, 0.0333, 0.0425, 0.0269], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-29 12:22:41,698 INFO [train.py:904] (3/8) Epoch 11, batch 9400, loss[loss=0.1997, simple_loss=0.3019, pruned_loss=0.04871, over 16336.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.27, pruned_loss=0.04376, over 3052830.77 frames. ], batch size: 146, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:23:36,574 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:23:57,274 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:21,277 INFO [train.py:904] (3/8) Epoch 11, batch 9450, loss[loss=0.1699, simple_loss=0.2603, pruned_loss=0.03977, over 16739.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2721, pruned_loss=0.04405, over 3073721.07 frames. ], batch size: 83, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:24:27,346 INFO [optim.py:368] (3/8) 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 12:24:33,041 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:34,617 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:50,265 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:06,691 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:00,934 INFO [train.py:904] (3/8) Epoch 11, batch 9500, loss[loss=0.1961, simple_loss=0.2927, pruned_loss=0.04971, over 16999.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2711, pruned_loss=0.04375, over 3072983.73 frames. ], batch size: 109, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:26:15,566 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:38,099 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:47,189 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3844, 4.6572, 4.5016, 4.4878, 4.1690, 4.1284, 4.2319, 4.7018], device='cuda:3'), covar=tensor([0.1098, 0.0886, 0.0882, 0.0637, 0.0743, 0.1442, 0.0913, 0.0816], device='cuda:3'), in_proj_covar=tensor([0.0514, 0.0637, 0.0521, 0.0441, 0.0399, 0.0417, 0.0530, 0.0483], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:27:46,818 INFO [train.py:904] (3/8) Epoch 11, batch 9550, loss[loss=0.198, simple_loss=0.2932, pruned_loss=0.05137, over 15543.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2704, pruned_loss=0.04376, over 3074244.88 frames. ], batch size: 192, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:27:55,299 INFO [optim.py:368] (3/8) 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 12:29:10,467 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 9600, loss[loss=0.1951, simple_loss=0.2926, pruned_loss=0.04877, over 15482.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2712, pruned_loss=0.04414, over 3075152.21 frames. ], batch size: 191, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:29:43,745 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0605, 1.3798, 1.6700, 2.0484, 2.0823, 2.1415, 1.5797, 2.1568], device='cuda:3'), covar=tensor([0.0208, 0.0384, 0.0221, 0.0247, 0.0234, 0.0162, 0.0373, 0.0078], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0166, 0.0147, 0.0151, 0.0161, 0.0118, 0.0166, 0.0108], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 12:30:14,557 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9601, 2.0416, 2.3303, 3.2193, 2.0861, 2.2484, 2.2264, 2.0822], device='cuda:3'), covar=tensor([0.0914, 0.3302, 0.2046, 0.0526, 0.4041, 0.2319, 0.2993, 0.3426], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0374, 0.0320, 0.0304, 0.0399, 0.0423, 0.0339, 0.0436], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:31:14,961 INFO [train.py:904] (3/8) Epoch 11, batch 9650, loss[loss=0.1681, simple_loss=0.2577, pruned_loss=0.03928, over 16648.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2731, pruned_loss=0.04489, over 3053818.02 frames. ], batch size: 57, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:31:24,135 INFO [optim.py:368] (3/8) 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 12:31:52,298 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 12:33:03,901 INFO [train.py:904] (3/8) Epoch 11, batch 9700, loss[loss=0.1752, simple_loss=0.2672, pruned_loss=0.04164, over 16520.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2718, pruned_loss=0.04459, over 3049035.97 frames. ], batch size: 147, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:33:49,083 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:33:52,465 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:34:46,293 INFO [train.py:904] (3/8) Epoch 11, batch 9750, loss[loss=0.1796, simple_loss=0.2597, pruned_loss=0.04975, over 12437.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2702, pruned_loss=0.04437, over 3044623.08 frames. ], batch size: 248, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:34:53,730 INFO [optim.py:368] (3/8) 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 12:35:17,239 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:35:39,297 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6864, 3.1999, 3.4164, 1.8943, 2.9312, 2.2633, 3.2742, 3.2167], device='cuda:3'), covar=tensor([0.0284, 0.0651, 0.0415, 0.1841, 0.0698, 0.0890, 0.0654, 0.1022], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0134, 0.0153, 0.0141, 0.0133, 0.0123, 0.0133, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 12:35:58,543 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:09,650 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5562, 2.6000, 2.2510, 4.0202, 2.6291, 4.0042, 1.2724, 2.8903], device='cuda:3'), covar=tensor([0.1436, 0.0698, 0.1188, 0.0106, 0.0143, 0.0346, 0.1645, 0.0679], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0156, 0.0177, 0.0138, 0.0184, 0.0204, 0.0181, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 12:36:25,538 INFO [train.py:904] (3/8) Epoch 11, batch 9800, loss[loss=0.186, simple_loss=0.2866, pruned_loss=0.0427, over 16529.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2705, pruned_loss=0.04345, over 3054978.77 frames. ], batch size: 147, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:36:44,334 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-04-29 12:36:49,118 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:50,547 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:37:34,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0436, 4.0428, 3.8958, 3.4581, 3.9393, 1.6509, 3.7508, 3.6306], device='cuda:3'), covar=tensor([0.0080, 0.0071, 0.0138, 0.0193, 0.0077, 0.2286, 0.0103, 0.0173], device='cuda:3'), in_proj_covar=tensor([0.0124, 0.0109, 0.0156, 0.0142, 0.0128, 0.0176, 0.0144, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:38:11,446 INFO [train.py:904] (3/8) Epoch 11, batch 9850, loss[loss=0.1758, simple_loss=0.2786, pruned_loss=0.03646, over 15344.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2723, pruned_loss=0.04324, over 3056732.64 frames. ], batch size: 190, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:38:20,188 INFO [optim.py:368] (3/8) 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 12:38:23,113 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3376, 3.5154, 2.0100, 3.8768, 2.5359, 3.7775, 2.1156, 2.7090], device='cuda:3'), covar=tensor([0.0262, 0.0322, 0.1625, 0.0146, 0.0901, 0.0492, 0.1721, 0.0769], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0157, 0.0185, 0.0121, 0.0164, 0.0196, 0.0195, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 12:39:42,455 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:40:02,813 INFO [train.py:904] (3/8) Epoch 11, batch 9900, loss[loss=0.1739, simple_loss=0.2621, pruned_loss=0.04282, over 12483.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2722, pruned_loss=0.0432, over 3033052.30 frames. ], batch size: 249, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:41:34,259 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:42:01,104 INFO [train.py:904] (3/8) Epoch 11, batch 9950, loss[loss=0.1921, simple_loss=0.288, pruned_loss=0.04806, over 16646.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2744, pruned_loss=0.04383, over 3035740.29 frames. ], batch size: 134, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:42:11,464 INFO [optim.py:368] (3/8) 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 12:42:40,495 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:43:28,189 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7385, 3.6661, 4.0813, 4.0739, 4.0934, 3.8653, 3.8673, 3.8645], device='cuda:3'), covar=tensor([0.0288, 0.0723, 0.0428, 0.0441, 0.0408, 0.0403, 0.0761, 0.0386], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0309, 0.0313, 0.0294, 0.0354, 0.0330, 0.0419, 0.0267], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:3') 2023-04-29 12:44:02,040 INFO [train.py:904] (3/8) Epoch 11, batch 10000, loss[loss=0.1971, simple_loss=0.2908, pruned_loss=0.05177, over 15201.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2729, pruned_loss=0.04316, over 3057147.97 frames. ], batch size: 190, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:44:16,671 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1326, 2.7678, 2.9027, 2.0605, 2.7406, 2.1244, 2.8180, 2.8696], device='cuda:3'), covar=tensor([0.0290, 0.0683, 0.0434, 0.1504, 0.0597, 0.0872, 0.0552, 0.0749], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0132, 0.0152, 0.0139, 0.0132, 0.0121, 0.0131, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 12:44:45,195 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:55,135 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:45:10,173 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2142, 5.5818, 5.3558, 5.3928, 5.0122, 4.9169, 5.0263, 5.6332], device='cuda:3'), covar=tensor([0.1000, 0.0754, 0.0877, 0.0572, 0.0738, 0.0686, 0.0906, 0.0860], device='cuda:3'), in_proj_covar=tensor([0.0505, 0.0626, 0.0508, 0.0434, 0.0392, 0.0411, 0.0520, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:45:42,134 INFO [train.py:904] (3/8) Epoch 11, batch 10050, loss[loss=0.1652, simple_loss=0.2586, pruned_loss=0.0359, over 16682.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2736, pruned_loss=0.04317, over 3073438.57 frames. ], batch size: 62, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:45:50,236 INFO [optim.py:368] (3/8) 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 12:46:06,654 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 12:46:22,294 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:46:38,809 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 11, batch 10100, loss[loss=0.1645, simple_loss=0.256, pruned_loss=0.03655, over 12573.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2744, pruned_loss=0.04355, over 3092092.60 frames. ], batch size: 248, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:47:35,953 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 0, loss[loss=0.1776, simple_loss=0.2594, pruned_loss=0.04786, over 16968.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2594, pruned_loss=0.04786, over 16968.00 frames. ], batch size: 41, lr: 5.82e-03, grad_scale: 8.0 2023-04-29 12:48:58,023 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 12:49:05,315 INFO [train.py:938] (3/8) Epoch 12, validation: loss=0.1543, simple_loss=0.2577, pruned_loss=0.0254, over 944034.00 frames. 2023-04-29 12:49:05,316 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 12:49:12,535 INFO [optim.py:368] (3/8) 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 12:49:21,578 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:49:22,798 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9052, 5.2227, 5.3371, 5.1925, 5.1774, 5.7469, 5.2981, 5.0279], device='cuda:3'), covar=tensor([0.1061, 0.1797, 0.2272, 0.2108, 0.2465, 0.1004, 0.1591, 0.2355], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0454, 0.0497, 0.0394, 0.0519, 0.0532, 0.0400, 0.0527], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:50:01,797 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1775, 3.0129, 3.4715, 2.2576, 3.1697, 3.4738, 3.2264, 2.1145], device='cuda:3'), covar=tensor([0.0441, 0.0157, 0.0041, 0.0331, 0.0086, 0.0077, 0.0091, 0.0369], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0066, 0.0067, 0.0124, 0.0076, 0.0085, 0.0075, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 12:50:15,964 INFO [train.py:904] (3/8) Epoch 12, batch 50, loss[loss=0.2087, simple_loss=0.2843, pruned_loss=0.06653, over 16314.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2845, pruned_loss=0.06216, over 751418.76 frames. ], batch size: 145, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:25,703 INFO [train.py:904] (3/8) Epoch 12, batch 100, loss[loss=0.2346, simple_loss=0.314, pruned_loss=0.07766, over 16259.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2809, pruned_loss=0.0582, over 1328979.14 frames. ], batch size: 165, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:34,342 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.444e+02 2.855e+02 3.643e+02 7.519e+02, threshold=5.710e+02, percent-clipped=2.0 2023-04-29 12:52:18,900 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:52:31,943 INFO [train.py:904] (3/8) Epoch 12, batch 150, loss[loss=0.2378, simple_loss=0.3164, pruned_loss=0.07957, over 12132.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2786, pruned_loss=0.05606, over 1774358.16 frames. ], batch size: 248, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:02,428 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:53:07,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4839, 3.7840, 4.2832, 2.1609, 3.3115, 2.6039, 4.0714, 3.9040], device='cuda:3'), covar=tensor([0.0253, 0.0695, 0.0383, 0.1704, 0.0681, 0.0905, 0.0555, 0.1008], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0136, 0.0155, 0.0141, 0.0134, 0.0123, 0.0133, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 12:53:31,034 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3574, 3.6976, 4.1897, 2.0834, 3.2182, 2.4719, 3.8286, 3.7482], device='cuda:3'), covar=tensor([0.0273, 0.0722, 0.0372, 0.1748, 0.0678, 0.0910, 0.0571, 0.1059], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0137, 0.0156, 0.0142, 0.0135, 0.0124, 0.0134, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 12:53:40,719 INFO [train.py:904] (3/8) Epoch 12, batch 200, loss[loss=0.2238, simple_loss=0.2898, pruned_loss=0.07893, over 16520.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2776, pruned_loss=0.05595, over 2118352.39 frames. ], batch size: 75, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:41,290 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:53:50,234 INFO [optim.py:368] (3/8) 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 12:54:21,334 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 250, loss[loss=0.2468, simple_loss=0.3106, pruned_loss=0.09156, over 16472.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2754, pruned_loss=0.05608, over 2380335.28 frames. ], batch size: 68, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:55:11,926 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 2023-04-29 12:55:27,223 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:55:57,752 INFO [train.py:904] (3/8) Epoch 12, batch 300, loss[loss=0.1912, simple_loss=0.2693, pruned_loss=0.05654, over 16862.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.273, pruned_loss=0.05562, over 2585033.79 frames. ], batch size: 90, lr: 5.82e-03, grad_scale: 1.0 2023-04-29 12:56:00,457 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0043, 3.9811, 3.8973, 3.3876, 3.8933, 1.9122, 3.6852, 3.4592], device='cuda:3'), covar=tensor([0.0103, 0.0079, 0.0138, 0.0248, 0.0080, 0.2121, 0.0125, 0.0213], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0114, 0.0164, 0.0150, 0.0134, 0.0183, 0.0151, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:56:02,198 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5155, 1.6679, 1.4668, 1.4851, 1.8354, 1.5132, 1.6171, 1.8783], device='cuda:3'), covar=tensor([0.0170, 0.0241, 0.0357, 0.0326, 0.0174, 0.0227, 0.0137, 0.0178], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0208, 0.0200, 0.0201, 0.0206, 0.0205, 0.0209, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 12:56:09,471 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.376e+02 2.773e+02 3.192e+02 6.381e+02, threshold=5.545e+02, percent-clipped=1.0 2023-04-29 12:57:10,667 INFO [train.py:904] (3/8) Epoch 12, batch 350, loss[loss=0.1562, simple_loss=0.2404, pruned_loss=0.03603, over 16822.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2703, pruned_loss=0.05417, over 2742341.92 frames. ], batch size: 39, lr: 5.81e-03, grad_scale: 1.0 2023-04-29 12:58:17,743 INFO [train.py:904] (3/8) Epoch 12, batch 400, loss[loss=0.1529, simple_loss=0.2313, pruned_loss=0.03728, over 15842.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2679, pruned_loss=0.05307, over 2869280.68 frames. ], batch size: 35, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:58:27,670 INFO [optim.py:368] (3/8) 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 12:58:37,030 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:59:26,047 INFO [train.py:904] (3/8) Epoch 12, batch 450, loss[loss=0.1904, simple_loss=0.2691, pruned_loss=0.05589, over 12244.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2657, pruned_loss=0.05247, over 2960639.99 frames. ], batch size: 247, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:59:55,902 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:00,660 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:27,067 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:00:33,800 INFO [train.py:904] (3/8) Epoch 12, batch 500, loss[loss=0.1809, simple_loss=0.2694, pruned_loss=0.04622, over 17127.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2653, pruned_loss=0.05172, over 3039592.65 frames. ], batch size: 49, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:00:45,214 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.270e+02 2.759e+02 3.532e+02 6.724e+02, threshold=5.519e+02, percent-clipped=2.0 2023-04-29 13:01:02,759 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:01:19,606 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 13:01:29,252 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 13:01:32,812 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8288, 4.9191, 5.4734, 5.4287, 5.4202, 5.0797, 5.0230, 4.8035], device='cuda:3'), covar=tensor([0.0347, 0.0497, 0.0352, 0.0433, 0.0387, 0.0328, 0.0898, 0.0398], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0346, 0.0348, 0.0328, 0.0389, 0.0367, 0.0468, 0.0298], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 13:01:44,739 INFO [train.py:904] (3/8) Epoch 12, batch 550, loss[loss=0.1699, simple_loss=0.2614, pruned_loss=0.03923, over 17088.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2639, pruned_loss=0.05089, over 3101208.09 frames. ], batch size: 53, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:02:55,015 INFO [train.py:904] (3/8) Epoch 12, batch 600, loss[loss=0.1559, simple_loss=0.2499, pruned_loss=0.03091, over 17201.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2635, pruned_loss=0.05084, over 3145808.91 frames. ], batch size: 46, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:03:06,854 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.365e+02 2.773e+02 3.420e+02 1.272e+03, threshold=5.547e+02, percent-clipped=1.0 2023-04-29 13:04:02,096 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 13:04:04,844 INFO [train.py:904] (3/8) Epoch 12, batch 650, loss[loss=0.1917, simple_loss=0.2668, pruned_loss=0.05828, over 16696.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2617, pruned_loss=0.05055, over 3191997.75 frames. ], batch size: 134, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:14,225 INFO [train.py:904] (3/8) Epoch 12, batch 700, loss[loss=0.1868, simple_loss=0.2607, pruned_loss=0.05646, over 16809.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.261, pruned_loss=0.05026, over 3208236.44 frames. ], batch size: 96, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:26,018 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.509e+02 2.916e+02 3.539e+02 5.225e+02, threshold=5.832e+02, percent-clipped=0.0 2023-04-29 13:06:07,001 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:06:24,426 INFO [train.py:904] (3/8) Epoch 12, batch 750, loss[loss=0.1644, simple_loss=0.2529, pruned_loss=0.03796, over 17186.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2616, pruned_loss=0.05021, over 3233220.26 frames. ], batch size: 46, lr: 5.80e-03, grad_scale: 2.0 2023-04-29 13:06:51,816 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:19,613 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0029, 2.3821, 2.5587, 4.8428, 2.3944, 2.8889, 2.5412, 2.6038], device='cuda:3'), covar=tensor([0.0867, 0.3276, 0.2231, 0.0294, 0.3696, 0.2219, 0.2756, 0.3184], device='cuda:3'), in_proj_covar=tensor([0.0366, 0.0394, 0.0333, 0.0322, 0.0415, 0.0449, 0.0357, 0.0461], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:07:29,143 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:07:32,716 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:34,637 INFO [train.py:904] (3/8) Epoch 12, batch 800, loss[loss=0.1682, simple_loss=0.2461, pruned_loss=0.04515, over 15909.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2615, pruned_loss=0.05037, over 3248624.91 frames. ], batch size: 35, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:07:45,050 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:08:42,816 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-29 13:08:42,877 INFO [train.py:904] (3/8) Epoch 12, batch 850, loss[loss=0.1544, simple_loss=0.2314, pruned_loss=0.03873, over 16817.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2603, pruned_loss=0.04923, over 3271083.59 frames. ], batch size: 39, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:08:58,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5054, 4.3154, 4.5278, 4.7492, 4.8224, 4.3781, 4.7220, 4.7991], device='cuda:3'), covar=tensor([0.1329, 0.1044, 0.1319, 0.0526, 0.0487, 0.0995, 0.1210, 0.0577], device='cuda:3'), in_proj_covar=tensor([0.0560, 0.0693, 0.0841, 0.0703, 0.0534, 0.0540, 0.0553, 0.0635], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:09:26,307 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 2023-04-29 13:09:51,999 INFO [train.py:904] (3/8) Epoch 12, batch 900, loss[loss=0.1852, simple_loss=0.2569, pruned_loss=0.05673, over 16855.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2601, pruned_loss=0.04878, over 3292794.62 frames. ], batch size: 96, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:10:02,362 INFO [optim.py:368] (3/8) 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 13:10:05,142 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:26,066 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:35,734 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5587, 3.6563, 2.8410, 2.2025, 2.3231, 2.1747, 3.6030, 3.2331], device='cuda:3'), covar=tensor([0.2487, 0.0572, 0.1478, 0.2446, 0.2457, 0.1840, 0.0515, 0.1251], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0259, 0.0284, 0.0277, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:10:59,465 INFO [train.py:904] (3/8) Epoch 12, batch 950, loss[loss=0.1756, simple_loss=0.2477, pruned_loss=0.05176, over 16889.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2599, pruned_loss=0.04894, over 3280150.44 frames. ], batch size: 96, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:11:21,328 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:28,369 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:50,152 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:12:07,126 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7606, 3.8884, 2.1712, 4.0964, 2.8385, 4.0164, 2.2770, 3.1146], device='cuda:3'), covar=tensor([0.0220, 0.0297, 0.1481, 0.0239, 0.0660, 0.0585, 0.1329, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0166, 0.0189, 0.0134, 0.0167, 0.0208, 0.0198, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 13:12:07,851 INFO [train.py:904] (3/8) Epoch 12, batch 1000, loss[loss=0.182, simple_loss=0.2434, pruned_loss=0.0603, over 16759.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.259, pruned_loss=0.04887, over 3278758.28 frames. ], batch size: 83, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:12:18,367 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.238e+02 2.624e+02 3.056e+02 7.407e+02, threshold=5.248e+02, percent-clipped=1.0 2023-04-29 13:12:43,287 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:13:15,633 INFO [train.py:904] (3/8) Epoch 12, batch 1050, loss[loss=0.1828, simple_loss=0.2562, pruned_loss=0.05469, over 16853.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2596, pruned_loss=0.04953, over 3295008.68 frames. ], batch size: 90, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:13:42,836 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:14,519 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:17,443 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-29 13:14:22,681 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5864, 4.2502, 4.4090, 3.2174, 3.7107, 4.2553, 3.8588, 2.5154], device='cuda:3'), covar=tensor([0.0426, 0.0048, 0.0029, 0.0261, 0.0073, 0.0075, 0.0060, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0070, 0.0069, 0.0125, 0.0079, 0.0089, 0.0078, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 13:14:23,334 INFO [train.py:904] (3/8) Epoch 12, batch 1100, loss[loss=0.1906, simple_loss=0.2828, pruned_loss=0.04917, over 17017.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2598, pruned_loss=0.04941, over 3296852.37 frames. ], batch size: 50, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:14:34,070 INFO [optim.py:368] (3/8) 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 13:14:48,133 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:15:16,034 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9368, 1.7972, 2.3836, 2.9277, 2.7780, 2.9519, 2.0617, 3.1071], device='cuda:3'), covar=tensor([0.0160, 0.0380, 0.0268, 0.0189, 0.0208, 0.0195, 0.0359, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0174, 0.0155, 0.0160, 0.0169, 0.0127, 0.0173, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 13:15:33,615 INFO [train.py:904] (3/8) Epoch 12, batch 1150, loss[loss=0.16, simple_loss=0.2363, pruned_loss=0.04186, over 16910.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2595, pruned_loss=0.04901, over 3302723.57 frames. ], batch size: 90, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:16:42,797 INFO [train.py:904] (3/8) Epoch 12, batch 1200, loss[loss=0.1667, simple_loss=0.2509, pruned_loss=0.04124, over 17207.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2585, pruned_loss=0.04798, over 3312781.27 frames. ], batch size: 44, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:16:52,675 INFO [optim.py:368] (3/8) 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 13:17:49,333 INFO [train.py:904] (3/8) Epoch 12, batch 1250, loss[loss=0.1443, simple_loss=0.2246, pruned_loss=0.03203, over 15969.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2579, pruned_loss=0.04781, over 3313834.16 frames. ], batch size: 35, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:18:12,303 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:16,592 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:18:41,657 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 13:18:57,920 INFO [train.py:904] (3/8) Epoch 12, batch 1300, loss[loss=0.193, simple_loss=0.2672, pruned_loss=0.05939, over 16831.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2579, pruned_loss=0.04747, over 3317234.03 frames. ], batch size: 116, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:19:09,587 INFO [optim.py:368] (3/8) 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 13:19:11,423 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3885, 2.1152, 2.2747, 4.0729, 2.1138, 2.5259, 2.2190, 2.3235], device='cuda:3'), covar=tensor([0.1042, 0.3242, 0.2356, 0.0469, 0.3626, 0.2195, 0.3222, 0.2841], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0397, 0.0336, 0.0324, 0.0416, 0.0455, 0.0361, 0.0465], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:19:27,112 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:19:40,988 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 1350, loss[loss=0.1447, simple_loss=0.2322, pruned_loss=0.02857, over 16785.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2584, pruned_loss=0.04718, over 3320026.15 frames. ], batch size: 39, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:20:44,692 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7725, 2.4582, 1.9583, 2.3624, 2.8430, 2.6417, 2.9431, 2.9903], device='cuda:3'), covar=tensor([0.0166, 0.0267, 0.0376, 0.0313, 0.0171, 0.0232, 0.0183, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0213, 0.0205, 0.0204, 0.0212, 0.0212, 0.0217, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:21:07,353 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:21:16,368 INFO [train.py:904] (3/8) Epoch 12, batch 1400, loss[loss=0.1712, simple_loss=0.2429, pruned_loss=0.04973, over 12167.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2583, pruned_loss=0.04765, over 3300508.81 frames. ], batch size: 246, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:21:26,436 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.440e+02 3.018e+02 3.818e+02 8.239e+02, threshold=6.035e+02, percent-clipped=8.0 2023-04-29 13:21:59,480 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4725, 2.9900, 2.6981, 2.2627, 2.2690, 2.2414, 2.8826, 2.8616], device='cuda:3'), covar=tensor([0.2138, 0.0799, 0.1351, 0.1965, 0.2174, 0.1781, 0.0492, 0.1105], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0260, 0.0284, 0.0279, 0.0279, 0.0225, 0.0267, 0.0301], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:22:13,293 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:22:24,192 INFO [train.py:904] (3/8) Epoch 12, batch 1450, loss[loss=0.171, simple_loss=0.2638, pruned_loss=0.03906, over 16710.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2581, pruned_loss=0.04789, over 3301157.79 frames. ], batch size: 57, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:22:36,773 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 13:23:35,102 INFO [train.py:904] (3/8) Epoch 12, batch 1500, loss[loss=0.1915, simple_loss=0.2584, pruned_loss=0.06229, over 16907.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2576, pruned_loss=0.04766, over 3305896.24 frames. ], batch size: 109, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:23:45,775 INFO [optim.py:368] (3/8) 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 13:24:43,439 INFO [train.py:904] (3/8) Epoch 12, batch 1550, loss[loss=0.1714, simple_loss=0.2671, pruned_loss=0.03789, over 17259.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2577, pruned_loss=0.04829, over 3314669.62 frames. ], batch size: 52, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:25:06,546 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:27,338 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:28,527 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:41,539 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8141, 4.2451, 3.1313, 2.2029, 2.9654, 2.4463, 4.5528, 3.8396], device='cuda:3'), covar=tensor([0.2603, 0.0663, 0.1481, 0.2534, 0.2263, 0.1745, 0.0425, 0.1125], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0255, 0.0279, 0.0274, 0.0274, 0.0222, 0.0263, 0.0297], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:25:54,130 INFO [train.py:904] (3/8) Epoch 12, batch 1600, loss[loss=0.1994, simple_loss=0.282, pruned_loss=0.05837, over 16741.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2607, pruned_loss=0.04988, over 3309073.16 frames. ], batch size: 134, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:26:04,702 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:22,747 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:26:28,587 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:53,631 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 1650, loss[loss=0.1908, simple_loss=0.2671, pruned_loss=0.05729, over 16755.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.263, pruned_loss=0.0511, over 3302353.46 frames. ], batch size: 134, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:27:29,757 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:27:45,264 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8134, 3.9905, 3.0675, 2.3309, 2.7811, 2.4665, 4.2540, 3.7269], device='cuda:3'), covar=tensor([0.2460, 0.0687, 0.1476, 0.2229, 0.2435, 0.1781, 0.0441, 0.1071], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0258, 0.0282, 0.0278, 0.0278, 0.0224, 0.0266, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:27:57,986 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7191, 3.4473, 2.8793, 5.1995, 4.3683, 4.6263, 1.8887, 3.3559], device='cuda:3'), covar=tensor([0.1296, 0.0561, 0.1031, 0.0166, 0.0293, 0.0390, 0.1340, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0158, 0.0179, 0.0148, 0.0193, 0.0209, 0.0180, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 13:28:12,390 INFO [train.py:904] (3/8) Epoch 12, batch 1700, loss[loss=0.2035, simple_loss=0.2836, pruned_loss=0.06165, over 16274.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2651, pruned_loss=0.05108, over 3310990.72 frames. ], batch size: 165, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:28:23,616 INFO [optim.py:368] (3/8) 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 13:29:08,151 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3254, 2.0792, 2.2391, 4.0177, 2.1054, 2.5740, 2.1741, 2.2946], device='cuda:3'), covar=tensor([0.1135, 0.3448, 0.2340, 0.0482, 0.3548, 0.2094, 0.3457, 0.2791], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0398, 0.0336, 0.0325, 0.0416, 0.0456, 0.0361, 0.0467], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:29:22,273 INFO [train.py:904] (3/8) Epoch 12, batch 1750, loss[loss=0.2126, simple_loss=0.2926, pruned_loss=0.06623, over 16751.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2658, pruned_loss=0.05109, over 3322038.85 frames. ], batch size: 83, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:29:53,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6938, 4.4927, 4.7702, 4.9463, 5.1040, 4.5941, 5.0438, 5.0718], device='cuda:3'), covar=tensor([0.1556, 0.1120, 0.1561, 0.0644, 0.0463, 0.0853, 0.0622, 0.0550], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0703, 0.0851, 0.0717, 0.0539, 0.0550, 0.0557, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:30:32,314 INFO [train.py:904] (3/8) Epoch 12, batch 1800, loss[loss=0.1727, simple_loss=0.2629, pruned_loss=0.04126, over 16734.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2681, pruned_loss=0.05187, over 3314800.56 frames. ], batch size: 57, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:30:43,443 INFO [optim.py:368] (3/8) 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 13:31:03,064 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9632, 4.9168, 4.7567, 4.3031, 4.8153, 1.9800, 4.5678, 4.6163], device='cuda:3'), covar=tensor([0.0081, 0.0074, 0.0145, 0.0288, 0.0079, 0.2202, 0.0113, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0124, 0.0173, 0.0162, 0.0144, 0.0188, 0.0161, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:31:42,920 INFO [train.py:904] (3/8) Epoch 12, batch 1850, loss[loss=0.1812, simple_loss=0.2716, pruned_loss=0.04542, over 17262.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.268, pruned_loss=0.05121, over 3322295.58 frames. ], batch size: 52, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:32:53,320 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 13:32:53,583 INFO [train.py:904] (3/8) Epoch 12, batch 1900, loss[loss=0.1644, simple_loss=0.2553, pruned_loss=0.03675, over 17216.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2664, pruned_loss=0.05, over 3325298.12 frames. ], batch size: 43, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:33:04,793 INFO [optim.py:368] (3/8) 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 13:33:29,759 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:05,628 INFO [train.py:904] (3/8) Epoch 12, batch 1950, loss[loss=0.1541, simple_loss=0.2365, pruned_loss=0.03586, over 16845.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.265, pruned_loss=0.0491, over 3334075.61 frames. ], batch size: 42, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:34:27,408 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:08,138 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:16,103 INFO [train.py:904] (3/8) Epoch 12, batch 2000, loss[loss=0.1472, simple_loss=0.2288, pruned_loss=0.03283, over 16788.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2636, pruned_loss=0.04828, over 3329376.48 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:35:22,957 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 13:35:27,900 INFO [optim.py:368] (3/8) 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 13:35:52,955 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 2050, loss[loss=0.2102, simple_loss=0.2756, pruned_loss=0.07237, over 16821.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2642, pruned_loss=0.04887, over 3324814.38 frames. ], batch size: 116, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:36:32,436 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:37:33,720 INFO [train.py:904] (3/8) Epoch 12, batch 2100, loss[loss=0.2293, simple_loss=0.2981, pruned_loss=0.08025, over 16877.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2652, pruned_loss=0.0494, over 3324595.88 frames. ], batch size: 116, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:37:45,423 INFO [optim.py:368] (3/8) 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 13:37:52,286 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 13:38:44,795 INFO [train.py:904] (3/8) Epoch 12, batch 2150, loss[loss=0.1424, simple_loss=0.222, pruned_loss=0.03145, over 16743.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2668, pruned_loss=0.04965, over 3322617.64 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:39:54,111 INFO [train.py:904] (3/8) Epoch 12, batch 2200, loss[loss=0.2477, simple_loss=0.3166, pruned_loss=0.0894, over 11771.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2675, pruned_loss=0.0501, over 3316115.73 frames. ], batch size: 246, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:40:02,384 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2900, 4.4820, 4.4915, 3.3785, 3.8311, 4.3820, 3.9662, 2.6354], device='cuda:3'), covar=tensor([0.0307, 0.0051, 0.0026, 0.0246, 0.0083, 0.0071, 0.0072, 0.0343], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0072, 0.0070, 0.0125, 0.0080, 0.0090, 0.0079, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 13:40:05,118 INFO [optim.py:368] (3/8) 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 13:40:05,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0073, 2.8622, 3.1502, 2.1466, 2.9177, 3.1506, 2.9550, 1.9299], device='cuda:3'), covar=tensor([0.0408, 0.0125, 0.0047, 0.0317, 0.0104, 0.0087, 0.0082, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0072, 0.0071, 0.0125, 0.0080, 0.0090, 0.0079, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 13:40:48,886 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:41:03,704 INFO [train.py:904] (3/8) Epoch 12, batch 2250, loss[loss=0.1569, simple_loss=0.2365, pruned_loss=0.03864, over 16816.00 frames. ], tot_loss[loss=0.186, simple_loss=0.269, pruned_loss=0.05155, over 3308196.93 frames. ], batch size: 39, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:41:17,201 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7599, 2.6985, 2.4928, 2.7069, 2.9609, 2.7974, 3.5221, 3.2599], device='cuda:3'), covar=tensor([0.0083, 0.0263, 0.0320, 0.0271, 0.0202, 0.0255, 0.0163, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0211, 0.0203, 0.0203, 0.0211, 0.0209, 0.0218, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:41:54,646 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:42:14,191 INFO [train.py:904] (3/8) Epoch 12, batch 2300, loss[loss=0.1771, simple_loss=0.274, pruned_loss=0.04005, over 17254.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.269, pruned_loss=0.05123, over 3316117.17 frames. ], batch size: 52, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:42:24,215 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.413e+02 2.819e+02 3.398e+02 7.599e+02, threshold=5.637e+02, percent-clipped=2.0 2023-04-29 13:42:32,888 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8688, 3.0536, 2.5861, 4.8156, 3.6187, 4.3031, 1.5243, 2.9264], device='cuda:3'), covar=tensor([0.1321, 0.0707, 0.1218, 0.0170, 0.0274, 0.0448, 0.1633, 0.0906], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0160, 0.0180, 0.0150, 0.0196, 0.0211, 0.0181, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 13:42:43,119 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:25,572 INFO [train.py:904] (3/8) Epoch 12, batch 2350, loss[loss=0.2342, simple_loss=0.2947, pruned_loss=0.08687, over 16872.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2703, pruned_loss=0.0516, over 3309892.35 frames. ], batch size: 109, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:43:25,833 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:44:02,406 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8713, 1.8284, 2.3168, 2.7902, 2.7699, 3.0313, 1.9629, 3.1110], device='cuda:3'), covar=tensor([0.0177, 0.0399, 0.0271, 0.0217, 0.0239, 0.0154, 0.0389, 0.0102], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0178, 0.0159, 0.0165, 0.0172, 0.0130, 0.0176, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 13:44:21,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6157, 2.1803, 2.2592, 4.4009, 2.1082, 2.6668, 2.2679, 2.4001], device='cuda:3'), covar=tensor([0.0970, 0.3615, 0.2460, 0.0416, 0.4000, 0.2475, 0.3183, 0.3336], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0397, 0.0336, 0.0324, 0.0413, 0.0459, 0.0362, 0.0466], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:44:35,987 INFO [train.py:904] (3/8) Epoch 12, batch 2400, loss[loss=0.2455, simple_loss=0.3198, pruned_loss=0.08555, over 11857.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2703, pruned_loss=0.05162, over 3314688.23 frames. ], batch size: 246, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:44:48,486 INFO [optim.py:368] (3/8) 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 13:45:06,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 13:45:22,957 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-29 13:45:49,047 INFO [train.py:904] (3/8) Epoch 12, batch 2450, loss[loss=0.1581, simple_loss=0.249, pruned_loss=0.03354, over 17217.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2702, pruned_loss=0.05142, over 3318694.47 frames. ], batch size: 44, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:45:51,511 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1970, 3.4490, 3.6178, 3.5832, 3.5700, 3.4073, 3.4451, 3.4515], device='cuda:3'), covar=tensor([0.0412, 0.0562, 0.0429, 0.0452, 0.0538, 0.0470, 0.0886, 0.0508], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0365, 0.0371, 0.0344, 0.0413, 0.0389, 0.0495, 0.0314], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 13:45:52,825 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7653, 3.0804, 2.5862, 4.7065, 3.8783, 4.2675, 1.6264, 3.0487], device='cuda:3'), covar=tensor([0.1303, 0.0641, 0.1142, 0.0177, 0.0290, 0.0404, 0.1426, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0160, 0.0181, 0.0150, 0.0196, 0.0212, 0.0180, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 13:46:23,277 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:46:57,384 INFO [train.py:904] (3/8) Epoch 12, batch 2500, loss[loss=0.2324, simple_loss=0.3147, pruned_loss=0.07501, over 15450.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2703, pruned_loss=0.05092, over 3321897.35 frames. ], batch size: 191, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:47:04,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2243, 4.6260, 4.6204, 3.2745, 3.9369, 4.5101, 4.1522, 2.7722], device='cuda:3'), covar=tensor([0.0338, 0.0047, 0.0023, 0.0259, 0.0073, 0.0059, 0.0058, 0.0338], device='cuda:3'), in_proj_covar=tensor([0.0126, 0.0070, 0.0070, 0.0124, 0.0080, 0.0089, 0.0079, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 13:47:09,683 INFO [optim.py:368] (3/8) 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 13:47:21,281 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 13:47:42,650 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-29 13:47:42,847 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 13:47:48,652 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:48:06,997 INFO [train.py:904] (3/8) Epoch 12, batch 2550, loss[loss=0.1513, simple_loss=0.2351, pruned_loss=0.03372, over 16948.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2697, pruned_loss=0.05068, over 3322367.00 frames. ], batch size: 41, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:48:17,289 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 13:48:31,147 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8949, 1.8190, 2.2685, 2.7694, 2.7351, 2.9172, 1.9890, 3.0480], device='cuda:3'), covar=tensor([0.0156, 0.0388, 0.0285, 0.0238, 0.0221, 0.0192, 0.0367, 0.0101], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0179, 0.0159, 0.0166, 0.0174, 0.0131, 0.0176, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:48:52,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7143, 2.9076, 2.6034, 4.2664, 3.3856, 4.1421, 1.7771, 2.7964], device='cuda:3'), covar=tensor([0.1350, 0.0604, 0.0993, 0.0145, 0.0181, 0.0355, 0.1304, 0.0812], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0160, 0.0182, 0.0152, 0.0197, 0.0213, 0.0182, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 13:49:05,328 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1642, 5.1540, 4.8992, 4.2481, 4.9449, 2.0840, 4.7478, 4.9375], device='cuda:3'), covar=tensor([0.0082, 0.0062, 0.0174, 0.0386, 0.0087, 0.2252, 0.0120, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0125, 0.0176, 0.0163, 0.0146, 0.0187, 0.0163, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:49:15,389 INFO [train.py:904] (3/8) Epoch 12, batch 2600, loss[loss=0.1554, simple_loss=0.2418, pruned_loss=0.03444, over 16958.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2704, pruned_loss=0.05108, over 3321588.31 frames. ], batch size: 41, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:25,933 INFO [optim.py:368] (3/8) 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 13:49:45,773 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:24,179 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 13:50:24,314 INFO [train.py:904] (3/8) Epoch 12, batch 2650, loss[loss=0.171, simple_loss=0.2578, pruned_loss=0.04211, over 17011.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2711, pruned_loss=0.05102, over 3316351.87 frames. ], batch size: 41, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:50:24,617 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:41,620 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6332, 2.5302, 2.1006, 2.3613, 2.8565, 2.6795, 3.3810, 3.1478], device='cuda:3'), covar=tensor([0.0093, 0.0321, 0.0418, 0.0367, 0.0210, 0.0283, 0.0209, 0.0201], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0214, 0.0205, 0.0206, 0.0213, 0.0213, 0.0222, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:50:45,263 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0209, 1.5990, 2.3440, 2.8903, 2.8073, 2.8926, 1.7630, 3.0670], device='cuda:3'), covar=tensor([0.0137, 0.0479, 0.0259, 0.0181, 0.0196, 0.0172, 0.0492, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0177, 0.0157, 0.0164, 0.0172, 0.0130, 0.0174, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 13:50:51,416 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:52,013 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 2023-04-29 13:51:32,419 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:35,147 INFO [train.py:904] (3/8) Epoch 12, batch 2700, loss[loss=0.1888, simple_loss=0.2619, pruned_loss=0.05785, over 16740.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2709, pruned_loss=0.05037, over 3318911.45 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:51:45,484 INFO [optim.py:368] (3/8) 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 13:52:24,793 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5590, 4.5192, 4.4831, 3.9322, 4.4914, 1.8057, 4.2630, 4.2543], device='cuda:3'), covar=tensor([0.0096, 0.0083, 0.0143, 0.0322, 0.0095, 0.2330, 0.0127, 0.0185], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0126, 0.0177, 0.0165, 0.0147, 0.0189, 0.0164, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:52:28,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7602, 2.5264, 2.3467, 3.4621, 2.8269, 3.6327, 1.4375, 2.7154], device='cuda:3'), covar=tensor([0.1238, 0.0634, 0.1076, 0.0152, 0.0174, 0.0396, 0.1437, 0.0735], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0159, 0.0180, 0.0151, 0.0196, 0.0211, 0.0181, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 13:52:36,523 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 13:52:44,857 INFO [train.py:904] (3/8) Epoch 12, batch 2750, loss[loss=0.1944, simple_loss=0.2654, pruned_loss=0.06172, over 16769.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.271, pruned_loss=0.04974, over 3327943.53 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:53:28,989 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 13:53:49,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6946, 2.5751, 2.1258, 2.3282, 2.9489, 2.7125, 3.4059, 3.2416], device='cuda:3'), covar=tensor([0.0102, 0.0347, 0.0418, 0.0399, 0.0222, 0.0315, 0.0210, 0.0224], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0214, 0.0204, 0.0206, 0.0213, 0.0212, 0.0222, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:53:54,633 INFO [train.py:904] (3/8) Epoch 12, batch 2800, loss[loss=0.1862, simple_loss=0.2591, pruned_loss=0.05668, over 16819.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2697, pruned_loss=0.04864, over 3335935.10 frames. ], batch size: 116, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:54:05,298 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2819, 5.6057, 5.3239, 5.4282, 4.9914, 4.9320, 5.0428, 5.6876], device='cuda:3'), covar=tensor([0.1021, 0.0810, 0.0986, 0.0669, 0.0838, 0.0716, 0.0984, 0.0891], device='cuda:3'), in_proj_covar=tensor([0.0581, 0.0721, 0.0592, 0.0504, 0.0456, 0.0461, 0.0599, 0.0556], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:54:06,072 INFO [optim.py:368] (3/8) 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 13:54:24,440 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4695, 2.1995, 1.6662, 1.9677, 2.5798, 2.3623, 2.6456, 2.7146], device='cuda:3'), covar=tensor([0.0143, 0.0289, 0.0415, 0.0371, 0.0179, 0.0255, 0.0191, 0.0213], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0214, 0.0205, 0.0207, 0.0214, 0.0213, 0.0224, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:54:38,676 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:54:39,900 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6731, 4.5299, 4.6607, 4.8850, 4.9895, 4.4470, 4.9212, 4.9849], device='cuda:3'), covar=tensor([0.1469, 0.0978, 0.1543, 0.0591, 0.0578, 0.1024, 0.0814, 0.0572], device='cuda:3'), in_proj_covar=tensor([0.0572, 0.0715, 0.0868, 0.0728, 0.0549, 0.0563, 0.0569, 0.0656], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:55:04,045 INFO [train.py:904] (3/8) Epoch 12, batch 2850, loss[loss=0.1608, simple_loss=0.2488, pruned_loss=0.03638, over 17215.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2698, pruned_loss=0.04911, over 3335409.56 frames. ], batch size: 45, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:55:57,162 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 13:56:13,202 INFO [train.py:904] (3/8) Epoch 12, batch 2900, loss[loss=0.174, simple_loss=0.2504, pruned_loss=0.0488, over 16734.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.269, pruned_loss=0.05003, over 3331215.00 frames. ], batch size: 83, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:56:24,537 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.459e+02 2.961e+02 3.423e+02 5.764e+02, threshold=5.923e+02, percent-clipped=1.0 2023-04-29 13:56:46,813 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7242, 2.7222, 2.3560, 3.9519, 3.2772, 3.9935, 1.4445, 2.8044], device='cuda:3'), covar=tensor([0.1253, 0.0544, 0.1015, 0.0134, 0.0132, 0.0331, 0.1341, 0.0728], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0160, 0.0180, 0.0152, 0.0196, 0.0212, 0.0181, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 13:56:52,080 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 13:57:20,462 INFO [train.py:904] (3/8) Epoch 12, batch 2950, loss[loss=0.1806, simple_loss=0.2548, pruned_loss=0.05325, over 16805.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2682, pruned_loss=0.05066, over 3331702.88 frames. ], batch size: 102, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:57:25,086 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9109, 4.6542, 4.8933, 5.1141, 5.2798, 4.6469, 5.2602, 5.2252], device='cuda:3'), covar=tensor([0.1483, 0.1050, 0.1579, 0.0605, 0.0488, 0.0803, 0.0493, 0.0571], device='cuda:3'), in_proj_covar=tensor([0.0570, 0.0710, 0.0859, 0.0722, 0.0543, 0.0557, 0.0565, 0.0648], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 13:58:28,575 INFO [train.py:904] (3/8) Epoch 12, batch 3000, loss[loss=0.1739, simple_loss=0.2653, pruned_loss=0.04125, over 17175.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2685, pruned_loss=0.05059, over 3334484.92 frames. ], batch size: 46, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:28,575 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 13:58:38,474 INFO [train.py:938] (3/8) Epoch 12, validation: loss=0.14, simple_loss=0.2459, pruned_loss=0.01708, over 944034.00 frames. 2023-04-29 13:58:38,475 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 13:58:43,747 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2252, 4.0334, 4.4972, 2.0612, 4.6494, 4.6544, 3.4743, 3.6613], device='cuda:3'), covar=tensor([0.0517, 0.0159, 0.0122, 0.0998, 0.0045, 0.0094, 0.0280, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0101, 0.0088, 0.0137, 0.0070, 0.0108, 0.0120, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 13:58:50,228 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 12, batch 3050, loss[loss=0.2009, simple_loss=0.2926, pruned_loss=0.05462, over 16746.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.268, pruned_loss=0.05066, over 3327331.96 frames. ], batch size: 57, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 14:00:08,898 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:00:33,000 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7403, 1.7311, 1.5018, 1.5399, 1.7778, 1.5115, 1.6459, 1.9513], device='cuda:3'), covar=tensor([0.0135, 0.0226, 0.0292, 0.0265, 0.0166, 0.0233, 0.0148, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0212, 0.0203, 0.0205, 0.0214, 0.0213, 0.0221, 0.0204], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:00:56,624 INFO [train.py:904] (3/8) Epoch 12, batch 3100, loss[loss=0.1872, simple_loss=0.2668, pruned_loss=0.0538, over 16510.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2678, pruned_loss=0.05089, over 3315984.86 frames. ], batch size: 68, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:01:04,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0113, 1.8017, 2.3430, 2.8694, 2.5760, 3.2705, 1.9130, 3.2230], device='cuda:3'), covar=tensor([0.0155, 0.0408, 0.0262, 0.0222, 0.0244, 0.0151, 0.0400, 0.0131], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0177, 0.0159, 0.0165, 0.0175, 0.0131, 0.0175, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 14:01:10,357 INFO [optim.py:368] (3/8) 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 14:01:31,911 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:01:40,391 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:01:56,913 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 14:02:05,275 INFO [train.py:904] (3/8) Epoch 12, batch 3150, loss[loss=0.1715, simple_loss=0.257, pruned_loss=0.04299, over 17245.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2674, pruned_loss=0.05052, over 3321986.93 frames. ], batch size: 45, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:02:45,191 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:03:14,012 INFO [train.py:904] (3/8) Epoch 12, batch 3200, loss[loss=0.1853, simple_loss=0.2747, pruned_loss=0.04793, over 17040.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2666, pruned_loss=0.04971, over 3313847.56 frames. ], batch size: 55, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:03:26,054 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:03:43,252 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 14:04:22,405 INFO [train.py:904] (3/8) Epoch 12, batch 3250, loss[loss=0.237, simple_loss=0.3095, pruned_loss=0.08225, over 15691.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2671, pruned_loss=0.05045, over 3303474.60 frames. ], batch size: 191, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:04:56,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 14:05:05,297 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 3300, loss[loss=0.1798, simple_loss=0.2702, pruned_loss=0.04466, over 16689.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2671, pruned_loss=0.04981, over 3317638.18 frames. ], batch size: 57, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:45,364 INFO [optim.py:368] (3/8) 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 14:06:01,177 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4943, 4.1994, 4.6019, 2.1223, 4.7848, 4.7188, 3.4235, 3.9632], device='cuda:3'), covar=tensor([0.0506, 0.0174, 0.0175, 0.1043, 0.0053, 0.0111, 0.0345, 0.0251], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0138, 0.0070, 0.0109, 0.0122, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 14:06:10,281 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-29 14:06:22,011 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1344, 3.4061, 3.3701, 1.9809, 2.8167, 2.4667, 3.6514, 3.5180], device='cuda:3'), covar=tensor([0.0242, 0.0756, 0.0629, 0.1763, 0.0787, 0.0886, 0.0500, 0.0863], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0149, 0.0159, 0.0145, 0.0137, 0.0125, 0.0138, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 14:06:41,221 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1239, 5.5971, 5.8277, 5.5069, 5.5709, 6.1785, 5.7019, 5.4021], device='cuda:3'), covar=tensor([0.0887, 0.1715, 0.1793, 0.2139, 0.2658, 0.0970, 0.1255, 0.2308], device='cuda:3'), in_proj_covar=tensor([0.0361, 0.0516, 0.0558, 0.0441, 0.0599, 0.0583, 0.0443, 0.0593], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 14:06:42,021 INFO [train.py:904] (3/8) Epoch 12, batch 3350, loss[loss=0.1744, simple_loss=0.2642, pruned_loss=0.04227, over 17218.00 frames. ], tot_loss[loss=0.184, simple_loss=0.268, pruned_loss=0.04996, over 3317947.27 frames. ], batch size: 46, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:07:50,786 INFO [train.py:904] (3/8) Epoch 12, batch 3400, loss[loss=0.2276, simple_loss=0.2939, pruned_loss=0.0806, over 16882.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2671, pruned_loss=0.04939, over 3320583.45 frames. ], batch size: 109, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:08:04,044 INFO [optim.py:368] (3/8) 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 14:08:08,697 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7940, 4.0638, 3.1336, 2.3747, 2.7993, 2.4894, 4.1159, 3.6910], device='cuda:3'), covar=tensor([0.2491, 0.0596, 0.1537, 0.2361, 0.2326, 0.1734, 0.0500, 0.1057], device='cuda:3'), in_proj_covar=tensor([0.0303, 0.0261, 0.0285, 0.0282, 0.0285, 0.0225, 0.0271, 0.0305], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 14:08:18,294 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:37,801 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:56,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9941, 2.1136, 2.5834, 2.9180, 2.8840, 3.4647, 2.2217, 3.3884], device='cuda:3'), covar=tensor([0.0178, 0.0361, 0.0228, 0.0248, 0.0231, 0.0125, 0.0358, 0.0120], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0160, 0.0165, 0.0177, 0.0132, 0.0176, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:09:00,023 INFO [train.py:904] (3/8) Epoch 12, batch 3450, loss[loss=0.1598, simple_loss=0.2385, pruned_loss=0.04056, over 16852.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2661, pruned_loss=0.04916, over 3331750.45 frames. ], batch size: 96, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:09:15,955 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 14:10:01,667 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:10:03,552 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6725, 2.2938, 2.2712, 4.4126, 2.2578, 2.6985, 2.3251, 2.4227], device='cuda:3'), covar=tensor([0.0900, 0.3322, 0.2433, 0.0393, 0.3637, 0.2379, 0.2924, 0.3420], device='cuda:3'), in_proj_covar=tensor([0.0375, 0.0400, 0.0337, 0.0328, 0.0415, 0.0463, 0.0364, 0.0469], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:10:08,328 INFO [train.py:904] (3/8) Epoch 12, batch 3500, loss[loss=0.1993, simple_loss=0.2746, pruned_loss=0.06194, over 16721.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.265, pruned_loss=0.04954, over 3326271.14 frames. ], batch size: 134, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:10:23,310 INFO [optim.py:368] (3/8) 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 14:10:36,835 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6642, 4.6316, 4.5713, 4.0647, 4.5487, 1.8679, 4.3756, 4.3274], device='cuda:3'), covar=tensor([0.0098, 0.0078, 0.0135, 0.0288, 0.0099, 0.2298, 0.0125, 0.0167], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0128, 0.0178, 0.0168, 0.0149, 0.0189, 0.0167, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:11:07,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6829, 3.9021, 2.3452, 4.3111, 2.8942, 4.3190, 2.2332, 2.9895], device='cuda:3'), covar=tensor([0.0254, 0.0308, 0.1350, 0.0229, 0.0703, 0.0422, 0.1480, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0167, 0.0188, 0.0140, 0.0168, 0.0213, 0.0197, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 14:11:19,452 INFO [train.py:904] (3/8) Epoch 12, batch 3550, loss[loss=0.1908, simple_loss=0.2666, pruned_loss=0.05749, over 12006.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2639, pruned_loss=0.04918, over 3322998.25 frames. ], batch size: 247, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:11:45,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9251, 4.5716, 3.3109, 2.3764, 2.9135, 2.6322, 4.7929, 3.7629], device='cuda:3'), covar=tensor([0.2507, 0.0582, 0.1470, 0.2335, 0.2599, 0.1732, 0.0374, 0.1179], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0259, 0.0284, 0.0280, 0.0284, 0.0223, 0.0268, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:11:53,651 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:12:28,659 INFO [train.py:904] (3/8) Epoch 12, batch 3600, loss[loss=0.1425, simple_loss=0.235, pruned_loss=0.02504, over 17233.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.262, pruned_loss=0.04813, over 3322899.70 frames. ], batch size: 45, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:12:43,852 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.259e+02 2.631e+02 3.384e+02 1.021e+03, threshold=5.262e+02, percent-clipped=2.0 2023-04-29 14:13:40,320 INFO [train.py:904] (3/8) Epoch 12, batch 3650, loss[loss=0.1963, simple_loss=0.2739, pruned_loss=0.05933, over 11441.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2617, pruned_loss=0.04902, over 3312565.68 frames. ], batch size: 246, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:14:05,923 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9641, 3.2886, 2.8630, 4.7196, 3.9442, 4.4036, 1.7996, 3.2727], device='cuda:3'), covar=tensor([0.1208, 0.0543, 0.0945, 0.0151, 0.0241, 0.0373, 0.1301, 0.0720], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0161, 0.0181, 0.0154, 0.0200, 0.0214, 0.0183, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 14:14:10,782 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7663, 3.8988, 2.2012, 4.2708, 2.9271, 4.1704, 2.2558, 2.9179], device='cuda:3'), covar=tensor([0.0222, 0.0322, 0.1449, 0.0219, 0.0676, 0.0508, 0.1406, 0.0688], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0168, 0.0189, 0.0141, 0.0169, 0.0214, 0.0199, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 14:14:49,269 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0043, 1.8924, 2.4804, 2.9927, 2.9024, 2.9505, 1.9231, 3.1309], device='cuda:3'), covar=tensor([0.0141, 0.0363, 0.0245, 0.0195, 0.0195, 0.0185, 0.0380, 0.0103], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0160, 0.0165, 0.0177, 0.0132, 0.0177, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:14:55,143 INFO [train.py:904] (3/8) Epoch 12, batch 3700, loss[loss=0.1952, simple_loss=0.2745, pruned_loss=0.05795, over 16556.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2603, pruned_loss=0.05033, over 3275015.69 frames. ], batch size: 62, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:15:09,328 INFO [optim.py:368] (3/8) 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 14:15:22,976 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:24,687 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:25,800 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:32,844 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9152, 3.9583, 4.2654, 4.2760, 4.2675, 4.0159, 4.0510, 3.9692], device='cuda:3'), covar=tensor([0.0335, 0.0580, 0.0360, 0.0368, 0.0449, 0.0380, 0.0720, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0371, 0.0369, 0.0345, 0.0416, 0.0387, 0.0496, 0.0312], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 14:16:10,075 INFO [train.py:904] (3/8) Epoch 12, batch 3750, loss[loss=0.1687, simple_loss=0.2428, pruned_loss=0.04724, over 16772.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2615, pruned_loss=0.05232, over 3262053.69 frames. ], batch size: 83, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:16:26,453 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8221, 1.7771, 2.2714, 2.7003, 2.7765, 2.6065, 1.6832, 2.8886], device='cuda:3'), covar=tensor([0.0132, 0.0360, 0.0261, 0.0201, 0.0183, 0.0203, 0.0389, 0.0089], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0160, 0.0165, 0.0177, 0.0132, 0.0177, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:16:36,698 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:54,171 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:55,359 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 12, batch 3800, loss[loss=0.1819, simple_loss=0.2524, pruned_loss=0.05567, over 16732.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.263, pruned_loss=0.05357, over 3256594.30 frames. ], batch size: 89, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:17:38,965 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 12, batch 3850, loss[loss=0.2055, simple_loss=0.2765, pruned_loss=0.06726, over 16751.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2624, pruned_loss=0.05386, over 3272559.98 frames. ], batch size: 134, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:19:16,981 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:19:26,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9701, 1.7904, 2.5284, 2.9098, 2.9606, 2.9040, 1.8140, 3.0471], device='cuda:3'), covar=tensor([0.0100, 0.0396, 0.0198, 0.0164, 0.0150, 0.0178, 0.0424, 0.0086], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0177, 0.0159, 0.0163, 0.0175, 0.0131, 0.0175, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:19:52,700 INFO [train.py:904] (3/8) Epoch 12, batch 3900, loss[loss=0.1694, simple_loss=0.2484, pruned_loss=0.04524, over 16255.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2618, pruned_loss=0.05391, over 3282651.57 frames. ], batch size: 165, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:20:07,961 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.153e+02 2.652e+02 3.211e+02 6.333e+02, threshold=5.304e+02, percent-clipped=2.0 2023-04-29 14:20:29,517 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:21:08,905 INFO [train.py:904] (3/8) Epoch 12, batch 3950, loss[loss=0.2181, simple_loss=0.2819, pruned_loss=0.07716, over 16913.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.262, pruned_loss=0.05468, over 3282612.00 frames. ], batch size: 116, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:21:12,117 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-29 14:22:00,867 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 14:22:21,470 INFO [train.py:904] (3/8) Epoch 12, batch 4000, loss[loss=0.1761, simple_loss=0.2564, pruned_loss=0.04792, over 16652.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2622, pruned_loss=0.05493, over 3284505.64 frames. ], batch size: 57, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:22:34,737 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.376e+02 2.718e+02 3.245e+02 5.190e+02, threshold=5.435e+02, percent-clipped=0.0 2023-04-29 14:23:35,774 INFO [train.py:904] (3/8) Epoch 12, batch 4050, loss[loss=0.2059, simple_loss=0.2834, pruned_loss=0.06425, over 16265.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2627, pruned_loss=0.05424, over 3269499.72 frames. ], batch size: 165, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:12,118 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:13,874 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:30,915 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1056, 5.0545, 4.8933, 3.9220, 5.0101, 1.7456, 4.7200, 4.5021], device='cuda:3'), covar=tensor([0.0078, 0.0075, 0.0133, 0.0458, 0.0073, 0.2810, 0.0128, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0127, 0.0175, 0.0166, 0.0147, 0.0188, 0.0165, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:24:35,723 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:50,500 INFO [train.py:904] (3/8) Epoch 12, batch 4100, loss[loss=0.189, simple_loss=0.2781, pruned_loss=0.04998, over 16849.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.264, pruned_loss=0.05331, over 3264742.58 frames. ], batch size: 102, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:57,530 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9105, 5.0227, 5.2375, 5.0692, 5.0566, 5.6703, 5.2059, 4.9920], device='cuda:3'), covar=tensor([0.0785, 0.1565, 0.1564, 0.1695, 0.2264, 0.0817, 0.1150, 0.1939], device='cuda:3'), in_proj_covar=tensor([0.0357, 0.0507, 0.0548, 0.0433, 0.0583, 0.0572, 0.0438, 0.0587], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 14:24:57,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7066, 1.8143, 2.2673, 2.7139, 2.7105, 3.0562, 1.9516, 2.9372], device='cuda:3'), covar=tensor([0.0191, 0.0381, 0.0260, 0.0218, 0.0211, 0.0126, 0.0375, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0177, 0.0160, 0.0164, 0.0176, 0.0132, 0.0176, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:25:05,531 INFO [optim.py:368] (3/8) 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 14:25:27,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2736, 2.5010, 2.0302, 2.3395, 2.8761, 2.5338, 3.0957, 3.1265], device='cuda:3'), covar=tensor([0.0087, 0.0308, 0.0397, 0.0349, 0.0170, 0.0318, 0.0144, 0.0184], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0209, 0.0202, 0.0203, 0.0210, 0.0208, 0.0217, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:25:48,280 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:26:05,810 INFO [train.py:904] (3/8) Epoch 12, batch 4150, loss[loss=0.1862, simple_loss=0.2755, pruned_loss=0.04845, over 16561.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2714, pruned_loss=0.05599, over 3229974.66 frames. ], batch size: 62, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:26:52,472 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-29 14:27:23,037 INFO [train.py:904] (3/8) Epoch 12, batch 4200, loss[loss=0.211, simple_loss=0.3087, pruned_loss=0.05667, over 16718.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2782, pruned_loss=0.05733, over 3213989.73 frames. ], batch size: 89, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:37,142 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.546e+02 2.889e+02 3.538e+02 7.743e+02, threshold=5.778e+02, percent-clipped=11.0 2023-04-29 14:27:47,233 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8245, 3.8003, 4.2442, 1.8906, 4.5461, 4.4900, 3.1367, 3.3520], device='cuda:3'), covar=tensor([0.0690, 0.0210, 0.0162, 0.1140, 0.0039, 0.0084, 0.0385, 0.0348], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0101, 0.0088, 0.0139, 0.0071, 0.0108, 0.0123, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 14:28:22,481 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:28:36,619 INFO [train.py:904] (3/8) Epoch 12, batch 4250, loss[loss=0.2023, simple_loss=0.2936, pruned_loss=0.05553, over 17006.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2816, pruned_loss=0.05783, over 3181758.64 frames. ], batch size: 55, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:29:09,590 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4709, 4.2579, 4.2762, 2.7569, 3.7925, 4.2905, 3.7906, 2.2973], device='cuda:3'), covar=tensor([0.0441, 0.0025, 0.0033, 0.0338, 0.0060, 0.0067, 0.0057, 0.0382], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0071, 0.0072, 0.0127, 0.0081, 0.0090, 0.0080, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 14:29:49,145 INFO [train.py:904] (3/8) Epoch 12, batch 4300, loss[loss=0.1906, simple_loss=0.2755, pruned_loss=0.05283, over 11404.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2823, pruned_loss=0.05661, over 3165655.58 frames. ], batch size: 247, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:29:50,936 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:30:01,827 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9894, 4.0877, 3.8329, 3.6510, 3.6210, 3.9831, 3.6071, 3.7280], device='cuda:3'), covar=tensor([0.0550, 0.0372, 0.0258, 0.0244, 0.0758, 0.0358, 0.0953, 0.0501], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0329, 0.0301, 0.0278, 0.0320, 0.0321, 0.0203, 0.0349], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:30:04,741 INFO [optim.py:368] (3/8) 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 14:31:07,410 INFO [train.py:904] (3/8) Epoch 12, batch 4350, loss[loss=0.2082, simple_loss=0.2951, pruned_loss=0.0607, over 16807.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2861, pruned_loss=0.05815, over 3166257.49 frames. ], batch size: 124, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:31:14,495 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8426, 4.8287, 4.6217, 4.0622, 4.7814, 1.5742, 4.5358, 4.3378], device='cuda:3'), covar=tensor([0.0051, 0.0040, 0.0105, 0.0262, 0.0055, 0.2529, 0.0079, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0123, 0.0169, 0.0161, 0.0142, 0.0183, 0.0159, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:31:27,417 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:46,023 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:47,877 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:06,615 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3010, 1.6208, 1.9531, 2.3270, 2.3993, 2.6475, 1.6537, 2.5199], device='cuda:3'), covar=tensor([0.0185, 0.0350, 0.0208, 0.0214, 0.0200, 0.0114, 0.0352, 0.0099], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0176, 0.0158, 0.0162, 0.0174, 0.0129, 0.0175, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 14:32:22,077 INFO [train.py:904] (3/8) Epoch 12, batch 4400, loss[loss=0.2313, simple_loss=0.2953, pruned_loss=0.08367, over 11763.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2885, pruned_loss=0.05903, over 3169540.36 frames. ], batch size: 248, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:32:37,554 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.512e+02 3.045e+02 3.645e+02 6.621e+02, threshold=6.090e+02, percent-clipped=4.0 2023-04-29 14:32:54,755 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:57,359 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:57,514 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:33:35,426 INFO [train.py:904] (3/8) Epoch 12, batch 4450, loss[loss=0.2029, simple_loss=0.2935, pruned_loss=0.05616, over 16660.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2924, pruned_loss=0.06016, over 3192893.04 frames. ], batch size: 62, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:33:40,078 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:33:56,995 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:34:17,898 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 14:34:18,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3640, 2.9522, 2.5958, 2.2212, 2.1706, 2.1316, 2.9239, 2.7945], device='cuda:3'), covar=tensor([0.2214, 0.0780, 0.1422, 0.2090, 0.2090, 0.1859, 0.0525, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0261, 0.0288, 0.0283, 0.0289, 0.0226, 0.0272, 0.0302], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:34:32,467 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0138, 4.1453, 4.4113, 4.3572, 4.3913, 4.0786, 4.1116, 3.9801], device='cuda:3'), covar=tensor([0.0270, 0.0392, 0.0318, 0.0384, 0.0390, 0.0358, 0.0851, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0350, 0.0352, 0.0333, 0.0398, 0.0372, 0.0476, 0.0298], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 14:34:49,327 INFO [train.py:904] (3/8) Epoch 12, batch 4500, loss[loss=0.2282, simple_loss=0.3126, pruned_loss=0.07193, over 16785.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2932, pruned_loss=0.0611, over 3205378.28 frames. ], batch size: 124, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:35:03,473 INFO [optim.py:368] (3/8) 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 14:35:10,922 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:25,617 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 4550, loss[loss=0.2328, simple_loss=0.3172, pruned_loss=0.07418, over 15356.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2932, pruned_loss=0.06157, over 3213899.95 frames. ], batch size: 190, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:36:05,666 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7072, 3.0615, 2.8126, 4.4886, 3.6287, 4.2551, 1.5819, 2.9080], device='cuda:3'), covar=tensor([0.1309, 0.0608, 0.0993, 0.0104, 0.0302, 0.0306, 0.1498, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0181, 0.0151, 0.0200, 0.0209, 0.0181, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 14:36:23,022 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5276, 3.5565, 3.4232, 2.7817, 3.3513, 2.0067, 3.2240, 2.5970], device='cuda:3'), covar=tensor([0.0101, 0.0077, 0.0131, 0.0209, 0.0077, 0.2116, 0.0100, 0.0177], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0122, 0.0169, 0.0160, 0.0141, 0.0182, 0.0158, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:37:08,830 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:37:14,079 INFO [train.py:904] (3/8) Epoch 12, batch 4600, loss[loss=0.1788, simple_loss=0.2717, pruned_loss=0.04296, over 16618.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2942, pruned_loss=0.06188, over 3216832.23 frames. ], batch size: 75, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:37:28,974 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-29 14:37:29,430 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.985e+02 2.258e+02 2.658e+02 3.690e+02, threshold=4.517e+02, percent-clipped=0.0 2023-04-29 14:38:26,067 INFO [train.py:904] (3/8) Epoch 12, batch 4650, loss[loss=0.2215, simple_loss=0.2933, pruned_loss=0.0749, over 11716.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2933, pruned_loss=0.06196, over 3201779.48 frames. ], batch size: 246, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:38:55,759 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:39:26,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9766, 3.4259, 3.4901, 2.0015, 2.8463, 2.2739, 3.4054, 3.6571], device='cuda:3'), covar=tensor([0.0375, 0.0745, 0.0599, 0.1855, 0.0922, 0.0917, 0.0796, 0.0851], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0148, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 14:39:38,311 INFO [train.py:904] (3/8) Epoch 12, batch 4700, loss[loss=0.1715, simple_loss=0.2618, pruned_loss=0.04059, over 16795.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2902, pruned_loss=0.06032, over 3200775.21 frames. ], batch size: 83, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:39:53,945 INFO [optim.py:368] (3/8) 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 14:40:05,806 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:40:27,505 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:40:54,105 INFO [train.py:904] (3/8) Epoch 12, batch 4750, loss[loss=0.1937, simple_loss=0.2753, pruned_loss=0.05607, over 16655.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2863, pruned_loss=0.05833, over 3208525.70 frames. ], batch size: 134, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:41:59,005 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:07,120 INFO [train.py:904] (3/8) Epoch 12, batch 4800, loss[loss=0.1971, simple_loss=0.2931, pruned_loss=0.05059, over 16358.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2828, pruned_loss=0.05613, over 3195736.39 frames. ], batch size: 146, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:42:22,127 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.880e+02 2.223e+02 2.728e+02 4.441e+02, threshold=4.446e+02, percent-clipped=0.0 2023-04-29 14:42:27,840 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9542, 4.8269, 4.6755, 3.3598, 3.8896, 4.6543, 4.0521, 2.5115], device='cuda:3'), covar=tensor([0.0359, 0.0015, 0.0023, 0.0253, 0.0070, 0.0053, 0.0066, 0.0338], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0069, 0.0070, 0.0125, 0.0081, 0.0090, 0.0080, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 14:42:39,712 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:23,406 INFO [train.py:904] (3/8) Epoch 12, batch 4850, loss[loss=0.1774, simple_loss=0.2727, pruned_loss=0.04108, over 16952.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.283, pruned_loss=0.05528, over 3184958.05 frames. ], batch size: 109, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:43:28,314 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:30,807 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:43:46,613 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 14:44:04,984 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 14:44:05,276 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 14:44:31,767 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:44:38,142 INFO [train.py:904] (3/8) Epoch 12, batch 4900, loss[loss=0.2081, simple_loss=0.2835, pruned_loss=0.0663, over 12263.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2824, pruned_loss=0.05437, over 3166266.27 frames. ], batch size: 246, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:44:52,630 INFO [optim.py:368] (3/8) 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 14:44:58,520 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:45:14,794 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 14:45:42,831 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 4950, loss[loss=0.2088, simple_loss=0.2954, pruned_loss=0.06109, over 15379.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2814, pruned_loss=0.05357, over 3174560.41 frames. ], batch size: 191, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:46:16,257 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:46:27,632 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 14:47:04,268 INFO [train.py:904] (3/8) Epoch 12, batch 5000, loss[loss=0.1937, simple_loss=0.2824, pruned_loss=0.05247, over 12382.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2827, pruned_loss=0.0538, over 3187817.22 frames. ], batch size: 247, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:47:17,030 INFO [optim.py:368] (3/8) 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 14:47:30,353 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:47:43,242 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:47:43,363 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:05,558 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4435, 2.4575, 2.0884, 2.3059, 2.8519, 2.5470, 3.1745, 3.1317], device='cuda:3'), covar=tensor([0.0059, 0.0299, 0.0388, 0.0298, 0.0180, 0.0246, 0.0126, 0.0167], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0206, 0.0203, 0.0200, 0.0206, 0.0204, 0.0210, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:48:07,635 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-29 14:48:15,688 INFO [train.py:904] (3/8) Epoch 12, batch 5050, loss[loss=0.1925, simple_loss=0.2894, pruned_loss=0.04784, over 16480.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2826, pruned_loss=0.05335, over 3216194.89 frames. ], batch size: 146, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:48:18,315 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:38,166 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:54,632 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 14:49:24,608 INFO [train.py:904] (3/8) Epoch 12, batch 5100, loss[loss=0.1735, simple_loss=0.2638, pruned_loss=0.04156, over 16264.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2811, pruned_loss=0.05283, over 3216886.39 frames. ], batch size: 165, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:49:37,402 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:49:43,669 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:52,568 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:20,508 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2944, 3.1801, 3.1122, 3.4549, 3.4124, 3.2344, 3.4325, 3.5051], device='cuda:3'), covar=tensor([0.1167, 0.1122, 0.1516, 0.0763, 0.0931, 0.2697, 0.1156, 0.0907], device='cuda:3'), in_proj_covar=tensor([0.0524, 0.0657, 0.0787, 0.0670, 0.0497, 0.0513, 0.0520, 0.0605], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:50:35,772 INFO [train.py:904] (3/8) Epoch 12, batch 5150, loss[loss=0.1837, simple_loss=0.285, pruned_loss=0.04125, over 16243.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2812, pruned_loss=0.05204, over 3216751.69 frames. ], batch size: 165, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:50:36,769 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:50:36,813 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:47,787 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:03,103 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:47,912 INFO [train.py:904] (3/8) Epoch 12, batch 5200, loss[loss=0.19, simple_loss=0.2769, pruned_loss=0.05159, over 16655.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.281, pruned_loss=0.05179, over 3205351.94 frames. ], batch size: 57, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:52:00,671 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:52:03,331 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:58,580 INFO [train.py:904] (3/8) Epoch 12, batch 5250, loss[loss=0.1859, simple_loss=0.2743, pruned_loss=0.04874, over 16908.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2791, pruned_loss=0.05198, over 3194344.51 frames. ], batch size: 96, lr: 5.69e-03, grad_scale: 16.0 2023-04-29 14:53:24,787 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5767, 4.5923, 4.4348, 3.7831, 4.4734, 1.6353, 4.1804, 4.1667], device='cuda:3'), covar=tensor([0.0074, 0.0065, 0.0124, 0.0354, 0.0078, 0.2504, 0.0117, 0.0184], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0119, 0.0164, 0.0157, 0.0136, 0.0179, 0.0154, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:54:11,425 INFO [train.py:904] (3/8) Epoch 12, batch 5300, loss[loss=0.1757, simple_loss=0.2567, pruned_loss=0.04731, over 16547.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2758, pruned_loss=0.05098, over 3187448.62 frames. ], batch size: 68, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:54:27,261 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.208e+02 2.588e+02 3.045e+02 5.450e+02, threshold=5.175e+02, percent-clipped=1.0 2023-04-29 14:54:41,985 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:54:49,704 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:55:21,987 INFO [train.py:904] (3/8) Epoch 12, batch 5350, loss[loss=0.1839, simple_loss=0.2809, pruned_loss=0.04338, over 16719.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2739, pruned_loss=0.04993, over 3206259.17 frames. ], batch size: 89, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:55:32,776 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7161, 4.7009, 4.5596, 3.8634, 4.5808, 1.6425, 4.3284, 4.3997], device='cuda:3'), covar=tensor([0.0083, 0.0077, 0.0129, 0.0423, 0.0093, 0.2421, 0.0124, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0119, 0.0163, 0.0157, 0.0136, 0.0178, 0.0153, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:55:45,244 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8018, 1.3310, 1.5921, 1.6849, 1.8070, 1.8864, 1.5624, 1.8074], device='cuda:3'), covar=tensor([0.0185, 0.0306, 0.0151, 0.0231, 0.0207, 0.0152, 0.0286, 0.0086], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0172, 0.0154, 0.0160, 0.0170, 0.0125, 0.0172, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 14:55:58,442 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 5400, loss[loss=0.1844, simple_loss=0.2825, pruned_loss=0.04319, over 16314.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2756, pruned_loss=0.05012, over 3212510.48 frames. ], batch size: 165, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:56:43,918 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:49,066 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 12, batch 5450, loss[loss=0.2045, simple_loss=0.2892, pruned_loss=0.05988, over 17086.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2792, pruned_loss=0.05204, over 3199030.74 frames. ], batch size: 55, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:57:46,713 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:57:54,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1308, 5.0768, 4.9629, 4.7144, 4.5233, 4.9695, 5.0388, 4.6573], device='cuda:3'), covar=tensor([0.0587, 0.0653, 0.0313, 0.0272, 0.1155, 0.0591, 0.0275, 0.0667], device='cuda:3'), in_proj_covar=tensor([0.0250, 0.0330, 0.0300, 0.0278, 0.0319, 0.0324, 0.0203, 0.0348], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-29 14:58:57,078 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:00,358 INFO [train.py:904] (3/8) Epoch 12, batch 5500, loss[loss=0.2417, simple_loss=0.3229, pruned_loss=0.0802, over 15275.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2874, pruned_loss=0.05748, over 3159201.50 frames. ], batch size: 190, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:59:09,313 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:18,173 INFO [optim.py:368] (3/8) 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 14:59:47,074 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8752, 4.1648, 3.8109, 3.6672, 3.2569, 4.0608, 3.7793, 3.6624], device='cuda:3'), covar=tensor([0.0891, 0.0592, 0.0477, 0.0357, 0.1480, 0.0489, 0.0963, 0.0684], device='cuda:3'), in_proj_covar=tensor([0.0248, 0.0327, 0.0297, 0.0276, 0.0317, 0.0321, 0.0200, 0.0344], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:00:18,009 INFO [train.py:904] (3/8) Epoch 12, batch 5550, loss[loss=0.22, simple_loss=0.2995, pruned_loss=0.07028, over 16436.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2956, pruned_loss=0.06396, over 3117300.82 frames. ], batch size: 146, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:00:30,330 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:49,861 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:51,087 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6702, 4.4493, 4.6643, 4.8693, 4.9669, 4.4510, 4.9605, 4.9485], device='cuda:3'), covar=tensor([0.1394, 0.1069, 0.1389, 0.0512, 0.0548, 0.0875, 0.0519, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0525, 0.0661, 0.0793, 0.0669, 0.0503, 0.0519, 0.0524, 0.0608], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:01:39,143 INFO [train.py:904] (3/8) Epoch 12, batch 5600, loss[loss=0.2571, simple_loss=0.3388, pruned_loss=0.08768, over 16804.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3007, pruned_loss=0.0687, over 3082369.34 frames. ], batch size: 90, lr: 5.68e-03, grad_scale: 8.0 2023-04-29 15:01:58,901 INFO [optim.py:368] (3/8) 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 15:02:03,751 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2318, 4.2280, 4.6866, 4.6383, 4.6562, 4.3387, 4.3707, 4.1386], device='cuda:3'), covar=tensor([0.0315, 0.0550, 0.0309, 0.0393, 0.0441, 0.0356, 0.0847, 0.0505], device='cuda:3'), in_proj_covar=tensor([0.0342, 0.0359, 0.0358, 0.0339, 0.0407, 0.0380, 0.0486, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 15:02:16,814 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:02:30,337 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 5650, loss[loss=0.2337, simple_loss=0.3268, pruned_loss=0.07032, over 17212.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3057, pruned_loss=0.07268, over 3069828.41 frames. ], batch size: 44, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:03:10,844 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7626, 1.7959, 1.5647, 1.5555, 1.8780, 1.5843, 1.6571, 1.9009], device='cuda:3'), covar=tensor([0.0127, 0.0177, 0.0270, 0.0252, 0.0126, 0.0170, 0.0129, 0.0134], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0206, 0.0201, 0.0200, 0.0206, 0.0204, 0.0211, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:03:32,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8448, 2.3758, 2.0127, 2.1528, 2.7073, 2.3915, 2.8097, 2.9328], device='cuda:3'), covar=tensor([0.0112, 0.0276, 0.0362, 0.0323, 0.0170, 0.0257, 0.0160, 0.0178], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0206, 0.0201, 0.0200, 0.0205, 0.0203, 0.0210, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:03:33,931 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:45,008 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:53,253 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6229, 3.6426, 2.1347, 4.1903, 2.6819, 4.1206, 2.1756, 2.9032], device='cuda:3'), covar=tensor([0.0225, 0.0395, 0.1612, 0.0145, 0.0758, 0.0466, 0.1637, 0.0757], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0166, 0.0189, 0.0131, 0.0167, 0.0206, 0.0197, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 15:04:08,960 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2310, 4.2218, 4.0984, 3.4848, 4.1443, 1.6965, 3.9748, 3.8288], device='cuda:3'), covar=tensor([0.0089, 0.0070, 0.0125, 0.0291, 0.0071, 0.2446, 0.0100, 0.0167], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0117, 0.0161, 0.0155, 0.0134, 0.0176, 0.0150, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:04:14,571 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1861, 4.0336, 4.2248, 4.4093, 4.4909, 4.0902, 4.4132, 4.4834], device='cuda:3'), covar=tensor([0.1531, 0.1129, 0.1405, 0.0564, 0.0558, 0.1139, 0.0696, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0519, 0.0655, 0.0784, 0.0663, 0.0498, 0.0516, 0.0519, 0.0602], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:04:18,926 INFO [train.py:904] (3/8) Epoch 12, batch 5700, loss[loss=0.2328, simple_loss=0.3155, pruned_loss=0.07504, over 15257.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3081, pruned_loss=0.07536, over 3024210.69 frames. ], batch size: 190, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:04:32,711 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:05:21,402 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:05:39,177 INFO [train.py:904] (3/8) Epoch 12, batch 5750, loss[loss=0.2381, simple_loss=0.3279, pruned_loss=0.07422, over 16634.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3109, pruned_loss=0.07681, over 3009807.77 frames. ], batch size: 134, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:05:48,655 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:06:53,761 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7597, 3.8465, 2.1893, 4.2303, 2.8196, 4.2500, 2.3736, 3.0585], device='cuda:3'), covar=tensor([0.0179, 0.0306, 0.1512, 0.0201, 0.0698, 0.0382, 0.1396, 0.0624], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0165, 0.0189, 0.0131, 0.0166, 0.0205, 0.0196, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 15:07:00,140 INFO [train.py:904] (3/8) Epoch 12, batch 5800, loss[loss=0.2149, simple_loss=0.3034, pruned_loss=0.0632, over 15347.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3108, pruned_loss=0.07567, over 3017165.48 frames. ], batch size: 191, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:07:09,800 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 15:08:16,560 INFO [train.py:904] (3/8) Epoch 12, batch 5850, loss[loss=0.2329, simple_loss=0.3025, pruned_loss=0.08168, over 11372.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3075, pruned_loss=0.07278, over 3043024.73 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:08:24,195 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:09:37,155 INFO [train.py:904] (3/8) Epoch 12, batch 5900, loss[loss=0.2169, simple_loss=0.3066, pruned_loss=0.06358, over 16737.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3064, pruned_loss=0.0716, over 3065508.65 frames. ], batch size: 83, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:10:01,565 INFO [optim.py:368] (3/8) 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 15:10:18,437 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 5950, loss[loss=0.251, simple_loss=0.3248, pruned_loss=0.08855, over 11269.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3078, pruned_loss=0.07015, over 3076325.88 frames. ], batch size: 247, lr: 5.67e-03, grad_scale: 2.0 2023-04-29 15:12:14,104 INFO [train.py:904] (3/8) Epoch 12, batch 6000, loss[loss=0.2223, simple_loss=0.3044, pruned_loss=0.07008, over 16393.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3063, pruned_loss=0.06963, over 3087531.92 frames. ], batch size: 146, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:12:14,104 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 15:12:25,321 INFO [train.py:938] (3/8) Epoch 12, validation: loss=0.161, simple_loss=0.2739, pruned_loss=0.02405, over 944034.00 frames. 2023-04-29 15:12:25,321 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 15:12:31,062 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 15:12:40,177 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-29 15:12:46,505 INFO [optim.py:368] (3/8) 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 15:13:18,108 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 6050, loss[loss=0.2227, simple_loss=0.3095, pruned_loss=0.0679, over 17056.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3049, pruned_loss=0.0695, over 3079523.76 frames. ], batch size: 41, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:14:12,639 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3931, 2.5062, 2.1761, 2.2701, 2.9086, 2.5679, 3.1049, 3.1753], device='cuda:3'), covar=tensor([0.0084, 0.0321, 0.0397, 0.0352, 0.0191, 0.0286, 0.0169, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0202, 0.0199, 0.0197, 0.0202, 0.0200, 0.0208, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:14:34,283 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5603, 4.4049, 4.3857, 3.0720, 3.9995, 4.4773, 4.0825, 2.3142], device='cuda:3'), covar=tensor([0.0446, 0.0031, 0.0039, 0.0291, 0.0061, 0.0086, 0.0056, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0069, 0.0071, 0.0126, 0.0081, 0.0091, 0.0080, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:15:02,175 INFO [train.py:904] (3/8) Epoch 12, batch 6100, loss[loss=0.1834, simple_loss=0.2797, pruned_loss=0.04359, over 16661.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3045, pruned_loss=0.06827, over 3109501.55 frames. ], batch size: 76, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:24,819 INFO [optim.py:368] (3/8) 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 15:16:19,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3086, 2.0996, 2.2608, 3.9532, 2.0177, 2.4992, 2.1927, 2.2362], device='cuda:3'), covar=tensor([0.0988, 0.3177, 0.2213, 0.0412, 0.3781, 0.2245, 0.3115, 0.3134], device='cuda:3'), in_proj_covar=tensor([0.0365, 0.0395, 0.0331, 0.0316, 0.0412, 0.0455, 0.0361, 0.0461], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:16:19,682 INFO [train.py:904] (3/8) Epoch 12, batch 6150, loss[loss=0.2106, simple_loss=0.2931, pruned_loss=0.06402, over 16843.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3017, pruned_loss=0.06708, over 3111347.02 frames. ], batch size: 116, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:17:38,949 INFO [train.py:904] (3/8) Epoch 12, batch 6200, loss[loss=0.2519, simple_loss=0.3122, pruned_loss=0.09575, over 11365.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3004, pruned_loss=0.06681, over 3122797.21 frames. ], batch size: 246, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:18:00,666 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.055e+02 3.626e+02 4.280e+02 7.203e+02, threshold=7.253e+02, percent-clipped=0.0 2023-04-29 15:18:18,065 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:18:21,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5054, 3.5756, 3.3409, 3.1394, 3.1670, 3.4276, 3.2777, 3.2649], device='cuda:3'), covar=tensor([0.0554, 0.0521, 0.0271, 0.0241, 0.0538, 0.0396, 0.1182, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0244, 0.0325, 0.0290, 0.0271, 0.0310, 0.0313, 0.0199, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:18:52,875 INFO [train.py:904] (3/8) Epoch 12, batch 6250, loss[loss=0.211, simple_loss=0.297, pruned_loss=0.06248, over 16616.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3006, pruned_loss=0.06699, over 3114405.11 frames. ], batch size: 62, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:19:27,990 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:20:06,472 INFO [train.py:904] (3/8) Epoch 12, batch 6300, loss[loss=0.2307, simple_loss=0.2943, pruned_loss=0.08352, over 11853.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.3002, pruned_loss=0.0662, over 3124091.73 frames. ], batch size: 248, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:20:28,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5170, 4.5263, 4.9433, 4.9034, 4.9437, 4.5916, 4.5945, 4.3737], device='cuda:3'), covar=tensor([0.0308, 0.0506, 0.0329, 0.0395, 0.0442, 0.0347, 0.0867, 0.0463], device='cuda:3'), in_proj_covar=tensor([0.0347, 0.0360, 0.0361, 0.0344, 0.0413, 0.0384, 0.0489, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 15:20:28,834 INFO [optim.py:368] (3/8) 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 15:20:42,181 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6594, 3.8910, 2.9813, 2.2701, 2.7110, 2.4199, 4.1668, 3.5686], device='cuda:3'), covar=tensor([0.2621, 0.0654, 0.1536, 0.2260, 0.2210, 0.1735, 0.0416, 0.0937], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0258, 0.0287, 0.0283, 0.0282, 0.0223, 0.0268, 0.0299], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:20:57,612 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 6350, loss[loss=0.2185, simple_loss=0.2894, pruned_loss=0.07375, over 16647.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.301, pruned_loss=0.06792, over 3089049.30 frames. ], batch size: 57, lr: 5.66e-03, grad_scale: 4.0 2023-04-29 15:22:11,805 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:14,890 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5387, 4.7102, 4.8670, 4.6668, 4.6868, 5.2324, 4.7036, 4.4972], device='cuda:3'), covar=tensor([0.1151, 0.1620, 0.1881, 0.1952, 0.2551, 0.1077, 0.1635, 0.2445], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0500, 0.0551, 0.0432, 0.0586, 0.0572, 0.0436, 0.0589], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:22:21,806 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:31,091 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7469, 1.7573, 2.2907, 2.6312, 2.5989, 3.0122, 1.8287, 2.9489], device='cuda:3'), covar=tensor([0.0152, 0.0367, 0.0225, 0.0222, 0.0234, 0.0124, 0.0408, 0.0083], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0173, 0.0155, 0.0160, 0.0170, 0.0126, 0.0173, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 15:22:37,169 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-04-29 15:22:37,545 INFO [train.py:904] (3/8) Epoch 12, batch 6400, loss[loss=0.1895, simple_loss=0.2706, pruned_loss=0.05421, over 16475.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3004, pruned_loss=0.06849, over 3079958.53 frames. ], batch size: 68, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:22:57,664 INFO [optim.py:368] (3/8) 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 15:23:49,764 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:23:50,965 INFO [train.py:904] (3/8) Epoch 12, batch 6450, loss[loss=0.2143, simple_loss=0.2945, pruned_loss=0.06705, over 17023.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2995, pruned_loss=0.06751, over 3072507.82 frames. ], batch size: 55, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:08,002 INFO [train.py:904] (3/8) Epoch 12, batch 6500, loss[loss=0.1913, simple_loss=0.2743, pruned_loss=0.05413, over 16974.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2974, pruned_loss=0.06645, over 3102514.81 frames. ], batch size: 41, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:29,347 INFO [optim.py:368] (3/8) 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 15:25:32,620 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:26:07,709 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 15:26:28,542 INFO [train.py:904] (3/8) Epoch 12, batch 6550, loss[loss=0.2243, simple_loss=0.3217, pruned_loss=0.06344, over 16210.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3014, pruned_loss=0.06822, over 3090546.92 frames. ], batch size: 165, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:27:10,773 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 15:27:44,389 INFO [train.py:904] (3/8) Epoch 12, batch 6600, loss[loss=0.237, simple_loss=0.3173, pruned_loss=0.07841, over 16776.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3044, pruned_loss=0.06883, over 3121062.75 frames. ], batch size: 124, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:28:05,473 INFO [optim.py:368] (3/8) 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 15:28:10,262 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:28:41,326 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:01,033 INFO [train.py:904] (3/8) Epoch 12, batch 6650, loss[loss=0.2006, simple_loss=0.2819, pruned_loss=0.0597, over 16893.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3049, pruned_loss=0.06984, over 3108769.94 frames. ], batch size: 116, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:29:43,620 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:45,515 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:47,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 15:29:56,079 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2570, 3.9839, 3.9353, 2.6147, 3.5756, 3.9392, 3.5894, 2.2339], device='cuda:3'), covar=tensor([0.0459, 0.0024, 0.0033, 0.0316, 0.0066, 0.0071, 0.0064, 0.0365], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0069, 0.0071, 0.0126, 0.0080, 0.0092, 0.0081, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:30:16,378 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:18,851 INFO [train.py:904] (3/8) Epoch 12, batch 6700, loss[loss=0.2183, simple_loss=0.2987, pruned_loss=0.06892, over 16871.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3038, pruned_loss=0.07029, over 3082860.03 frames. ], batch size: 116, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:30:24,497 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 15:30:39,917 INFO [optim.py:368] (3/8) 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 15:31:17,451 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:27,183 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 6750, loss[loss=0.2166, simple_loss=0.3008, pruned_loss=0.0662, over 16480.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3029, pruned_loss=0.07019, over 3066961.38 frames. ], batch size: 68, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:31:48,740 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 15:32:33,221 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 15:32:48,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2353, 2.2618, 2.6950, 3.2081, 2.9891, 3.8832, 2.2656, 3.6048], device='cuda:3'), covar=tensor([0.0143, 0.0333, 0.0230, 0.0180, 0.0201, 0.0082, 0.0359, 0.0091], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0155, 0.0158, 0.0170, 0.0126, 0.0171, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 15:32:49,915 INFO [train.py:904] (3/8) Epoch 12, batch 6800, loss[loss=0.239, simple_loss=0.3106, pruned_loss=0.08369, over 11865.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3023, pruned_loss=0.06925, over 3087364.83 frames. ], batch size: 248, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:33:11,648 INFO [optim.py:368] (3/8) 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 15:33:17,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8671, 3.2486, 3.3214, 2.0992, 3.1084, 3.2890, 3.0942, 1.7700], device='cuda:3'), covar=tensor([0.0551, 0.0043, 0.0047, 0.0409, 0.0080, 0.0105, 0.0086, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0069, 0.0071, 0.0127, 0.0080, 0.0092, 0.0081, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:34:04,797 INFO [train.py:904] (3/8) Epoch 12, batch 6850, loss[loss=0.2116, simple_loss=0.3202, pruned_loss=0.05153, over 16691.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3036, pruned_loss=0.069, over 3113396.88 frames. ], batch size: 89, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:34:36,242 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 15:35:04,180 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2346, 1.8808, 2.5228, 3.0669, 2.7375, 3.4870, 1.7685, 3.4380], device='cuda:3'), covar=tensor([0.0134, 0.0412, 0.0271, 0.0188, 0.0246, 0.0109, 0.0484, 0.0091], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0155, 0.0159, 0.0171, 0.0126, 0.0172, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 15:35:15,879 INFO [train.py:904] (3/8) Epoch 12, batch 6900, loss[loss=0.1984, simple_loss=0.2927, pruned_loss=0.05203, over 16890.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3058, pruned_loss=0.06927, over 3110988.73 frames. ], batch size: 96, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:35:36,844 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 3.007e+02 3.634e+02 4.757e+02 1.105e+03, threshold=7.268e+02, percent-clipped=1.0 2023-04-29 15:36:30,544 INFO [train.py:904] (3/8) Epoch 12, batch 6950, loss[loss=0.2232, simple_loss=0.3046, pruned_loss=0.07089, over 16435.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3081, pruned_loss=0.07185, over 3090274.81 frames. ], batch size: 146, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:37:04,588 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:14,739 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:35,443 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 7000, loss[loss=0.2056, simple_loss=0.3077, pruned_loss=0.05173, over 16819.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3073, pruned_loss=0.07053, over 3095892.61 frames. ], batch size: 102, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:37:48,470 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 15:38:05,443 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:44,745 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:50,483 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 7050, loss[loss=0.2396, simple_loss=0.3115, pruned_loss=0.08386, over 11344.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3075, pruned_loss=0.06932, over 3119027.82 frames. ], batch size: 248, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:39:00,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9803, 3.5318, 2.9107, 1.7344, 2.5447, 1.9193, 3.2456, 3.6583], device='cuda:3'), covar=tensor([0.0244, 0.0575, 0.0685, 0.2036, 0.0996, 0.1070, 0.0699, 0.0754], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0144, 0.0158, 0.0143, 0.0136, 0.0124, 0.0136, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 15:39:37,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8331, 5.0197, 5.2676, 5.0834, 5.0793, 5.6755, 5.1475, 4.9283], device='cuda:3'), covar=tensor([0.0912, 0.1812, 0.1708, 0.1793, 0.2407, 0.0918, 0.1401, 0.2260], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0498, 0.0548, 0.0431, 0.0584, 0.0571, 0.0432, 0.0589], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:40:01,768 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:40:14,616 INFO [train.py:904] (3/8) Epoch 12, batch 7100, loss[loss=0.198, simple_loss=0.2807, pruned_loss=0.0577, over 16458.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3059, pruned_loss=0.06899, over 3120734.10 frames. ], batch size: 35, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:36,859 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 3.120e+02 3.747e+02 4.649e+02 1.761e+03, threshold=7.495e+02, percent-clipped=5.0 2023-04-29 15:41:29,295 INFO [train.py:904] (3/8) Epoch 12, batch 7150, loss[loss=0.2162, simple_loss=0.3026, pruned_loss=0.06486, over 16748.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.304, pruned_loss=0.06859, over 3131750.22 frames. ], batch size: 83, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:42:01,967 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 7200, loss[loss=0.1919, simple_loss=0.2718, pruned_loss=0.056, over 11818.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3017, pruned_loss=0.06743, over 3093798.75 frames. ], batch size: 248, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:42:46,649 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9228, 2.6913, 2.7633, 2.0397, 2.5713, 2.1178, 2.6957, 2.8789], device='cuda:3'), covar=tensor([0.0311, 0.0721, 0.0537, 0.1729, 0.0810, 0.0919, 0.0631, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0147, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 15:43:03,912 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:44:00,069 INFO [train.py:904] (3/8) Epoch 12, batch 7250, loss[loss=0.1909, simple_loss=0.2674, pruned_loss=0.05723, over 16705.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2992, pruned_loss=0.06616, over 3091230.08 frames. ], batch size: 62, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:44:35,473 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:04,503 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:15,124 INFO [train.py:904] (3/8) Epoch 12, batch 7300, loss[loss=0.2181, simple_loss=0.3119, pruned_loss=0.06212, over 16180.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2986, pruned_loss=0.0658, over 3090707.10 frames. ], batch size: 165, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:45:36,390 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.034e+02 3.599e+02 4.380e+02 7.583e+02, threshold=7.199e+02, percent-clipped=5.0 2023-04-29 15:45:45,781 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:45,994 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7484, 1.7325, 1.5275, 1.4946, 1.8167, 1.4187, 1.6263, 1.7895], device='cuda:3'), covar=tensor([0.0101, 0.0201, 0.0289, 0.0240, 0.0139, 0.0196, 0.0112, 0.0133], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0203, 0.0198, 0.0198, 0.0204, 0.0202, 0.0206, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:46:03,484 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:06,454 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:28,512 INFO [train.py:904] (3/8) Epoch 12, batch 7350, loss[loss=0.2243, simple_loss=0.3036, pruned_loss=0.07245, over 16242.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2997, pruned_loss=0.06683, over 3083334.73 frames. ], batch size: 165, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:46:38,534 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2312, 3.9793, 3.9371, 2.7152, 3.5452, 3.9722, 3.6322, 1.9495], device='cuda:3'), covar=tensor([0.0430, 0.0025, 0.0027, 0.0284, 0.0071, 0.0072, 0.0064, 0.0381], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0067, 0.0070, 0.0124, 0.0079, 0.0090, 0.0079, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:47:14,410 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:47:34,814 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0726, 3.5685, 3.5218, 2.4755, 3.3137, 3.5461, 3.3714, 1.7986], device='cuda:3'), covar=tensor([0.0452, 0.0033, 0.0037, 0.0296, 0.0069, 0.0094, 0.0070, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0067, 0.0070, 0.0124, 0.0079, 0.0090, 0.0079, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:47:42,938 INFO [train.py:904] (3/8) Epoch 12, batch 7400, loss[loss=0.2164, simple_loss=0.3036, pruned_loss=0.06463, over 16580.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3007, pruned_loss=0.06752, over 3084576.11 frames. ], batch size: 62, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:47:43,734 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 15:48:06,306 INFO [optim.py:368] (3/8) 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 15:48:12,412 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9601, 5.0087, 4.8299, 4.5829, 4.4375, 4.8686, 4.7953, 4.5809], device='cuda:3'), covar=tensor([0.0582, 0.0466, 0.0257, 0.0249, 0.0884, 0.0417, 0.0303, 0.0637], device='cuda:3'), in_proj_covar=tensor([0.0236, 0.0316, 0.0282, 0.0261, 0.0301, 0.0303, 0.0194, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:48:19,127 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:44,377 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5483, 4.3605, 4.3409, 3.1585, 3.8856, 4.3244, 3.9736, 2.2455], device='cuda:3'), covar=tensor([0.0423, 0.0024, 0.0026, 0.0252, 0.0069, 0.0077, 0.0054, 0.0387], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0068, 0.0071, 0.0125, 0.0080, 0.0091, 0.0080, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:48:47,069 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:55,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8526, 1.8301, 2.2974, 2.7811, 2.6760, 3.1450, 1.9560, 3.1414], device='cuda:3'), covar=tensor([0.0141, 0.0358, 0.0246, 0.0215, 0.0210, 0.0116, 0.0369, 0.0090], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0153, 0.0157, 0.0169, 0.0124, 0.0171, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 15:49:01,208 INFO [train.py:904] (3/8) Epoch 12, batch 7450, loss[loss=0.2583, simple_loss=0.322, pruned_loss=0.09728, over 11399.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3026, pruned_loss=0.06917, over 3069551.76 frames. ], batch size: 249, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:49:21,000 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2810, 2.4187, 1.8660, 2.1171, 2.7640, 2.3600, 3.0221, 3.0148], device='cuda:3'), covar=tensor([0.0093, 0.0313, 0.0428, 0.0383, 0.0212, 0.0318, 0.0159, 0.0195], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0202, 0.0198, 0.0198, 0.0203, 0.0201, 0.0206, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:49:55,938 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:50:15,157 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8746, 2.7002, 2.8187, 2.0581, 2.6610, 2.2365, 2.7335, 2.9168], device='cuda:3'), covar=tensor([0.0279, 0.0672, 0.0464, 0.1709, 0.0712, 0.0811, 0.0570, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0145, 0.0158, 0.0143, 0.0136, 0.0125, 0.0137, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 15:50:20,486 INFO [train.py:904] (3/8) Epoch 12, batch 7500, loss[loss=0.2396, simple_loss=0.3037, pruned_loss=0.08775, over 11391.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3031, pruned_loss=0.06905, over 3045742.68 frames. ], batch size: 248, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:50:24,068 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 15:50:28,440 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8000, 1.2906, 1.6355, 1.6888, 1.7901, 1.8881, 1.5420, 1.8131], device='cuda:3'), covar=tensor([0.0184, 0.0296, 0.0140, 0.0189, 0.0190, 0.0127, 0.0267, 0.0090], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0155, 0.0158, 0.0171, 0.0125, 0.0172, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 15:50:42,269 INFO [optim.py:368] (3/8) 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 15:51:25,295 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0263, 4.0262, 3.9309, 3.3230, 3.9293, 1.6734, 3.7239, 3.5836], device='cuda:3'), covar=tensor([0.0094, 0.0075, 0.0139, 0.0298, 0.0086, 0.2441, 0.0121, 0.0200], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0118, 0.0162, 0.0154, 0.0135, 0.0177, 0.0150, 0.0150], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:51:35,616 INFO [train.py:904] (3/8) Epoch 12, batch 7550, loss[loss=0.2072, simple_loss=0.294, pruned_loss=0.06021, over 16721.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3016, pruned_loss=0.06844, over 3061237.21 frames. ], batch size: 76, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:51:55,743 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.61 vs. limit=5.0 2023-04-29 15:52:28,152 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4909, 3.8865, 3.9266, 2.6982, 3.5367, 3.9408, 3.6751, 2.2266], device='cuda:3'), covar=tensor([0.0394, 0.0033, 0.0034, 0.0302, 0.0080, 0.0089, 0.0067, 0.0362], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0068, 0.0071, 0.0125, 0.0080, 0.0091, 0.0080, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 15:52:50,119 INFO [train.py:904] (3/8) Epoch 12, batch 7600, loss[loss=0.2472, simple_loss=0.3071, pruned_loss=0.09361, over 11284.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3013, pruned_loss=0.06877, over 3069625.00 frames. ], batch size: 246, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:53:12,404 INFO [optim.py:368] (3/8) 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 15:53:31,124 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 15:53:43,952 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:54:05,865 INFO [train.py:904] (3/8) Epoch 12, batch 7650, loss[loss=0.2107, simple_loss=0.294, pruned_loss=0.06368, over 16701.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.302, pruned_loss=0.06951, over 3078277.46 frames. ], batch size: 134, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:54:55,612 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:55:07,377 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3442, 5.3679, 5.1707, 4.9069, 4.7833, 5.1977, 5.1684, 4.8903], device='cuda:3'), covar=tensor([0.0548, 0.0356, 0.0249, 0.0229, 0.0914, 0.0364, 0.0212, 0.0560], device='cuda:3'), in_proj_covar=tensor([0.0237, 0.0320, 0.0282, 0.0263, 0.0301, 0.0305, 0.0196, 0.0331], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 15:55:20,165 INFO [train.py:904] (3/8) Epoch 12, batch 7700, loss[loss=0.2131, simple_loss=0.2939, pruned_loss=0.06615, over 16369.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3031, pruned_loss=0.07096, over 3054599.00 frames. ], batch size: 146, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:55:42,613 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.234e+02 3.396e+02 4.418e+02 5.495e+02 1.012e+03, threshold=8.835e+02, percent-clipped=5.0 2023-04-29 15:56:32,029 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6012, 2.6730, 2.4345, 4.2319, 3.0921, 4.0200, 1.3379, 3.0262], device='cuda:3'), covar=tensor([0.1353, 0.0716, 0.1172, 0.0143, 0.0302, 0.0420, 0.1598, 0.0734], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0162, 0.0182, 0.0151, 0.0199, 0.0208, 0.0184, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 15:56:36,113 INFO [train.py:904] (3/8) Epoch 12, batch 7750, loss[loss=0.215, simple_loss=0.3016, pruned_loss=0.06425, over 16381.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3021, pruned_loss=0.06979, over 3066360.39 frames. ], batch size: 165, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:57:18,831 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:57:45,521 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 15:57:48,704 INFO [train.py:904] (3/8) Epoch 12, batch 7800, loss[loss=0.2147, simple_loss=0.3012, pruned_loss=0.06411, over 16442.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3036, pruned_loss=0.07064, over 3062767.40 frames. ], batch size: 146, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:58:11,184 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.141e+02 3.879e+02 4.510e+02 7.221e+02, threshold=7.757e+02, percent-clipped=0.0 2023-04-29 15:59:04,882 INFO [train.py:904] (3/8) Epoch 12, batch 7850, loss[loss=0.2124, simple_loss=0.3008, pruned_loss=0.06202, over 16772.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3036, pruned_loss=0.06992, over 3054631.94 frames. ], batch size: 124, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 15:59:24,429 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:59:50,641 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-04-29 16:00:18,022 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8402, 1.7467, 1.5055, 1.4740, 1.8392, 1.5597, 1.6446, 1.9290], device='cuda:3'), covar=tensor([0.0151, 0.0230, 0.0347, 0.0292, 0.0167, 0.0225, 0.0148, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0207, 0.0203, 0.0202, 0.0208, 0.0205, 0.0211, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:00:21,535 INFO [train.py:904] (3/8) Epoch 12, batch 7900, loss[loss=0.2057, simple_loss=0.2976, pruned_loss=0.05692, over 16663.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3023, pruned_loss=0.06911, over 3046063.30 frames. ], batch size: 89, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:00:23,368 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 16:00:45,732 INFO [optim.py:368] (3/8) 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 16:00:56,354 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:01:38,576 INFO [train.py:904] (3/8) Epoch 12, batch 7950, loss[loss=0.2231, simple_loss=0.3012, pruned_loss=0.07252, over 15507.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3018, pruned_loss=0.06879, over 3074640.52 frames. ], batch size: 192, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:02:39,412 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9805, 2.6646, 2.5841, 1.9324, 2.5055, 2.6831, 2.6169, 1.9113], device='cuda:3'), covar=tensor([0.0334, 0.0055, 0.0050, 0.0292, 0.0091, 0.0085, 0.0078, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0069, 0.0071, 0.0127, 0.0080, 0.0092, 0.0081, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 16:02:53,352 INFO [train.py:904] (3/8) Epoch 12, batch 8000, loss[loss=0.2247, simple_loss=0.3036, pruned_loss=0.07287, over 15383.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.3031, pruned_loss=0.06997, over 3057982.47 frames. ], batch size: 191, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 16:02:58,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1307, 4.2143, 3.9811, 3.7814, 3.7025, 4.1065, 3.8539, 3.7855], device='cuda:3'), covar=tensor([0.0641, 0.0536, 0.0302, 0.0281, 0.0821, 0.0440, 0.0629, 0.0608], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0324, 0.0285, 0.0264, 0.0303, 0.0307, 0.0197, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:03:12,785 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:03:17,120 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 12, batch 8050, loss[loss=0.2262, simple_loss=0.3141, pruned_loss=0.0691, over 15283.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3023, pruned_loss=0.06912, over 3071520.34 frames. ], batch size: 190, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:04:26,747 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0538, 3.9693, 4.1202, 4.2741, 4.3627, 3.9954, 4.3234, 4.3515], device='cuda:3'), covar=tensor([0.1534, 0.0994, 0.1352, 0.0580, 0.0574, 0.1134, 0.0673, 0.0641], device='cuda:3'), in_proj_covar=tensor([0.0523, 0.0655, 0.0785, 0.0665, 0.0505, 0.0513, 0.0535, 0.0607], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:04:42,917 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:05:17,376 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 16:05:21,186 INFO [train.py:904] (3/8) Epoch 12, batch 8100, loss[loss=0.2057, simple_loss=0.2932, pruned_loss=0.05916, over 16419.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3016, pruned_loss=0.06868, over 3076012.30 frames. ], batch size: 75, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:05:47,745 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.274e+02 3.055e+02 3.528e+02 4.290e+02 7.598e+02, threshold=7.056e+02, percent-clipped=1.0 2023-04-29 16:05:48,726 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 16:05:52,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5625, 5.5868, 5.3272, 4.6880, 5.3925, 2.2244, 5.1750, 5.1773], device='cuda:3'), covar=tensor([0.0057, 0.0039, 0.0126, 0.0306, 0.0065, 0.2045, 0.0082, 0.0145], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0119, 0.0165, 0.0156, 0.0137, 0.0181, 0.0152, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:06:00,177 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:28,060 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:34,773 INFO [train.py:904] (3/8) Epoch 12, batch 8150, loss[loss=0.1879, simple_loss=0.2691, pruned_loss=0.05331, over 15328.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2989, pruned_loss=0.0675, over 3088369.35 frames. ], batch size: 191, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:06:57,707 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5535, 4.8347, 4.5883, 4.5806, 4.3531, 4.2317, 4.3181, 4.8762], device='cuda:3'), covar=tensor([0.1011, 0.0752, 0.0920, 0.0682, 0.0732, 0.1209, 0.1019, 0.0845], device='cuda:3'), in_proj_covar=tensor([0.0544, 0.0672, 0.0559, 0.0478, 0.0429, 0.0447, 0.0566, 0.0523], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:07:50,608 INFO [train.py:904] (3/8) Epoch 12, batch 8200, loss[loss=0.1987, simple_loss=0.2774, pruned_loss=0.06002, over 16628.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2968, pruned_loss=0.06682, over 3095885.10 frames. ], batch size: 57, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:08:18,228 INFO [optim.py:368] (3/8) 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 16:08:19,311 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:08:37,006 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 16:08:38,908 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 16:09:09,078 INFO [train.py:904] (3/8) Epoch 12, batch 8250, loss[loss=0.1789, simple_loss=0.2663, pruned_loss=0.04577, over 11947.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2962, pruned_loss=0.06474, over 3075725.34 frames. ], batch size: 247, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:28,023 INFO [train.py:904] (3/8) Epoch 12, batch 8300, loss[loss=0.1877, simple_loss=0.2819, pruned_loss=0.04668, over 16747.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2933, pruned_loss=0.06164, over 3073617.29 frames. ], batch size: 124, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:57,564 INFO [optim.py:368] (3/8) 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 16:11:52,966 INFO [train.py:904] (3/8) Epoch 12, batch 8350, loss[loss=0.196, simple_loss=0.2912, pruned_loss=0.0504, over 16861.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2922, pruned_loss=0.06005, over 3050722.82 frames. ], batch size: 116, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:12:24,537 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=120021.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:13:14,450 INFO [train.py:904] (3/8) Epoch 12, batch 8400, loss[loss=0.1713, simple_loss=0.2721, pruned_loss=0.0353, over 16821.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2895, pruned_loss=0.05784, over 3038474.36 frames. ], batch size: 102, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:13:42,958 INFO [optim.py:368] (3/8) 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 16:14:31,500 INFO [train.py:904] (3/8) Epoch 12, batch 8450, loss[loss=0.1831, simple_loss=0.2766, pruned_loss=0.04484, over 15274.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2875, pruned_loss=0.05584, over 3049644.87 frames. ], batch size: 191, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:15:47,446 INFO [train.py:904] (3/8) Epoch 12, batch 8500, loss[loss=0.1683, simple_loss=0.2477, pruned_loss=0.04444, over 11961.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2831, pruned_loss=0.05304, over 3040305.08 frames. ], batch size: 246, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:16:15,430 INFO [optim.py:368] (3/8) 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 16:16:15,917 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:16:34,975 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-04-29 16:17:07,725 INFO [train.py:904] (3/8) Epoch 12, batch 8550, loss[loss=0.2185, simple_loss=0.308, pruned_loss=0.06448, over 16659.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2809, pruned_loss=0.05182, over 3036148.06 frames. ], batch size: 134, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:17:20,380 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 16:17:37,325 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:17:50,562 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2448, 4.2009, 4.0976, 3.4896, 4.1406, 1.6518, 3.9607, 3.7378], device='cuda:3'), covar=tensor([0.0068, 0.0068, 0.0124, 0.0213, 0.0075, 0.2371, 0.0096, 0.0196], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0116, 0.0159, 0.0150, 0.0133, 0.0177, 0.0148, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:18:33,696 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4023, 3.0265, 2.6846, 2.1286, 2.1827, 2.1276, 2.9806, 2.8555], device='cuda:3'), covar=tensor([0.2398, 0.0706, 0.1413, 0.2271, 0.2384, 0.1997, 0.0401, 0.1124], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0249, 0.0278, 0.0275, 0.0271, 0.0218, 0.0260, 0.0288], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:18:41,889 INFO [train.py:904] (3/8) Epoch 12, batch 8600, loss[loss=0.1929, simple_loss=0.2872, pruned_loss=0.0493, over 16812.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2808, pruned_loss=0.05063, over 3026868.23 frames. ], batch size: 124, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:19:19,444 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.485e+02 2.557e+02 3.073e+02 4.030e+02 6.956e+02, threshold=6.147e+02, percent-clipped=5.0 2023-04-29 16:20:19,200 INFO [train.py:904] (3/8) Epoch 12, batch 8650, loss[loss=0.1757, simple_loss=0.2627, pruned_loss=0.04433, over 12318.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2793, pruned_loss=0.04932, over 3025104.91 frames. ], batch size: 248, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:21:01,235 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:21:53,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0536, 3.0900, 1.8836, 3.3303, 2.2243, 3.2822, 2.0453, 2.5910], device='cuda:3'), covar=tensor([0.0269, 0.0342, 0.1544, 0.0185, 0.0843, 0.0503, 0.1580, 0.0722], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0158, 0.0184, 0.0126, 0.0163, 0.0198, 0.0194, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 16:22:02,145 INFO [train.py:904] (3/8) Epoch 12, batch 8700, loss[loss=0.1898, simple_loss=0.2741, pruned_loss=0.05278, over 12512.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2761, pruned_loss=0.04764, over 3038951.07 frames. ], batch size: 248, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:22:33,633 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:23:09,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6663, 3.2245, 3.3628, 1.8351, 2.8662, 2.2479, 3.1735, 3.3574], device='cuda:3'), covar=tensor([0.0295, 0.0691, 0.0491, 0.1957, 0.0769, 0.0961, 0.0707, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0141, 0.0155, 0.0141, 0.0134, 0.0123, 0.0134, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 16:23:36,530 INFO [train.py:904] (3/8) Epoch 12, batch 8750, loss[loss=0.2036, simple_loss=0.2996, pruned_loss=0.05375, over 15268.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.276, pruned_loss=0.04662, over 3060664.45 frames. ], batch size: 191, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:23:40,735 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5379, 3.5365, 3.4844, 2.9213, 3.4124, 1.9370, 3.2200, 2.8619], device='cuda:3'), covar=tensor([0.0098, 0.0083, 0.0123, 0.0170, 0.0080, 0.2117, 0.0097, 0.0179], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0115, 0.0158, 0.0149, 0.0132, 0.0177, 0.0148, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:23:47,396 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4953, 4.3309, 4.5522, 4.6890, 4.8305, 4.3879, 4.8665, 4.8559], device='cuda:3'), covar=tensor([0.1454, 0.1120, 0.1371, 0.0649, 0.0511, 0.0766, 0.0430, 0.0553], device='cuda:3'), in_proj_covar=tensor([0.0506, 0.0636, 0.0762, 0.0646, 0.0490, 0.0501, 0.0517, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:25:30,724 INFO [train.py:904] (3/8) Epoch 12, batch 8800, loss[loss=0.1696, simple_loss=0.2661, pruned_loss=0.03653, over 16850.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2743, pruned_loss=0.0452, over 3071532.49 frames. ], batch size: 96, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:26:08,427 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.538e+02 3.135e+02 3.629e+02 6.620e+02, threshold=6.271e+02, percent-clipped=2.0 2023-04-29 16:27:17,125 INFO [train.py:904] (3/8) Epoch 12, batch 8850, loss[loss=0.1948, simple_loss=0.2917, pruned_loss=0.04894, over 16955.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2779, pruned_loss=0.04508, over 3078865.35 frames. ], batch size: 125, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:27:31,960 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 16:29:04,467 INFO [train.py:904] (3/8) Epoch 12, batch 8900, loss[loss=0.1772, simple_loss=0.2779, pruned_loss=0.03827, over 16763.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2781, pruned_loss=0.04483, over 3077889.53 frames. ], batch size: 83, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:29:39,305 INFO [optim.py:368] (3/8) 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 16:31:10,994 INFO [train.py:904] (3/8) Epoch 12, batch 8950, loss[loss=0.1673, simple_loss=0.2621, pruned_loss=0.03624, over 16463.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2781, pruned_loss=0.04525, over 3100921.80 frames. ], batch size: 68, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,365 INFO [train.py:904] (3/8) Epoch 12, batch 9000, loss[loss=0.1484, simple_loss=0.2378, pruned_loss=0.02949, over 16771.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2744, pruned_loss=0.0438, over 3074621.33 frames. ], batch size: 83, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,365 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 16:33:10,349 INFO [train.py:938] (3/8) Epoch 12, validation: loss=0.1532, simple_loss=0.2571, pruned_loss=0.02465, over 944034.00 frames. 2023-04-29 16:33:10,350 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 16:33:49,110 INFO [optim.py:368] (3/8) 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 16:34:01,233 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 12, batch 9050, loss[loss=0.1495, simple_loss=0.2431, pruned_loss=0.02799, over 16565.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2746, pruned_loss=0.04424, over 3069243.03 frames. ], batch size: 62, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:35:27,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9918, 3.3789, 3.5193, 2.3313, 3.1629, 3.4773, 3.3714, 1.9336], device='cuda:3'), covar=tensor([0.0494, 0.0034, 0.0035, 0.0319, 0.0075, 0.0066, 0.0054, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0067, 0.0068, 0.0124, 0.0079, 0.0088, 0.0079, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 16:35:43,806 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8338, 4.8341, 4.6846, 4.2710, 4.3264, 4.7344, 4.6941, 4.3866], device='cuda:3'), covar=tensor([0.0538, 0.0478, 0.0274, 0.0300, 0.0890, 0.0415, 0.0303, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0312, 0.0275, 0.0256, 0.0291, 0.0296, 0.0191, 0.0320], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:36:04,419 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:36:36,157 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1995, 5.5800, 5.3259, 5.3127, 5.0062, 4.9391, 4.9384, 5.6574], device='cuda:3'), covar=tensor([0.1159, 0.0902, 0.0898, 0.0656, 0.0714, 0.0682, 0.1077, 0.0784], device='cuda:3'), in_proj_covar=tensor([0.0528, 0.0654, 0.0539, 0.0465, 0.0419, 0.0433, 0.0546, 0.0510], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:36:39,406 INFO [train.py:904] (3/8) Epoch 12, batch 9100, loss[loss=0.1661, simple_loss=0.2687, pruned_loss=0.0318, over 16664.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2743, pruned_loss=0.0448, over 3071975.46 frames. ], batch size: 89, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:37:15,444 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.513e+02 2.910e+02 3.503e+02 5.095e+02, threshold=5.819e+02, percent-clipped=0.0 2023-04-29 16:37:27,119 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3002, 3.5431, 3.6828, 2.4952, 3.2973, 3.6122, 3.4800, 2.1209], device='cuda:3'), covar=tensor([0.0413, 0.0037, 0.0031, 0.0304, 0.0083, 0.0066, 0.0053, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0125, 0.0066, 0.0067, 0.0122, 0.0078, 0.0086, 0.0077, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 16:38:36,952 INFO [train.py:904] (3/8) Epoch 12, batch 9150, loss[loss=0.1842, simple_loss=0.2669, pruned_loss=0.05072, over 12160.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2749, pruned_loss=0.04508, over 3042926.83 frames. ], batch size: 247, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:21,514 INFO [train.py:904] (3/8) Epoch 12, batch 9200, loss[loss=0.1869, simple_loss=0.2754, pruned_loss=0.04915, over 16184.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2696, pruned_loss=0.04349, over 3065809.54 frames. ], batch size: 165, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:55,538 INFO [optim.py:368] (3/8) 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 16:41:00,572 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 16:41:03,852 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8400, 2.9389, 2.4434, 4.5941, 2.8823, 4.0840, 1.4529, 3.0285], device='cuda:3'), covar=tensor([0.1409, 0.0757, 0.1316, 0.0136, 0.0190, 0.0380, 0.1763, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0158, 0.0179, 0.0143, 0.0187, 0.0202, 0.0181, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 16:41:14,940 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-04-29 16:42:00,503 INFO [train.py:904] (3/8) Epoch 12, batch 9250, loss[loss=0.1853, simple_loss=0.2743, pruned_loss=0.0482, over 15336.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2692, pruned_loss=0.04338, over 3065706.25 frames. ], batch size: 191, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:42:15,619 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9018, 4.2239, 3.9952, 4.0507, 3.6409, 3.7855, 3.8467, 4.2059], device='cuda:3'), covar=tensor([0.1028, 0.1016, 0.1055, 0.0746, 0.0892, 0.1637, 0.0916, 0.1075], device='cuda:3'), in_proj_covar=tensor([0.0526, 0.0655, 0.0538, 0.0463, 0.0418, 0.0435, 0.0545, 0.0510], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:42:24,098 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 16:43:49,104 INFO [train.py:904] (3/8) Epoch 12, batch 9300, loss[loss=0.1668, simple_loss=0.2481, pruned_loss=0.04277, over 12313.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.268, pruned_loss=0.04305, over 3063141.15 frames. ], batch size: 246, lr: 5.60e-03, grad_scale: 4.0 2023-04-29 16:44:07,077 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3512, 2.0730, 2.1212, 3.9208, 2.0519, 2.4873, 2.2167, 2.2643], device='cuda:3'), covar=tensor([0.0918, 0.3382, 0.2524, 0.0389, 0.3836, 0.2252, 0.3173, 0.3147], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0384, 0.0326, 0.0309, 0.0406, 0.0437, 0.0352, 0.0446], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:44:31,798 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 12, batch 9350, loss[loss=0.205, simple_loss=0.2897, pruned_loss=0.06013, over 16966.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2681, pruned_loss=0.04298, over 3063974.18 frames. ], batch size: 109, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:45:35,668 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:13,259 INFO [train.py:904] (3/8) Epoch 12, batch 9400, loss[loss=0.1859, simple_loss=0.2861, pruned_loss=0.04282, over 16063.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2686, pruned_loss=0.04245, over 3078187.26 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:47:26,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1672, 2.0671, 2.2042, 3.5552, 1.9837, 2.3603, 2.2243, 2.1457], device='cuda:3'), covar=tensor([0.0948, 0.3230, 0.2327, 0.0449, 0.4017, 0.2110, 0.2920, 0.3389], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0384, 0.0326, 0.0310, 0.0406, 0.0437, 0.0351, 0.0446], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:47:39,097 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121064.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:47:50,521 INFO [optim.py:368] (3/8) 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 16:48:04,419 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8337, 4.8011, 4.6166, 4.1670, 4.6259, 1.6679, 4.4227, 4.4172], device='cuda:3'), covar=tensor([0.0053, 0.0049, 0.0119, 0.0199, 0.0068, 0.2287, 0.0080, 0.0155], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0115, 0.0157, 0.0146, 0.0132, 0.0178, 0.0147, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 16:48:55,033 INFO [train.py:904] (3/8) Epoch 12, batch 9450, loss[loss=0.1662, simple_loss=0.2632, pruned_loss=0.03462, over 16796.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2699, pruned_loss=0.04278, over 3064144.48 frames. ], batch size: 83, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:49:39,726 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3230, 1.5643, 1.9730, 2.4727, 2.3724, 2.6472, 1.6810, 2.5933], device='cuda:3'), covar=tensor([0.0188, 0.0417, 0.0279, 0.0227, 0.0252, 0.0135, 0.0406, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0170, 0.0156, 0.0157, 0.0167, 0.0121, 0.0170, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 16:49:47,112 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4066, 2.9075, 3.1352, 1.8307, 2.7939, 2.1952, 3.0333, 2.9996], device='cuda:3'), covar=tensor([0.0296, 0.0816, 0.0521, 0.1960, 0.0773, 0.0938, 0.0702, 0.0976], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0138, 0.0155, 0.0141, 0.0133, 0.0123, 0.0133, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 16:50:34,751 INFO [train.py:904] (3/8) Epoch 12, batch 9500, loss[loss=0.1665, simple_loss=0.253, pruned_loss=0.04002, over 12831.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2687, pruned_loss=0.04234, over 3059585.56 frames. ], batch size: 246, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:51:01,842 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:51:13,616 INFO [optim.py:368] (3/8) 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 16:52:19,995 INFO [train.py:904] (3/8) Epoch 12, batch 9550, loss[loss=0.1639, simple_loss=0.2559, pruned_loss=0.0359, over 12563.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2679, pruned_loss=0.04221, over 3044511.38 frames. ], batch size: 246, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:53:08,750 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:53:59,926 INFO [train.py:904] (3/8) Epoch 12, batch 9600, loss[loss=0.1926, simple_loss=0.2895, pruned_loss=0.04785, over 16210.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2694, pruned_loss=0.04318, over 3029147.46 frames. ], batch size: 165, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:54:35,186 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.441e+02 3.037e+02 3.482e+02 7.490e+02, threshold=6.075e+02, percent-clipped=2.0 2023-04-29 16:55:12,438 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7686, 5.0115, 5.1186, 4.9053, 4.8999, 5.5262, 5.0472, 4.7786], device='cuda:3'), covar=tensor([0.0820, 0.1728, 0.1552, 0.2033, 0.2619, 0.0888, 0.1318, 0.2201], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0470, 0.0524, 0.0406, 0.0541, 0.0545, 0.0410, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 16:55:45,648 INFO [train.py:904] (3/8) Epoch 12, batch 9650, loss[loss=0.1688, simple_loss=0.268, pruned_loss=0.03479, over 16736.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2711, pruned_loss=0.04323, over 3034859.32 frames. ], batch size: 76, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:56:09,947 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 16:56:51,968 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:28,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4155, 3.2447, 3.3896, 1.7318, 3.5561, 3.6117, 2.8592, 2.8143], device='cuda:3'), covar=tensor([0.0669, 0.0206, 0.0174, 0.1151, 0.0064, 0.0124, 0.0386, 0.0412], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0098, 0.0084, 0.0135, 0.0067, 0.0102, 0.0119, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 16:57:30,748 INFO [train.py:904] (3/8) Epoch 12, batch 9700, loss[loss=0.1748, simple_loss=0.2578, pruned_loss=0.04587, over 12147.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2702, pruned_loss=0.04336, over 3021058.97 frames. ], batch size: 250, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:57:44,330 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:58:07,289 INFO [optim.py:368] (3/8) 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 16:58:10,422 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 16:58:31,089 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:58:33,105 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 16:59:04,708 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5594, 3.6240, 3.3482, 3.0762, 3.2399, 3.5185, 3.2981, 3.3675], device='cuda:3'), covar=tensor([0.0531, 0.0462, 0.0259, 0.0233, 0.0551, 0.0336, 0.1270, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0304, 0.0273, 0.0252, 0.0287, 0.0293, 0.0188, 0.0317], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-29 16:59:14,333 INFO [train.py:904] (3/8) Epoch 12, batch 9750, loss[loss=0.1851, simple_loss=0.2636, pruned_loss=0.05332, over 12400.00 frames. ], tot_loss[loss=0.178, simple_loss=0.269, pruned_loss=0.04347, over 3009906.39 frames. ], batch size: 249, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:59:23,931 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:59:52,984 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 16:59:58,296 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 17:00:39,136 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1440, 1.4916, 1.7798, 2.1112, 2.1944, 2.2589, 1.7120, 2.2364], device='cuda:3'), covar=tensor([0.0208, 0.0365, 0.0229, 0.0230, 0.0230, 0.0162, 0.0340, 0.0113], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0168, 0.0154, 0.0156, 0.0166, 0.0120, 0.0167, 0.0113], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 17:00:53,745 INFO [train.py:904] (3/8) Epoch 12, batch 9800, loss[loss=0.1893, simple_loss=0.2878, pruned_loss=0.04538, over 16987.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2691, pruned_loss=0.04274, over 3025671.71 frames. ], batch size: 109, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:01:24,641 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:01:29,898 INFO [optim.py:368] (3/8) 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 17:02:38,487 INFO [train.py:904] (3/8) Epoch 12, batch 9850, loss[loss=0.184, simple_loss=0.2758, pruned_loss=0.04612, over 16970.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2701, pruned_loss=0.04231, over 3037283.41 frames. ], batch size: 109, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:03:17,655 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:03:23,774 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6074, 4.4862, 4.7006, 4.8422, 5.0025, 4.4288, 4.9813, 4.9932], device='cuda:3'), covar=tensor([0.1509, 0.1050, 0.1327, 0.0620, 0.0481, 0.0826, 0.0492, 0.0543], device='cuda:3'), in_proj_covar=tensor([0.0500, 0.0625, 0.0750, 0.0640, 0.0482, 0.0495, 0.0508, 0.0581], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:04:30,193 INFO [train.py:904] (3/8) Epoch 12, batch 9900, loss[loss=0.1959, simple_loss=0.2893, pruned_loss=0.05131, over 15267.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2697, pruned_loss=0.04188, over 3007604.54 frames. ], batch size: 190, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:05:12,985 INFO [optim.py:368] (3/8) 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 17:06:15,896 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 17:06:26,646 INFO [train.py:904] (3/8) Epoch 12, batch 9950, loss[loss=0.1973, simple_loss=0.2844, pruned_loss=0.0551, over 11926.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2724, pruned_loss=0.04242, over 3027607.79 frames. ], batch size: 248, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:27,932 INFO [train.py:904] (3/8) Epoch 12, batch 10000, loss[loss=0.1864, simple_loss=0.2704, pruned_loss=0.05123, over 12980.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2707, pruned_loss=0.04181, over 3043798.45 frames. ], batch size: 250, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:44,243 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:08:50,133 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2740, 4.1160, 4.3612, 4.4670, 4.6152, 4.1225, 4.5753, 4.5991], device='cuda:3'), covar=tensor([0.1398, 0.1029, 0.1234, 0.0627, 0.0466, 0.0958, 0.0541, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0497, 0.0625, 0.0748, 0.0638, 0.0480, 0.0491, 0.0504, 0.0578], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:08:56,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9817, 4.2292, 4.0689, 4.0663, 3.7039, 3.7997, 3.8238, 4.2004], device='cuda:3'), covar=tensor([0.0824, 0.0809, 0.0839, 0.0581, 0.0685, 0.1462, 0.0792, 0.0830], device='cuda:3'), in_proj_covar=tensor([0.0517, 0.0651, 0.0531, 0.0459, 0.0415, 0.0428, 0.0538, 0.0499], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:08:59,915 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:09:06,353 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 12, batch 10050, loss[loss=0.2016, simple_loss=0.2937, pruned_loss=0.05471, over 16686.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2711, pruned_loss=0.04181, over 3049867.97 frames. ], batch size: 134, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:10:21,423 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:11:46,088 INFO [train.py:904] (3/8) Epoch 12, batch 10100, loss[loss=0.1612, simple_loss=0.2512, pruned_loss=0.03566, over 16773.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2715, pruned_loss=0.04206, over 3058338.03 frames. ], batch size: 83, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:12:06,075 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 13, batch 0, loss[loss=0.203, simple_loss=0.2914, pruned_loss=0.05729, over 17061.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2914, pruned_loss=0.05729, over 17061.00 frames. ], batch size: 55, lr: 5.36e-03, grad_scale: 8.0 2023-04-29 17:13:30,552 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 17:13:38,110 INFO [train.py:938] (3/8) Epoch 13, validation: loss=0.1523, simple_loss=0.2559, pruned_loss=0.02431, over 944034.00 frames. 2023-04-29 17:13:38,111 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 17:14:03,985 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:14:49,499 INFO [train.py:904] (3/8) Epoch 13, batch 50, loss[loss=0.2, simple_loss=0.2751, pruned_loss=0.0625, over 16298.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2828, pruned_loss=0.06239, over 754603.91 frames. ], batch size: 164, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:15:00,334 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 17:15:11,431 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:15:22,693 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8164, 3.0439, 2.5567, 4.4615, 3.6012, 4.1858, 1.5199, 2.9167], device='cuda:3'), covar=tensor([0.1359, 0.0587, 0.1120, 0.0117, 0.0176, 0.0381, 0.1583, 0.0781], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0156, 0.0178, 0.0143, 0.0184, 0.0202, 0.0180, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 17:15:22,754 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4658, 2.1562, 2.3413, 4.1931, 2.2513, 2.5856, 2.2431, 2.3505], device='cuda:3'), covar=tensor([0.1032, 0.3606, 0.2333, 0.0469, 0.3565, 0.2325, 0.3251, 0.2762], device='cuda:3'), in_proj_covar=tensor([0.0357, 0.0386, 0.0327, 0.0310, 0.0405, 0.0439, 0.0353, 0.0448], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:15:28,718 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9142, 1.8993, 2.2749, 2.6807, 2.7176, 2.5933, 1.8788, 2.9958], device='cuda:3'), covar=tensor([0.0140, 0.0342, 0.0269, 0.0233, 0.0231, 0.0252, 0.0385, 0.0099], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0156, 0.0159, 0.0169, 0.0122, 0.0171, 0.0115], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:3') 2023-04-29 17:15:58,270 INFO [train.py:904] (3/8) Epoch 13, batch 100, loss[loss=0.2083, simple_loss=0.2875, pruned_loss=0.06454, over 16720.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2758, pruned_loss=0.0569, over 1327278.52 frames. ], batch size: 83, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:16:26,390 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 17:16:30,910 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7800, 3.0713, 2.7289, 4.8963, 4.0367, 4.3358, 1.5755, 3.3329], device='cuda:3'), covar=tensor([0.1299, 0.0640, 0.1111, 0.0145, 0.0227, 0.0404, 0.1508, 0.0640], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0144, 0.0185, 0.0203, 0.0181, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 17:17:04,486 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:17:07,072 INFO [train.py:904] (3/8) Epoch 13, batch 150, loss[loss=0.1614, simple_loss=0.2536, pruned_loss=0.0346, over 17217.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2742, pruned_loss=0.05558, over 1763063.35 frames. ], batch size: 44, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:17:27,824 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:17:35,629 INFO [optim.py:368] (3/8) 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 17:18:00,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8126, 5.2008, 4.9495, 4.8986, 4.6989, 4.6732, 4.6366, 5.3027], device='cuda:3'), covar=tensor([0.1083, 0.0857, 0.1095, 0.0766, 0.0789, 0.0889, 0.1026, 0.0852], device='cuda:3'), in_proj_covar=tensor([0.0544, 0.0684, 0.0558, 0.0482, 0.0437, 0.0446, 0.0570, 0.0522], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:18:18,196 INFO [train.py:904] (3/8) Epoch 13, batch 200, loss[loss=0.1451, simple_loss=0.2376, pruned_loss=0.02637, over 16862.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2749, pruned_loss=0.05578, over 2104478.28 frames. ], batch size: 42, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:18:30,495 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:18:35,254 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 17:18:42,726 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-29 17:19:26,236 INFO [train.py:904] (3/8) Epoch 13, batch 250, loss[loss=0.1822, simple_loss=0.2537, pruned_loss=0.05535, over 16828.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2726, pruned_loss=0.05462, over 2373819.73 frames. ], batch size: 83, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:19:41,324 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:19:54,203 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 300, loss[loss=0.1508, simple_loss=0.2326, pruned_loss=0.0345, over 16783.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.269, pruned_loss=0.05308, over 2591429.13 frames. ], batch size: 39, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:20:47,418 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:20:55,801 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:21:42,721 INFO [train.py:904] (3/8) Epoch 13, batch 350, loss[loss=0.15, simple_loss=0.2335, pruned_loss=0.03324, over 16805.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2664, pruned_loss=0.05129, over 2758279.93 frames. ], batch size: 42, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:22:13,944 INFO [optim.py:368] (3/8) 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 17:22:20,587 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 17:22:43,846 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:22:53,137 INFO [train.py:904] (3/8) Epoch 13, batch 400, loss[loss=0.1675, simple_loss=0.2616, pruned_loss=0.03667, over 17091.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2654, pruned_loss=0.0514, over 2888603.92 frames. ], batch size: 47, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:03,760 INFO [train.py:904] (3/8) Epoch 13, batch 450, loss[loss=0.1702, simple_loss=0.2519, pruned_loss=0.04427, over 16685.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2638, pruned_loss=0.05015, over 2991647.94 frames. ], batch size: 62, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:09,623 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:24:29,359 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9681, 1.9116, 2.4091, 2.9112, 2.6771, 2.9509, 1.9356, 3.0595], device='cuda:3'), covar=tensor([0.0145, 0.0358, 0.0248, 0.0221, 0.0226, 0.0158, 0.0374, 0.0101], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0161, 0.0164, 0.0174, 0.0127, 0.0175, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 17:24:32,943 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.250e+02 2.134e+02 2.674e+02 3.325e+02 9.122e+02, threshold=5.349e+02, percent-clipped=2.0 2023-04-29 17:25:13,886 INFO [train.py:904] (3/8) Epoch 13, batch 500, loss[loss=0.1944, simple_loss=0.2683, pruned_loss=0.06021, over 16866.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2628, pruned_loss=0.04931, over 3063105.38 frames. ], batch size: 116, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:25:19,457 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:25:36,888 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:25:59,321 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9334, 4.8657, 4.7193, 4.3054, 4.3303, 4.7764, 4.6727, 4.4601], device='cuda:3'), covar=tensor([0.0584, 0.0590, 0.0312, 0.0324, 0.0990, 0.0441, 0.0377, 0.0692], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0338, 0.0301, 0.0280, 0.0319, 0.0320, 0.0204, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:26:07,848 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0814, 4.2130, 4.5428, 2.1440, 4.7316, 4.8027, 3.3149, 3.7527], device='cuda:3'), covar=tensor([0.0705, 0.0168, 0.0150, 0.1142, 0.0060, 0.0120, 0.0397, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0102, 0.0089, 0.0139, 0.0069, 0.0108, 0.0122, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 17:26:21,438 INFO [train.py:904] (3/8) Epoch 13, batch 550, loss[loss=0.1756, simple_loss=0.2472, pruned_loss=0.05201, over 16815.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2617, pruned_loss=0.04892, over 3116672.48 frames. ], batch size: 83, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:26:50,249 INFO [optim.py:368] (3/8) 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 17:27:00,776 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:27:22,529 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0187, 1.9532, 2.4662, 2.9292, 2.7847, 2.8559, 2.0622, 3.1144], device='cuda:3'), covar=tensor([0.0145, 0.0343, 0.0232, 0.0197, 0.0215, 0.0176, 0.0329, 0.0100], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0160, 0.0163, 0.0174, 0.0128, 0.0174, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 17:27:30,038 INFO [train.py:904] (3/8) Epoch 13, batch 600, loss[loss=0.1509, simple_loss=0.2363, pruned_loss=0.03272, over 16791.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2612, pruned_loss=0.04931, over 3163473.56 frames. ], batch size: 39, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:27:39,199 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7977, 3.8236, 1.9694, 4.3910, 2.7164, 4.3216, 2.2390, 3.0084], device='cuda:3'), covar=tensor([0.0239, 0.0381, 0.1752, 0.0261, 0.0858, 0.0484, 0.1536, 0.0673], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0164, 0.0188, 0.0136, 0.0170, 0.0205, 0.0197, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 17:27:49,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6900, 2.9267, 2.8612, 4.9499, 4.1191, 4.5139, 1.7210, 3.1229], device='cuda:3'), covar=tensor([0.1327, 0.0677, 0.1034, 0.0192, 0.0254, 0.0384, 0.1420, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0157, 0.0179, 0.0147, 0.0189, 0.0204, 0.0180, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 17:28:17,231 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-04-29 17:28:37,794 INFO [train.py:904] (3/8) Epoch 13, batch 650, loss[loss=0.1644, simple_loss=0.2496, pruned_loss=0.03966, over 16468.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2592, pruned_loss=0.04829, over 3190791.26 frames. ], batch size: 146, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:29:01,196 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:29:07,724 INFO [optim.py:368] (3/8) 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 17:29:08,033 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:29:47,551 INFO [train.py:904] (3/8) Epoch 13, batch 700, loss[loss=0.1651, simple_loss=0.2486, pruned_loss=0.04079, over 16632.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2603, pruned_loss=0.0484, over 3222115.81 frames. ], batch size: 76, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:30:26,078 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:56,883 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:57,588 INFO [train.py:904] (3/8) Epoch 13, batch 750, loss[loss=0.1723, simple_loss=0.2507, pruned_loss=0.04697, over 16763.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2594, pruned_loss=0.04801, over 3240390.44 frames. ], batch size: 83, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:31:00,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3164, 3.4798, 3.6489, 3.6378, 3.6456, 3.4556, 3.4885, 3.4791], device='cuda:3'), covar=tensor([0.0425, 0.0775, 0.0568, 0.0481, 0.0546, 0.0511, 0.0777, 0.0591], device='cuda:3'), in_proj_covar=tensor([0.0354, 0.0374, 0.0371, 0.0353, 0.0420, 0.0396, 0.0493, 0.0320], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 17:31:27,840 INFO [optim.py:368] (3/8) 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 17:32:02,991 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4960, 4.4118, 4.3703, 4.1205, 4.0861, 4.4419, 4.2066, 4.1826], device='cuda:3'), covar=tensor([0.0573, 0.0612, 0.0281, 0.0253, 0.0773, 0.0405, 0.0476, 0.0577], device='cuda:3'), in_proj_covar=tensor([0.0257, 0.0345, 0.0307, 0.0287, 0.0327, 0.0328, 0.0208, 0.0358], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:32:09,552 INFO [train.py:904] (3/8) Epoch 13, batch 800, loss[loss=0.1555, simple_loss=0.2443, pruned_loss=0.03335, over 17029.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2591, pruned_loss=0.04817, over 3254278.48 frames. ], batch size: 50, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:32:15,154 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 850, loss[loss=0.1583, simple_loss=0.2362, pruned_loss=0.04025, over 16884.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.259, pruned_loss=0.04774, over 3274127.92 frames. ], batch size: 90, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:33:18,234 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:23,654 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4872, 3.6551, 2.2165, 3.8642, 2.8155, 3.7999, 2.1636, 2.7727], device='cuda:3'), covar=tensor([0.0248, 0.0329, 0.1372, 0.0285, 0.0666, 0.0684, 0.1327, 0.0670], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0164, 0.0188, 0.0136, 0.0168, 0.0206, 0.0196, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 17:33:44,224 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:34:23,508 INFO [train.py:904] (3/8) Epoch 13, batch 900, loss[loss=0.1694, simple_loss=0.2477, pruned_loss=0.04553, over 16811.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2591, pruned_loss=0.0473, over 3284594.64 frames. ], batch size: 102, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:35:04,748 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1080, 1.7976, 2.6196, 3.0371, 2.7998, 3.5406, 1.8238, 3.3711], device='cuda:3'), covar=tensor([0.0153, 0.0424, 0.0227, 0.0204, 0.0230, 0.0106, 0.0477, 0.0109], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0174, 0.0159, 0.0163, 0.0174, 0.0127, 0.0175, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 17:35:33,155 INFO [train.py:904] (3/8) Epoch 13, batch 950, loss[loss=0.1642, simple_loss=0.2574, pruned_loss=0.03549, over 16868.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2591, pruned_loss=0.04747, over 3285970.44 frames. ], batch size: 42, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:36:02,667 INFO [optim.py:368] (3/8) 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 17:36:03,705 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:36:43,312 INFO [train.py:904] (3/8) Epoch 13, batch 1000, loss[loss=0.1475, simple_loss=0.2334, pruned_loss=0.03077, over 16993.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2584, pruned_loss=0.0478, over 3289931.41 frames. ], batch size: 41, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:36:47,270 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 17:37:09,707 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:14,619 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,069 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,802 INFO [train.py:904] (3/8) Epoch 13, batch 1050, loss[loss=0.148, simple_loss=0.2425, pruned_loss=0.02676, over 17220.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2576, pruned_loss=0.04655, over 3299799.34 frames. ], batch size: 46, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:38:20,430 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:38:21,176 INFO [optim.py:368] (3/8) 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 17:38:58,265 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 1100, loss[loss=0.18, simple_loss=0.2495, pruned_loss=0.05526, over 16692.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.257, pruned_loss=0.04615, over 3306654.76 frames. ], batch size: 134, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:39:32,568 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2038, 4.1799, 4.1923, 3.1749, 4.1764, 1.5564, 3.8678, 3.6943], device='cuda:3'), covar=tensor([0.0172, 0.0144, 0.0178, 0.0605, 0.0132, 0.3213, 0.0198, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0127, 0.0172, 0.0161, 0.0145, 0.0189, 0.0161, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:39:45,774 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 1150, loss[loss=0.1539, simple_loss=0.2424, pruned_loss=0.03274, over 16992.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2563, pruned_loss=0.04558, over 3302938.63 frames. ], batch size: 41, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:40:39,513 INFO [optim.py:368] (3/8) 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 17:40:43,437 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:41:20,386 INFO [train.py:904] (3/8) Epoch 13, batch 1200, loss[loss=0.1865, simple_loss=0.2534, pruned_loss=0.05976, over 16693.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2557, pruned_loss=0.04556, over 3303849.56 frames. ], batch size: 83, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:41:50,190 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 1250, loss[loss=0.1551, simple_loss=0.2394, pruned_loss=0.03536, over 15946.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2565, pruned_loss=0.04593, over 3296656.09 frames. ], batch size: 35, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:42:59,815 INFO [optim.py:368] (3/8) 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 17:43:32,695 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0471, 4.5344, 3.4566, 2.4436, 2.9740, 2.7079, 4.8261, 3.9290], device='cuda:3'), covar=tensor([0.2375, 0.0550, 0.1345, 0.2495, 0.2569, 0.1781, 0.0336, 0.1079], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0261, 0.0288, 0.0284, 0.0279, 0.0227, 0.0272, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:43:40,478 INFO [train.py:904] (3/8) Epoch 13, batch 1300, loss[loss=0.1705, simple_loss=0.2631, pruned_loss=0.03893, over 17080.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2568, pruned_loss=0.04626, over 3304531.66 frames. ], batch size: 55, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:44:06,151 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9942, 4.1990, 2.3732, 4.6456, 3.0142, 4.7145, 2.6034, 3.2503], device='cuda:3'), covar=tensor([0.0233, 0.0286, 0.1558, 0.0249, 0.0805, 0.0369, 0.1415, 0.0716], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0168, 0.0191, 0.0141, 0.0171, 0.0211, 0.0200, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 17:44:12,323 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 1350, loss[loss=0.165, simple_loss=0.2431, pruned_loss=0.04341, over 12598.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2577, pruned_loss=0.04661, over 3307706.33 frames. ], batch size: 246, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:45:17,070 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:45:17,976 INFO [optim.py:368] (3/8) 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 17:45:58,587 INFO [train.py:904] (3/8) Epoch 13, batch 1400, loss[loss=0.145, simple_loss=0.2298, pruned_loss=0.03016, over 16769.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2578, pruned_loss=0.04614, over 3313595.46 frames. ], batch size: 39, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:46:04,264 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 17:46:23,592 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3087, 5.2623, 5.0986, 4.5089, 5.1375, 1.9428, 4.8803, 5.0744], device='cuda:3'), covar=tensor([0.0070, 0.0062, 0.0143, 0.0366, 0.0097, 0.2346, 0.0125, 0.0164], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0128, 0.0173, 0.0162, 0.0146, 0.0189, 0.0162, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:46:35,857 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:47:05,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4393, 5.8554, 5.6483, 5.6975, 5.2065, 5.2373, 5.3324, 6.0060], device='cuda:3'), covar=tensor([0.1266, 0.1056, 0.0979, 0.0739, 0.0787, 0.0685, 0.1030, 0.0961], device='cuda:3'), in_proj_covar=tensor([0.0575, 0.0730, 0.0588, 0.0516, 0.0459, 0.0468, 0.0608, 0.0558], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:47:09,272 INFO [train.py:904] (3/8) Epoch 13, batch 1450, loss[loss=0.1904, simple_loss=0.2539, pruned_loss=0.06347, over 16900.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2565, pruned_loss=0.0455, over 3315845.77 frames. ], batch size: 109, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:47:35,863 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5321, 5.9423, 5.6389, 5.7136, 5.3061, 5.2061, 5.3195, 6.0748], device='cuda:3'), covar=tensor([0.1295, 0.0910, 0.1206, 0.0767, 0.0773, 0.0708, 0.1074, 0.0918], device='cuda:3'), in_proj_covar=tensor([0.0575, 0.0730, 0.0588, 0.0516, 0.0459, 0.0469, 0.0608, 0.0559], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:47:38,908 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.301e+02 2.596e+02 3.249e+02 6.793e+02, threshold=5.192e+02, percent-clipped=2.0 2023-04-29 17:47:40,336 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5984, 4.4843, 4.4570, 4.2252, 4.2180, 4.5105, 4.3153, 4.2840], device='cuda:3'), covar=tensor([0.0535, 0.0573, 0.0273, 0.0264, 0.0791, 0.0414, 0.0547, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0264, 0.0357, 0.0318, 0.0296, 0.0338, 0.0339, 0.0213, 0.0370], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:48:19,760 INFO [train.py:904] (3/8) Epoch 13, batch 1500, loss[loss=0.2138, simple_loss=0.2781, pruned_loss=0.07471, over 16761.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2567, pruned_loss=0.04612, over 3316948.02 frames. ], batch size: 134, lr: 5.33e-03, grad_scale: 4.0 2023-04-29 17:48:37,193 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:49:20,513 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7254, 2.8224, 2.4229, 2.6385, 3.0633, 2.9748, 3.5220, 3.2601], device='cuda:3'), covar=tensor([0.0094, 0.0292, 0.0351, 0.0320, 0.0219, 0.0264, 0.0189, 0.0217], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0217, 0.0208, 0.0208, 0.0217, 0.0215, 0.0222, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:49:30,722 INFO [train.py:904] (3/8) Epoch 13, batch 1550, loss[loss=0.1715, simple_loss=0.2584, pruned_loss=0.04233, over 17233.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2595, pruned_loss=0.04765, over 3320523.56 frames. ], batch size: 44, lr: 5.32e-03, grad_scale: 4.0 2023-04-29 17:49:31,207 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7590, 3.1971, 2.9212, 4.9482, 3.9835, 4.5065, 1.8660, 3.3637], device='cuda:3'), covar=tensor([0.1404, 0.0657, 0.1059, 0.0181, 0.0246, 0.0354, 0.1441, 0.0683], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0160, 0.0182, 0.0153, 0.0194, 0.0209, 0.0182, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 17:50:00,254 INFO [optim.py:368] (3/8) 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 17:50:01,820 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:50:11,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6872, 4.2543, 4.4104, 3.1959, 3.6659, 4.2609, 3.9547, 2.4806], device='cuda:3'), covar=tensor([0.0414, 0.0041, 0.0026, 0.0261, 0.0088, 0.0081, 0.0064, 0.0368], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0072, 0.0072, 0.0129, 0.0084, 0.0093, 0.0083, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 17:50:27,765 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4549, 4.3923, 4.5789, 4.4423, 4.4448, 4.9884, 4.5635, 4.3208], device='cuda:3'), covar=tensor([0.1500, 0.2002, 0.2113, 0.2031, 0.2894, 0.1089, 0.1478, 0.2425], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0526, 0.0574, 0.0455, 0.0616, 0.0597, 0.0454, 0.0606], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 17:50:39,392 INFO [train.py:904] (3/8) Epoch 13, batch 1600, loss[loss=0.1906, simple_loss=0.2833, pruned_loss=0.04896, over 17063.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2604, pruned_loss=0.04804, over 3327917.32 frames. ], batch size: 53, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:50:52,323 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9832, 2.4388, 2.6319, 1.8711, 2.7363, 2.7286, 2.3569, 2.3889], device='cuda:3'), covar=tensor([0.0723, 0.0219, 0.0223, 0.0906, 0.0090, 0.0216, 0.0465, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0101, 0.0090, 0.0138, 0.0069, 0.0110, 0.0121, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 17:50:59,391 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-29 17:51:15,064 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 1650, loss[loss=0.1509, simple_loss=0.2334, pruned_loss=0.03418, over 17013.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2612, pruned_loss=0.04805, over 3320517.03 frames. ], batch size: 41, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:52:11,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7528, 2.6867, 2.2101, 2.4778, 3.0077, 2.8328, 3.4649, 3.1727], device='cuda:3'), covar=tensor([0.0098, 0.0317, 0.0420, 0.0353, 0.0227, 0.0291, 0.0196, 0.0235], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0217, 0.0209, 0.0208, 0.0218, 0.0216, 0.0223, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:52:18,026 INFO [optim.py:368] (3/8) 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 17:52:38,837 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:52:55,811 INFO [train.py:904] (3/8) Epoch 13, batch 1700, loss[loss=0.1879, simple_loss=0.2705, pruned_loss=0.05268, over 15954.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.263, pruned_loss=0.049, over 3314039.74 frames. ], batch size: 35, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:53:12,653 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7353, 4.1783, 3.9920, 2.1853, 3.2473, 2.5286, 4.1083, 4.3080], device='cuda:3'), covar=tensor([0.0240, 0.0632, 0.0541, 0.1969, 0.0898, 0.1027, 0.0552, 0.0887], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0149, 0.0159, 0.0144, 0.0137, 0.0125, 0.0137, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 17:53:32,020 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:53:43,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7341, 5.0860, 4.7917, 4.8519, 4.5714, 4.4587, 4.4685, 5.1071], device='cuda:3'), covar=tensor([0.1063, 0.0796, 0.1055, 0.0723, 0.0823, 0.1255, 0.1113, 0.0949], device='cuda:3'), in_proj_covar=tensor([0.0585, 0.0737, 0.0595, 0.0524, 0.0467, 0.0477, 0.0613, 0.0565], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:54:04,798 INFO [train.py:904] (3/8) Epoch 13, batch 1750, loss[loss=0.2011, simple_loss=0.2759, pruned_loss=0.0632, over 16416.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2637, pruned_loss=0.04942, over 3315776.32 frames. ], batch size: 146, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:54:16,097 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5580, 3.4236, 2.7553, 2.0841, 2.3109, 2.1318, 3.4849, 3.1323], device='cuda:3'), covar=tensor([0.2449, 0.0725, 0.1554, 0.2515, 0.2438, 0.1963, 0.0548, 0.1342], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0262, 0.0289, 0.0283, 0.0280, 0.0228, 0.0272, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:54:34,130 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:55:14,401 INFO [train.py:904] (3/8) Epoch 13, batch 1800, loss[loss=0.1441, simple_loss=0.2222, pruned_loss=0.03302, over 17023.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2636, pruned_loss=0.04836, over 3322581.85 frames. ], batch size: 41, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:55:16,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5034, 5.4184, 5.2790, 4.7670, 5.3273, 2.4301, 5.0379, 5.3223], device='cuda:3'), covar=tensor([0.0057, 0.0067, 0.0140, 0.0298, 0.0072, 0.2017, 0.0112, 0.0113], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0131, 0.0176, 0.0165, 0.0149, 0.0191, 0.0165, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:56:23,364 INFO [train.py:904] (3/8) Epoch 13, batch 1850, loss[loss=0.2007, simple_loss=0.2839, pruned_loss=0.05879, over 16464.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.265, pruned_loss=0.04879, over 3315524.12 frames. ], batch size: 146, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:37,132 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-29 17:56:40,450 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5155, 2.2100, 2.4241, 4.2925, 2.2475, 2.6897, 2.2952, 2.4430], device='cuda:3'), covar=tensor([0.1033, 0.3475, 0.2384, 0.0501, 0.3574, 0.2241, 0.3417, 0.2725], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0402, 0.0339, 0.0326, 0.0417, 0.0464, 0.0368, 0.0471], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 17:56:47,435 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:56:52,511 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 1900, loss[loss=0.1433, simple_loss=0.2249, pruned_loss=0.0309, over 16705.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2641, pruned_loss=0.04823, over 3314677.36 frames. ], batch size: 39, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:39,331 INFO [train.py:904] (3/8) Epoch 13, batch 1950, loss[loss=0.1986, simple_loss=0.2713, pruned_loss=0.06289, over 16793.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.265, pruned_loss=0.04812, over 3316020.76 frames. ], batch size: 102, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:59:09,529 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 17:59:09,937 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.247e+02 2.704e+02 3.281e+02 7.395e+02, threshold=5.408e+02, percent-clipped=2.0 2023-04-29 17:59:25,260 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 2000, loss[loss=0.1957, simple_loss=0.2883, pruned_loss=0.05162, over 17271.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2648, pruned_loss=0.04799, over 3317725.06 frames. ], batch size: 52, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:00:59,626 INFO [train.py:904] (3/8) Epoch 13, batch 2050, loss[loss=0.2078, simple_loss=0.2839, pruned_loss=0.06589, over 16261.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2657, pruned_loss=0.04888, over 3309474.93 frames. ], batch size: 165, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:01:28,819 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 13, batch 2100, loss[loss=0.1513, simple_loss=0.2428, pruned_loss=0.02986, over 17231.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2657, pruned_loss=0.04887, over 3315928.04 frames. ], batch size: 45, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:02:33,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0625, 2.0276, 2.4678, 2.8857, 2.7777, 3.3741, 2.3136, 3.3340], device='cuda:3'), covar=tensor([0.0176, 0.0371, 0.0263, 0.0236, 0.0245, 0.0136, 0.0336, 0.0117], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0176, 0.0161, 0.0166, 0.0175, 0.0129, 0.0175, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 18:02:45,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8556, 5.1792, 5.3440, 5.0885, 5.0812, 5.7329, 5.1967, 4.9397], device='cuda:3'), covar=tensor([0.1125, 0.1719, 0.2120, 0.1881, 0.2782, 0.0982, 0.1386, 0.2315], device='cuda:3'), in_proj_covar=tensor([0.0375, 0.0537, 0.0583, 0.0466, 0.0624, 0.0609, 0.0462, 0.0615], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:02:54,425 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:09,966 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1537, 5.6582, 5.7928, 5.5260, 5.5765, 6.1246, 5.6289, 5.4486], device='cuda:3'), covar=tensor([0.0765, 0.1677, 0.2019, 0.1963, 0.2697, 0.0990, 0.1386, 0.2028], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0535, 0.0582, 0.0465, 0.0623, 0.0609, 0.0461, 0.0614], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:03:20,325 INFO [train.py:904] (3/8) Epoch 13, batch 2150, loss[loss=0.1861, simple_loss=0.281, pruned_loss=0.04554, over 17141.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2675, pruned_loss=0.05023, over 3308422.41 frames. ], batch size: 48, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:03:45,044 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:50,026 INFO [optim.py:368] (3/8) 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 18:04:30,751 INFO [train.py:904] (3/8) Epoch 13, batch 2200, loss[loss=0.1641, simple_loss=0.2432, pruned_loss=0.04246, over 16805.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2676, pruned_loss=0.05008, over 3315680.95 frames. ], batch size: 39, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:04:53,398 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:19,549 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:40,213 INFO [train.py:904] (3/8) Epoch 13, batch 2250, loss[loss=0.2117, simple_loss=0.278, pruned_loss=0.07275, over 16898.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2688, pruned_loss=0.05142, over 3310370.04 frames. ], batch size: 109, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:05:53,619 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:09,553 INFO [optim.py:368] (3/8) 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 18:06:22,744 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:42,865 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:48,018 INFO [train.py:904] (3/8) Epoch 13, batch 2300, loss[loss=0.2032, simple_loss=0.2799, pruned_loss=0.06324, over 15456.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2694, pruned_loss=0.05167, over 3297722.51 frames. ], batch size: 190, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:06:48,914 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 18:07:03,563 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-04-29 18:07:17,834 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:29,437 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 2350, loss[loss=0.1691, simple_loss=0.2593, pruned_loss=0.03948, over 16865.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2682, pruned_loss=0.0506, over 3312580.51 frames. ], batch size: 42, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:08:26,417 INFO [optim.py:368] (3/8) 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 18:09:06,167 INFO [train.py:904] (3/8) Epoch 13, batch 2400, loss[loss=0.1913, simple_loss=0.2824, pruned_loss=0.05008, over 17076.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2684, pruned_loss=0.05015, over 3325208.42 frames. ], batch size: 53, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:09:42,798 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:10:15,680 INFO [train.py:904] (3/8) Epoch 13, batch 2450, loss[loss=0.194, simple_loss=0.2694, pruned_loss=0.05927, over 16787.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2698, pruned_loss=0.05029, over 3331979.11 frames. ], batch size: 83, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:10:21,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4946, 4.3150, 4.5468, 4.6984, 4.8042, 4.3262, 4.6438, 4.7816], device='cuda:3'), covar=tensor([0.1498, 0.1373, 0.1330, 0.0810, 0.0607, 0.1062, 0.1636, 0.1018], device='cuda:3'), in_proj_covar=tensor([0.0572, 0.0714, 0.0863, 0.0722, 0.0543, 0.0572, 0.0577, 0.0664], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:10:22,482 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 18:10:37,289 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 18:10:43,289 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.400e+02 2.819e+02 3.354e+02 7.582e+02, threshold=5.638e+02, percent-clipped=2.0 2023-04-29 18:10:50,918 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1608, 4.9085, 5.1516, 5.3688, 5.5420, 4.7751, 5.5171, 5.5091], device='cuda:3'), covar=tensor([0.1514, 0.1217, 0.1598, 0.0627, 0.0423, 0.0687, 0.0440, 0.0526], device='cuda:3'), in_proj_covar=tensor([0.0570, 0.0711, 0.0858, 0.0717, 0.0540, 0.0569, 0.0574, 0.0660], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:11:24,236 INFO [train.py:904] (3/8) Epoch 13, batch 2500, loss[loss=0.1767, simple_loss=0.2597, pruned_loss=0.04691, over 16573.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2691, pruned_loss=0.05008, over 3327404.95 frames. ], batch size: 75, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:11:30,479 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4342, 3.2840, 3.7238, 1.8668, 3.7970, 3.8198, 3.0773, 2.8145], device='cuda:3'), covar=tensor([0.0677, 0.0225, 0.0135, 0.1040, 0.0075, 0.0159, 0.0336, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0103, 0.0090, 0.0140, 0.0071, 0.0112, 0.0122, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 18:12:09,475 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:12:31,503 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:12:37,094 INFO [train.py:904] (3/8) Epoch 13, batch 2550, loss[loss=0.205, simple_loss=0.2735, pruned_loss=0.06829, over 16723.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2691, pruned_loss=0.05039, over 3329529.22 frames. ], batch size: 134, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:01,680 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6537, 3.7846, 2.9166, 2.2097, 2.5209, 2.2765, 3.7871, 3.3505], device='cuda:3'), covar=tensor([0.2402, 0.0627, 0.1521, 0.2612, 0.2399, 0.1828, 0.0535, 0.1248], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0262, 0.0289, 0.0285, 0.0284, 0.0229, 0.0272, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:13:06,415 INFO [optim.py:368] (3/8) 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 18:13:31,386 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 2600, loss[loss=0.1711, simple_loss=0.2638, pruned_loss=0.0392, over 17123.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2694, pruned_loss=0.05007, over 3329421.01 frames. ], batch size: 47, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:48,150 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 18:13:54,829 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:14:06,294 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124418.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:26,635 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:48,653 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0634, 4.0862, 4.4709, 4.4721, 4.4992, 4.1613, 4.2117, 4.1003], device='cuda:3'), covar=tensor([0.0350, 0.0568, 0.0396, 0.0394, 0.0440, 0.0378, 0.0777, 0.0481], device='cuda:3'), in_proj_covar=tensor([0.0363, 0.0382, 0.0382, 0.0361, 0.0432, 0.0406, 0.0504, 0.0324], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 18:14:54,847 INFO [train.py:904] (3/8) Epoch 13, batch 2650, loss[loss=0.1821, simple_loss=0.2768, pruned_loss=0.04374, over 16672.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2692, pruned_loss=0.04914, over 3336268.05 frames. ], batch size: 62, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:15:26,000 INFO [optim.py:368] (3/8) 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 18:15:52,743 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:04,557 INFO [train.py:904] (3/8) Epoch 13, batch 2700, loss[loss=0.1786, simple_loss=0.2696, pruned_loss=0.0438, over 16461.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2694, pruned_loss=0.04867, over 3329875.46 frames. ], batch size: 68, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:16:11,794 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 18:16:40,265 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:50,915 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9208, 4.1102, 3.0817, 2.2869, 2.8694, 2.4715, 4.4778, 3.6514], device='cuda:3'), covar=tensor([0.2433, 0.0841, 0.1664, 0.2432, 0.2285, 0.1762, 0.0450, 0.1109], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0262, 0.0289, 0.0285, 0.0283, 0.0229, 0.0273, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:17:13,323 INFO [train.py:904] (3/8) Epoch 13, batch 2750, loss[loss=0.1945, simple_loss=0.2742, pruned_loss=0.05742, over 16884.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2687, pruned_loss=0.04781, over 3331206.65 frames. ], batch size: 109, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:17:44,015 INFO [optim.py:368] (3/8) 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 18:17:46,591 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:18:10,574 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 18:18:22,960 INFO [train.py:904] (3/8) Epoch 13, batch 2800, loss[loss=0.16, simple_loss=0.2416, pruned_loss=0.03917, over 16789.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2684, pruned_loss=0.04763, over 3332532.09 frames. ], batch size: 39, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:18:24,548 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9202, 1.8348, 2.3638, 2.7465, 2.6854, 3.2682, 2.0778, 3.2171], device='cuda:3'), covar=tensor([0.0192, 0.0439, 0.0287, 0.0290, 0.0268, 0.0144, 0.0401, 0.0119], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0182, 0.0165, 0.0171, 0.0180, 0.0133, 0.0179, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:18:36,389 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 18:18:59,117 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:12,817 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 2850, loss[loss=0.1848, simple_loss=0.2725, pruned_loss=0.04853, over 17084.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2679, pruned_loss=0.04735, over 3334102.12 frames. ], batch size: 53, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:19:46,097 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 18:20:00,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5553, 2.5727, 2.1652, 2.4623, 2.8547, 2.7443, 3.3483, 3.1696], device='cuda:3'), covar=tensor([0.0099, 0.0334, 0.0394, 0.0338, 0.0223, 0.0285, 0.0194, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0217, 0.0208, 0.0209, 0.0218, 0.0215, 0.0226, 0.0206], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:20:24,494 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:33,936 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 2900, loss[loss=0.1632, simple_loss=0.252, pruned_loss=0.03723, over 16781.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2674, pruned_loss=0.04843, over 3330694.12 frames. ], batch size: 39, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:20:40,109 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 18:20:58,985 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:07,466 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:08,678 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4679, 2.9053, 2.9264, 2.0480, 2.6080, 2.0955, 3.1660, 3.1185], device='cuda:3'), covar=tensor([0.0261, 0.0813, 0.0583, 0.1680, 0.0827, 0.1010, 0.0587, 0.0773], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0151, 0.0161, 0.0146, 0.0138, 0.0126, 0.0138, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 18:21:29,838 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:33,212 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7289, 2.5080, 2.4395, 3.4966, 2.7151, 3.6137, 1.4924, 2.7327], device='cuda:3'), covar=tensor([0.1365, 0.0654, 0.1082, 0.0201, 0.0163, 0.0424, 0.1534, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0164, 0.0186, 0.0160, 0.0200, 0.0212, 0.0186, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 18:21:34,622 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 18:21:37,309 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4116, 4.2908, 4.4631, 4.6387, 4.7275, 4.2880, 4.5630, 4.7453], device='cuda:3'), covar=tensor([0.1494, 0.1073, 0.1241, 0.0636, 0.0656, 0.1181, 0.1532, 0.0691], device='cuda:3'), in_proj_covar=tensor([0.0572, 0.0716, 0.0867, 0.0727, 0.0543, 0.0575, 0.0577, 0.0664], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:21:46,235 INFO [train.py:904] (3/8) Epoch 13, batch 2950, loss[loss=0.1636, simple_loss=0.2612, pruned_loss=0.03298, over 17204.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2668, pruned_loss=0.04907, over 3321000.56 frames. ], batch size: 44, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:05,795 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:16,910 INFO [optim.py:368] (3/8) 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 18:22:35,692 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:39,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4715, 5.4083, 5.2636, 4.7737, 4.8442, 5.2716, 5.3170, 4.8921], device='cuda:3'), covar=tensor([0.0562, 0.0398, 0.0256, 0.0307, 0.1156, 0.0427, 0.0210, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0366, 0.0324, 0.0305, 0.0348, 0.0350, 0.0221, 0.0380], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:22:48,684 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6489, 3.6038, 3.9252, 2.0661, 4.0010, 3.9892, 3.2363, 3.0076], device='cuda:3'), covar=tensor([0.0681, 0.0189, 0.0127, 0.1036, 0.0055, 0.0157, 0.0289, 0.0414], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0101, 0.0090, 0.0138, 0.0070, 0.0112, 0.0122, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 18:22:55,199 INFO [train.py:904] (3/8) Epoch 13, batch 3000, loss[loss=0.1773, simple_loss=0.2639, pruned_loss=0.04533, over 16468.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2669, pruned_loss=0.04953, over 3319738.09 frames. ], batch size: 75, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:55,199 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 18:23:04,002 INFO [train.py:938] (3/8) Epoch 13, validation: loss=0.1391, simple_loss=0.2452, pruned_loss=0.01648, over 944034.00 frames. 2023-04-29 18:23:04,003 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 18:23:08,403 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3322, 2.5429, 2.0098, 2.3664, 2.9182, 2.6668, 3.1981, 3.1078], device='cuda:3'), covar=tensor([0.0123, 0.0283, 0.0426, 0.0334, 0.0185, 0.0316, 0.0190, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0218, 0.0208, 0.0210, 0.0219, 0.0216, 0.0227, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:23:29,500 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 3050, loss[loss=0.1749, simple_loss=0.2678, pruned_loss=0.04097, over 16752.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2668, pruned_loss=0.04983, over 3312666.56 frames. ], batch size: 57, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:24:44,775 INFO [optim.py:368] (3/8) 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 18:24:55,344 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:25:09,260 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9099, 4.3794, 3.3105, 2.3107, 2.9825, 2.6147, 4.7479, 3.9317], device='cuda:3'), covar=tensor([0.2492, 0.0577, 0.1438, 0.2438, 0.2433, 0.1737, 0.0324, 0.1015], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0259, 0.0286, 0.0283, 0.0282, 0.0228, 0.0271, 0.0307], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:25:21,447 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9707, 5.4068, 5.6163, 5.3743, 5.5059, 6.0506, 5.5006, 5.2555], device='cuda:3'), covar=tensor([0.0901, 0.1924, 0.1888, 0.1982, 0.2184, 0.0803, 0.1366, 0.2185], device='cuda:3'), in_proj_covar=tensor([0.0375, 0.0529, 0.0576, 0.0460, 0.0622, 0.0602, 0.0458, 0.0611], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:25:25,280 INFO [train.py:904] (3/8) Epoch 13, batch 3100, loss[loss=0.1669, simple_loss=0.2503, pruned_loss=0.04176, over 16851.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2662, pruned_loss=0.04948, over 3323419.77 frames. ], batch size: 42, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:26:01,344 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:26:34,832 INFO [train.py:904] (3/8) Epoch 13, batch 3150, loss[loss=0.182, simple_loss=0.2601, pruned_loss=0.05196, over 16593.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2649, pruned_loss=0.04919, over 3327263.75 frames. ], batch size: 146, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:05,835 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:34,189 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:45,736 INFO [train.py:904] (3/8) Epoch 13, batch 3200, loss[loss=0.1716, simple_loss=0.2545, pruned_loss=0.04434, over 16819.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2641, pruned_loss=0.04875, over 3316152.74 frames. ], batch size: 102, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:49,042 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:28:10,526 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:28:29,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0917, 4.8158, 5.1323, 5.3396, 5.4906, 4.8081, 5.4814, 5.4707], device='cuda:3'), covar=tensor([0.1739, 0.1320, 0.1684, 0.0636, 0.0523, 0.0751, 0.0451, 0.0527], device='cuda:3'), in_proj_covar=tensor([0.0582, 0.0731, 0.0882, 0.0740, 0.0555, 0.0583, 0.0586, 0.0678], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:28:50,549 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8278, 2.9779, 2.8656, 4.8912, 3.9751, 4.4472, 1.7705, 3.2536], device='cuda:3'), covar=tensor([0.1362, 0.0748, 0.1071, 0.0207, 0.0257, 0.0365, 0.1526, 0.0708], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0162, 0.0182, 0.0158, 0.0198, 0.0210, 0.0183, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 18:28:56,249 INFO [train.py:904] (3/8) Epoch 13, batch 3250, loss[loss=0.1639, simple_loss=0.2429, pruned_loss=0.04244, over 17004.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2637, pruned_loss=0.04845, over 3315077.21 frames. ], batch size: 41, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:28:56,531 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:29:27,129 INFO [optim.py:368] (3/8) 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 18:29:46,790 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:05,563 INFO [train.py:904] (3/8) Epoch 13, batch 3300, loss[loss=0.2322, simple_loss=0.3076, pruned_loss=0.07835, over 16839.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2654, pruned_loss=0.04909, over 3321028.90 frames. ], batch size: 124, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:30:42,374 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:47,550 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0171, 2.3358, 2.4373, 4.6960, 2.3236, 2.8104, 2.4476, 2.5912], device='cuda:3'), covar=tensor([0.0863, 0.3384, 0.2430, 0.0384, 0.3871, 0.2506, 0.3293, 0.3267], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0404, 0.0338, 0.0324, 0.0417, 0.0467, 0.0368, 0.0474], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:30:53,994 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:57,872 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 18:31:15,061 INFO [train.py:904] (3/8) Epoch 13, batch 3350, loss[loss=0.1684, simple_loss=0.2632, pruned_loss=0.03679, over 17292.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2664, pruned_loss=0.04932, over 3321250.44 frames. ], batch size: 52, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:31:25,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2476, 2.1635, 2.6879, 3.2122, 3.0313, 3.5908, 2.3648, 3.5761], device='cuda:3'), covar=tensor([0.0170, 0.0380, 0.0237, 0.0218, 0.0218, 0.0143, 0.0344, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0162, 0.0169, 0.0177, 0.0132, 0.0177, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:31:38,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9987, 4.7061, 4.9362, 5.2304, 5.3926, 4.6216, 5.3456, 5.3745], device='cuda:3'), covar=tensor([0.1669, 0.1591, 0.2163, 0.0894, 0.0678, 0.1000, 0.0685, 0.0746], device='cuda:3'), in_proj_covar=tensor([0.0586, 0.0737, 0.0888, 0.0743, 0.0559, 0.0584, 0.0592, 0.0681], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:31:45,937 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.375e+02 2.737e+02 3.421e+02 8.798e+02, threshold=5.473e+02, percent-clipped=2.0 2023-04-29 18:31:49,124 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:31:56,688 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8332, 2.0067, 2.3914, 2.8556, 2.6446, 3.2529, 2.1283, 3.3127], device='cuda:3'), covar=tensor([0.0215, 0.0377, 0.0277, 0.0263, 0.0263, 0.0159, 0.0384, 0.0117], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0163, 0.0169, 0.0177, 0.0133, 0.0177, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:32:06,943 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:24,433 INFO [train.py:904] (3/8) Epoch 13, batch 3400, loss[loss=0.1828, simple_loss=0.2539, pruned_loss=0.05588, over 16878.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2661, pruned_loss=0.04875, over 3327943.14 frames. ], batch size: 116, lr: 5.29e-03, grad_scale: 4.0 2023-04-29 18:33:26,386 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5903, 3.6252, 3.9473, 2.1119, 4.0104, 4.0773, 3.1815, 3.2405], device='cuda:3'), covar=tensor([0.0706, 0.0194, 0.0161, 0.1055, 0.0076, 0.0144, 0.0332, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0137, 0.0070, 0.0111, 0.0121, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 18:33:27,528 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3180, 4.1550, 4.3651, 4.5324, 4.6074, 4.1310, 4.3782, 4.6142], device='cuda:3'), covar=tensor([0.1346, 0.1131, 0.1204, 0.0648, 0.0597, 0.1170, 0.1738, 0.0823], device='cuda:3'), in_proj_covar=tensor([0.0586, 0.0738, 0.0891, 0.0744, 0.0559, 0.0586, 0.0592, 0.0683], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:33:35,396 INFO [train.py:904] (3/8) Epoch 13, batch 3450, loss[loss=0.2125, simple_loss=0.2812, pruned_loss=0.07188, over 11613.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2652, pruned_loss=0.04885, over 3308911.80 frames. ], batch size: 246, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:33:44,579 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1872, 4.2740, 4.5886, 4.5897, 4.5922, 4.2716, 4.3105, 4.1673], device='cuda:3'), covar=tensor([0.0339, 0.0664, 0.0408, 0.0434, 0.0463, 0.0432, 0.0805, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0365, 0.0385, 0.0384, 0.0363, 0.0432, 0.0407, 0.0508, 0.0327], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 18:34:07,280 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.393e+02 2.670e+02 3.290e+02 7.203e+02, threshold=5.341e+02, percent-clipped=1.0 2023-04-29 18:34:35,531 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:34:46,233 INFO [train.py:904] (3/8) Epoch 13, batch 3500, loss[loss=0.1731, simple_loss=0.2576, pruned_loss=0.04428, over 16769.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2642, pruned_loss=0.04879, over 3312849.08 frames. ], batch size: 83, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:35:09,862 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:43,533 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 3550, loss[loss=0.1637, simple_loss=0.2571, pruned_loss=0.0352, over 17124.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2635, pruned_loss=0.04834, over 3314815.09 frames. ], batch size: 47, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:36:03,800 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:03,882 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8759, 1.3862, 1.6811, 1.6685, 1.7575, 1.8941, 1.5016, 1.7891], device='cuda:3'), covar=tensor([0.0183, 0.0314, 0.0159, 0.0230, 0.0200, 0.0137, 0.0317, 0.0093], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0180, 0.0164, 0.0170, 0.0179, 0.0134, 0.0178, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:36:18,663 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:30,597 INFO [optim.py:368] (3/8) 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 18:36:32,837 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:57,618 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6761, 3.6018, 3.7915, 3.5480, 3.7017, 4.1364, 3.8411, 3.4805], device='cuda:3'), covar=tensor([0.2054, 0.2362, 0.1859, 0.2826, 0.2745, 0.2105, 0.1399, 0.2748], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0530, 0.0573, 0.0458, 0.0620, 0.0597, 0.0455, 0.0609], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:37:08,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8146, 4.3433, 3.2473, 2.2383, 2.8824, 2.5534, 4.6900, 3.6290], device='cuda:3'), covar=tensor([0.2664, 0.0562, 0.1504, 0.2726, 0.2533, 0.1802, 0.0333, 0.1249], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0261, 0.0288, 0.0287, 0.0286, 0.0230, 0.0275, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:37:08,976 INFO [train.py:904] (3/8) Epoch 13, batch 3600, loss[loss=0.173, simple_loss=0.2642, pruned_loss=0.04087, over 17102.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2619, pruned_loss=0.04781, over 3299332.03 frames. ], batch size: 47, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:37:31,354 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 18:37:35,961 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:59,798 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:38:12,539 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8579, 5.2044, 4.9517, 4.9566, 4.6698, 4.6368, 4.6709, 5.2465], device='cuda:3'), covar=tensor([0.1037, 0.0851, 0.0991, 0.0833, 0.0767, 0.0956, 0.1039, 0.0961], device='cuda:3'), in_proj_covar=tensor([0.0582, 0.0737, 0.0595, 0.0523, 0.0464, 0.0469, 0.0612, 0.0566], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:38:21,942 INFO [train.py:904] (3/8) Epoch 13, batch 3650, loss[loss=0.1787, simple_loss=0.2498, pruned_loss=0.05377, over 16816.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2614, pruned_loss=0.04845, over 3291394.47 frames. ], batch size: 124, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:38:57,386 INFO [optim.py:368] (3/8) 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 18:38:59,589 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:07,719 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:11,359 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:36,233 INFO [train.py:904] (3/8) Epoch 13, batch 3700, loss[loss=0.1862, simple_loss=0.2614, pruned_loss=0.05548, over 16413.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2599, pruned_loss=0.04988, over 3269770.56 frames. ], batch size: 146, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:09,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5181, 5.8344, 5.5714, 5.6875, 5.2265, 4.9923, 5.2709, 5.8977], device='cuda:3'), covar=tensor([0.0974, 0.0649, 0.0913, 0.0639, 0.0724, 0.0709, 0.0916, 0.0802], device='cuda:3'), in_proj_covar=tensor([0.0577, 0.0728, 0.0589, 0.0517, 0.0461, 0.0465, 0.0605, 0.0561], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:40:10,905 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:40:31,265 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7249, 1.8334, 2.3348, 2.6348, 2.7192, 2.5204, 1.8266, 2.8370], device='cuda:3'), covar=tensor([0.0128, 0.0376, 0.0230, 0.0216, 0.0184, 0.0223, 0.0386, 0.0088], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0163, 0.0170, 0.0178, 0.0133, 0.0177, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:40:51,181 INFO [train.py:904] (3/8) Epoch 13, batch 3750, loss[loss=0.171, simple_loss=0.2448, pruned_loss=0.04863, over 16718.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2603, pruned_loss=0.0513, over 3266152.58 frames. ], batch size: 89, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:41:24,163 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.222e+02 2.735e+02 3.211e+02 5.083e+02, threshold=5.471e+02, percent-clipped=0.0 2023-04-29 18:41:27,355 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3755, 3.7084, 4.0633, 2.8629, 3.6104, 4.0180, 3.7600, 2.3195], device='cuda:3'), covar=tensor([0.0410, 0.0150, 0.0033, 0.0245, 0.0076, 0.0085, 0.0059, 0.0350], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0073, 0.0071, 0.0126, 0.0083, 0.0093, 0.0083, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:42:05,166 INFO [train.py:904] (3/8) Epoch 13, batch 3800, loss[loss=0.2074, simple_loss=0.2855, pruned_loss=0.06467, over 12611.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.262, pruned_loss=0.05262, over 3256702.57 frames. ], batch size: 248, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:42:07,403 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:42:11,261 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4244, 2.8932, 2.9329, 1.8888, 2.5748, 2.1160, 3.0694, 3.1183], device='cuda:3'), covar=tensor([0.0301, 0.0795, 0.0571, 0.1858, 0.0858, 0.0967, 0.0590, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0152, 0.0160, 0.0146, 0.0138, 0.0126, 0.0138, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 18:43:18,385 INFO [train.py:904] (3/8) Epoch 13, batch 3850, loss[loss=0.1879, simple_loss=0.2638, pruned_loss=0.056, over 16843.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2619, pruned_loss=0.05306, over 3257639.92 frames. ], batch size: 42, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:43:35,124 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:39,941 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6915, 3.8120, 4.0175, 4.0141, 4.0572, 3.8302, 3.7078, 3.7820], device='cuda:3'), covar=tensor([0.0520, 0.0710, 0.0571, 0.0636, 0.0689, 0.0608, 0.1236, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0361, 0.0380, 0.0376, 0.0360, 0.0427, 0.0401, 0.0500, 0.0321], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 18:43:50,561 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.444e+02 2.805e+02 3.444e+02 9.574e+02, threshold=5.610e+02, percent-clipped=2.0 2023-04-29 18:44:29,772 INFO [train.py:904] (3/8) Epoch 13, batch 3900, loss[loss=0.1792, simple_loss=0.2546, pruned_loss=0.05192, over 16753.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2607, pruned_loss=0.05298, over 3274058.42 frames. ], batch size: 124, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:44:45,080 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 18:44:49,744 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1994, 2.0256, 2.2587, 3.8951, 2.1468, 2.3942, 2.1541, 2.2332], device='cuda:3'), covar=tensor([0.1144, 0.3403, 0.2322, 0.0463, 0.3378, 0.2450, 0.3352, 0.2880], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0408, 0.0340, 0.0326, 0.0419, 0.0472, 0.0372, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:44:58,108 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 18:45:03,283 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 13, batch 3950, loss[loss=0.1686, simple_loss=0.2395, pruned_loss=0.04884, over 16727.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.26, pruned_loss=0.05313, over 3280777.85 frames. ], batch size: 83, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:46:16,760 INFO [optim.py:368] (3/8) 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 18:46:17,243 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2734, 3.2365, 3.5239, 1.8275, 3.5553, 3.6069, 2.8366, 2.7636], device='cuda:3'), covar=tensor([0.0780, 0.0204, 0.0165, 0.1069, 0.0100, 0.0217, 0.0414, 0.0429], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0138, 0.0070, 0.0111, 0.0121, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 18:46:20,056 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:29,307 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:32,705 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:44,278 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 18:46:53,453 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5920, 4.6641, 4.8020, 4.6895, 4.7298, 5.2996, 4.8385, 4.5817], device='cuda:3'), covar=tensor([0.1248, 0.1776, 0.1843, 0.2091, 0.2641, 0.0989, 0.1346, 0.2333], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0528, 0.0572, 0.0456, 0.0614, 0.0595, 0.0453, 0.0607], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:46:55,491 INFO [train.py:904] (3/8) Epoch 13, batch 4000, loss[loss=0.1934, simple_loss=0.278, pruned_loss=0.05447, over 16685.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2608, pruned_loss=0.05403, over 3271290.50 frames. ], batch size: 89, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:47:37,207 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 4050, loss[loss=0.1823, simple_loss=0.2575, pruned_loss=0.05351, over 17230.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2609, pruned_loss=0.05308, over 3279015.85 frames. ], batch size: 45, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:48:15,083 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1344, 3.1251, 3.6108, 1.6701, 3.7314, 3.7755, 2.8657, 2.7741], device='cuda:3'), covar=tensor([0.0844, 0.0277, 0.0146, 0.1214, 0.0051, 0.0106, 0.0421, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0137, 0.0070, 0.0110, 0.0121, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 18:48:23,627 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-29 18:48:36,496 INFO [optim.py:368] (3/8) 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 18:48:45,109 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6883, 3.1048, 3.2098, 1.8863, 2.7543, 2.1559, 3.3181, 3.2737], device='cuda:3'), covar=tensor([0.0245, 0.0663, 0.0560, 0.1888, 0.0798, 0.0968, 0.0537, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0152, 0.0160, 0.0146, 0.0138, 0.0126, 0.0137, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 18:49:15,115 INFO [train.py:904] (3/8) Epoch 13, batch 4100, loss[loss=0.2172, simple_loss=0.2925, pruned_loss=0.07093, over 16676.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2627, pruned_loss=0.05261, over 3268539.08 frames. ], batch size: 134, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:49:16,198 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 18:49:48,019 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3987, 2.9512, 2.5546, 2.2259, 2.1904, 2.1976, 2.8761, 2.7920], device='cuda:3'), covar=tensor([0.2204, 0.0751, 0.1512, 0.2072, 0.2025, 0.1801, 0.0477, 0.0990], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0260, 0.0290, 0.0287, 0.0288, 0.0229, 0.0275, 0.0307], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:50:12,059 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0122, 3.1623, 3.2136, 2.0504, 2.9756, 3.2125, 2.9835, 1.8806], device='cuda:3'), covar=tensor([0.0449, 0.0043, 0.0041, 0.0371, 0.0082, 0.0103, 0.0082, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0072, 0.0072, 0.0127, 0.0083, 0.0093, 0.0084, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:50:13,157 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6345, 3.5341, 4.1978, 1.7976, 4.3787, 4.4719, 2.8879, 3.2797], device='cuda:3'), covar=tensor([0.0758, 0.0235, 0.0146, 0.1230, 0.0047, 0.0074, 0.0465, 0.0404], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0101, 0.0088, 0.0136, 0.0070, 0.0109, 0.0121, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 18:50:19,778 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9664, 3.0539, 2.6307, 4.6312, 3.5832, 4.1259, 1.6258, 2.9382], device='cuda:3'), covar=tensor([0.1092, 0.0575, 0.1074, 0.0125, 0.0307, 0.0429, 0.1383, 0.0803], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0163, 0.0184, 0.0159, 0.0201, 0.0210, 0.0185, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 18:50:30,154 INFO [train.py:904] (3/8) Epoch 13, batch 4150, loss[loss=0.1983, simple_loss=0.2886, pruned_loss=0.05401, over 16731.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2704, pruned_loss=0.05539, over 3229498.28 frames. ], batch size: 124, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:42,637 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:05,850 INFO [optim.py:368] (3/8) 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 18:51:08,757 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6738, 2.6653, 2.1385, 2.4458, 2.9858, 2.6816, 3.3404, 3.2849], device='cuda:3'), covar=tensor([0.0062, 0.0295, 0.0406, 0.0330, 0.0196, 0.0285, 0.0172, 0.0163], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0214, 0.0207, 0.0209, 0.0215, 0.0213, 0.0223, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:51:27,938 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0954, 2.0508, 2.1388, 3.6348, 1.9632, 2.4291, 2.1578, 2.2240], device='cuda:3'), covar=tensor([0.1078, 0.3245, 0.2315, 0.0452, 0.3663, 0.2182, 0.2968, 0.2875], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0409, 0.0339, 0.0325, 0.0419, 0.0474, 0.0372, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 18:51:40,528 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:48,708 INFO [train.py:904] (3/8) Epoch 13, batch 4200, loss[loss=0.2332, simple_loss=0.3126, pruned_loss=0.07694, over 16659.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2768, pruned_loss=0.0565, over 3224271.38 frames. ], batch size: 57, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:52:04,528 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 18:52:12,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9367, 4.6989, 4.6431, 3.2788, 3.9666, 4.5286, 3.9716, 2.7612], device='cuda:3'), covar=tensor([0.0370, 0.0019, 0.0026, 0.0259, 0.0066, 0.0088, 0.0062, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0071, 0.0072, 0.0128, 0.0084, 0.0093, 0.0084, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:52:34,302 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 4250, loss[loss=0.1841, simple_loss=0.2758, pruned_loss=0.04618, over 16481.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.28, pruned_loss=0.05616, over 3208061.25 frames. ], batch size: 68, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:53:13,085 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:14,082 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:36,046 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.365e+02 2.942e+02 3.644e+02 5.722e+02, threshold=5.884e+02, percent-clipped=0.0 2023-04-29 18:53:39,677 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:43,861 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:45,088 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:54:16,796 INFO [train.py:904] (3/8) Epoch 13, batch 4300, loss[loss=0.2078, simple_loss=0.3012, pruned_loss=0.05724, over 16758.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2812, pruned_loss=0.05528, over 3212460.44 frames. ], batch size: 124, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:54:51,666 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:55:21,337 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1466, 3.6376, 3.6110, 2.2674, 3.2728, 3.5459, 3.2862, 1.9677], device='cuda:3'), covar=tensor([0.0419, 0.0029, 0.0035, 0.0346, 0.0083, 0.0089, 0.0079, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0071, 0.0072, 0.0127, 0.0083, 0.0092, 0.0083, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 18:55:21,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5633, 2.2666, 2.1465, 3.5547, 2.5564, 3.5402, 1.4337, 2.6418], device='cuda:3'), covar=tensor([0.1326, 0.0845, 0.1378, 0.0191, 0.0257, 0.0437, 0.1537, 0.0854], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0160, 0.0182, 0.0156, 0.0197, 0.0206, 0.0181, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 18:55:30,687 INFO [train.py:904] (3/8) Epoch 13, batch 4350, loss[loss=0.1918, simple_loss=0.2824, pruned_loss=0.05058, over 17233.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2847, pruned_loss=0.05652, over 3211162.97 frames. ], batch size: 45, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:56:06,023 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.441e+02 2.808e+02 3.338e+02 6.392e+02, threshold=5.616e+02, percent-clipped=2.0 2023-04-29 18:56:46,664 INFO [train.py:904] (3/8) Epoch 13, batch 4400, loss[loss=0.1985, simple_loss=0.2823, pruned_loss=0.05738, over 16666.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2866, pruned_loss=0.05728, over 3213480.86 frames. ], batch size: 57, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:57:59,273 INFO [train.py:904] (3/8) Epoch 13, batch 4450, loss[loss=0.2241, simple_loss=0.3159, pruned_loss=0.06615, over 16296.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2904, pruned_loss=0.05873, over 3216895.14 frames. ], batch size: 165, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:58:10,184 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:58:17,744 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8124, 3.9999, 2.4759, 4.7995, 2.9748, 4.6298, 2.5608, 2.9907], device='cuda:3'), covar=tensor([0.0259, 0.0340, 0.1521, 0.0098, 0.0792, 0.0342, 0.1393, 0.0753], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0167, 0.0189, 0.0140, 0.0168, 0.0212, 0.0197, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 18:58:33,037 INFO [optim.py:368] (3/8) 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 18:59:14,452 INFO [train.py:904] (3/8) Epoch 13, batch 4500, loss[loss=0.1932, simple_loss=0.274, pruned_loss=0.05615, over 16637.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2905, pruned_loss=0.05902, over 3219095.61 frames. ], batch size: 76, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:59:21,402 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:59:56,982 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7233, 1.3503, 1.6615, 1.6792, 1.8186, 1.9094, 1.5286, 1.8026], device='cuda:3'), covar=tensor([0.0174, 0.0284, 0.0134, 0.0175, 0.0172, 0.0128, 0.0290, 0.0081], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0176, 0.0159, 0.0165, 0.0176, 0.0132, 0.0175, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:00:10,268 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 19:00:25,572 INFO [train.py:904] (3/8) Epoch 13, batch 4550, loss[loss=0.2193, simple_loss=0.2968, pruned_loss=0.07092, over 16550.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2912, pruned_loss=0.06017, over 3220375.97 frames. ], batch size: 68, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:00:28,789 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:39,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0916, 2.8830, 3.1159, 1.8143, 3.2417, 3.3117, 2.5703, 2.6333], device='cuda:3'), covar=tensor([0.0847, 0.0264, 0.0209, 0.1069, 0.0073, 0.0137, 0.0481, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0103, 0.0089, 0.0138, 0.0070, 0.0110, 0.0123, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 19:00:59,196 INFO [optim.py:368] (3/8) 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 19:01:06,184 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:07,327 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:35,434 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6076, 4.5663, 4.3359, 3.5552, 4.4917, 1.5358, 4.1983, 3.8621], device='cuda:3'), covar=tensor([0.0055, 0.0051, 0.0125, 0.0336, 0.0055, 0.2755, 0.0094, 0.0243], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0126, 0.0171, 0.0161, 0.0144, 0.0182, 0.0160, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:01:36,168 INFO [train.py:904] (3/8) Epoch 13, batch 4600, loss[loss=0.2087, simple_loss=0.2981, pruned_loss=0.0597, over 16852.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2915, pruned_loss=0.05969, over 3243410.09 frames. ], batch size: 116, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:02:00,574 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.4429, 2.5754, 2.4658, 3.7692, 3.0546, 3.8504, 1.4243, 2.7861], device='cuda:3'), covar=tensor([0.1414, 0.0737, 0.1194, 0.0165, 0.0312, 0.0378, 0.1643, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0161, 0.0183, 0.0157, 0.0198, 0.0206, 0.0182, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-29 19:02:22,725 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:23,012 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1542, 2.1215, 2.1363, 3.9046, 2.0931, 2.5445, 2.2508, 2.2688], device='cuda:3'), covar=tensor([0.1112, 0.2971, 0.2342, 0.0396, 0.3728, 0.2170, 0.2817, 0.3230], device='cuda:3'), in_proj_covar=tensor([0.0372, 0.0405, 0.0336, 0.0319, 0.0419, 0.0470, 0.0369, 0.0473], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:02:26,994 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-29 19:02:39,289 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:53,216 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 4650, loss[loss=0.1906, simple_loss=0.278, pruned_loss=0.05163, over 16759.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2906, pruned_loss=0.05962, over 3242364.26 frames. ], batch size: 83, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:03:27,525 INFO [optim.py:368] (3/8) 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 19:03:35,743 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 19:04:07,652 INFO [train.py:904] (3/8) Epoch 13, batch 4700, loss[loss=0.1966, simple_loss=0.2724, pruned_loss=0.06034, over 11664.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2875, pruned_loss=0.05856, over 3235508.87 frames. ], batch size: 250, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:04:21,533 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:05:13,246 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 19:05:21,228 INFO [train.py:904] (3/8) Epoch 13, batch 4750, loss[loss=0.1676, simple_loss=0.2519, pruned_loss=0.04167, over 16989.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2839, pruned_loss=0.05695, over 3217723.45 frames. ], batch size: 53, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:05:53,852 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 4800, loss[loss=0.1794, simple_loss=0.2715, pruned_loss=0.0437, over 16427.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2798, pruned_loss=0.05455, over 3221594.16 frames. ], batch size: 75, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:07:26,318 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-29 19:07:43,476 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9863, 2.2998, 2.3087, 2.8442, 2.1093, 3.2479, 1.6641, 2.7541], device='cuda:3'), covar=tensor([0.1089, 0.0599, 0.1050, 0.0136, 0.0109, 0.0331, 0.1409, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0159, 0.0201, 0.0208, 0.0184, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 19:07:46,141 INFO [train.py:904] (3/8) Epoch 13, batch 4850, loss[loss=0.1919, simple_loss=0.284, pruned_loss=0.04994, over 16607.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2812, pruned_loss=0.05408, over 3201012.92 frames. ], batch size: 62, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:07:50,113 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:08:19,805 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 4900, loss[loss=0.1859, simple_loss=0.2812, pruned_loss=0.04531, over 16148.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2799, pruned_loss=0.05229, over 3193314.89 frames. ], batch size: 165, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:08:59,539 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:09:49,778 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 4950, loss[loss=0.216, simple_loss=0.2994, pruned_loss=0.06631, over 16517.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.28, pruned_loss=0.05195, over 3183409.17 frames. ], batch size: 68, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:10:44,873 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4246, 4.5099, 4.3063, 4.0037, 3.9381, 4.3876, 4.2258, 4.0985], device='cuda:3'), covar=tensor([0.0542, 0.0369, 0.0249, 0.0258, 0.0916, 0.0349, 0.0457, 0.0557], device='cuda:3'), in_proj_covar=tensor([0.0249, 0.0339, 0.0299, 0.0280, 0.0320, 0.0323, 0.0203, 0.0349], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:10:45,730 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 5000, loss[loss=0.1839, simple_loss=0.2777, pruned_loss=0.04507, over 16771.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2817, pruned_loss=0.05226, over 3192393.95 frames. ], batch size: 124, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:11:32,725 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:11:47,320 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:11:49,133 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6808, 3.6354, 4.4394, 1.9286, 4.5429, 4.5864, 3.1009, 3.1972], device='cuda:3'), covar=tensor([0.0744, 0.0224, 0.0099, 0.1231, 0.0031, 0.0067, 0.0369, 0.0446], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0101, 0.0087, 0.0135, 0.0069, 0.0107, 0.0120, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 19:12:38,197 INFO [train.py:904] (3/8) Epoch 13, batch 5050, loss[loss=0.2188, simple_loss=0.3, pruned_loss=0.0688, over 12128.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2817, pruned_loss=0.05194, over 3210703.16 frames. ], batch size: 246, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:13:11,124 INFO [optim.py:368] (3/8) 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 19:13:14,652 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:13:39,832 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1972, 3.4468, 3.7263, 2.0834, 3.1214, 2.4294, 3.5540, 3.4902], device='cuda:3'), covar=tensor([0.0216, 0.0685, 0.0473, 0.1811, 0.0691, 0.0810, 0.0643, 0.0952], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0151, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 19:13:51,463 INFO [train.py:904] (3/8) Epoch 13, batch 5100, loss[loss=0.1863, simple_loss=0.2672, pruned_loss=0.05265, over 16653.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2795, pruned_loss=0.05104, over 3222637.64 frames. ], batch size: 57, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:14:10,631 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7901, 4.8155, 4.6221, 4.2746, 4.2103, 4.6645, 4.5558, 4.4095], device='cuda:3'), covar=tensor([0.0533, 0.0311, 0.0274, 0.0269, 0.1051, 0.0392, 0.0418, 0.0612], device='cuda:3'), in_proj_covar=tensor([0.0251, 0.0343, 0.0302, 0.0282, 0.0324, 0.0327, 0.0206, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:15:04,900 INFO [train.py:904] (3/8) Epoch 13, batch 5150, loss[loss=0.2202, simple_loss=0.3002, pruned_loss=0.07012, over 12052.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2802, pruned_loss=0.05062, over 3217886.09 frames. ], batch size: 246, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:15:14,427 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4814, 4.2089, 4.4736, 4.6766, 4.8712, 4.3926, 4.8184, 4.8384], device='cuda:3'), covar=tensor([0.1452, 0.1425, 0.1714, 0.0764, 0.0446, 0.0868, 0.0525, 0.0628], device='cuda:3'), in_proj_covar=tensor([0.0541, 0.0678, 0.0810, 0.0681, 0.0514, 0.0535, 0.0547, 0.0624], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:15:37,787 INFO [optim.py:368] (3/8) 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 19:16:17,748 INFO [train.py:904] (3/8) Epoch 13, batch 5200, loss[loss=0.1839, simple_loss=0.2697, pruned_loss=0.04905, over 16872.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2788, pruned_loss=0.05012, over 3213303.72 frames. ], batch size: 116, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:17:08,292 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:17:31,277 INFO [train.py:904] (3/8) Epoch 13, batch 5250, loss[loss=0.1653, simple_loss=0.259, pruned_loss=0.0358, over 16683.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2767, pruned_loss=0.04975, over 3204290.55 frames. ], batch size: 89, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:17:49,414 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 19:17:50,426 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-04-29 19:18:04,188 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:44,211 INFO [train.py:904] (3/8) Epoch 13, batch 5300, loss[loss=0.1638, simple_loss=0.2458, pruned_loss=0.04092, over 16414.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2731, pruned_loss=0.04879, over 3214770.88 frames. ], batch size: 68, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:50,651 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:19:29,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3906, 1.5760, 2.0614, 2.3423, 2.4313, 2.7167, 1.7477, 2.6259], device='cuda:3'), covar=tensor([0.0147, 0.0425, 0.0238, 0.0264, 0.0212, 0.0144, 0.0402, 0.0089], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0176, 0.0160, 0.0165, 0.0174, 0.0130, 0.0175, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:19:58,315 INFO [train.py:904] (3/8) Epoch 13, batch 5350, loss[loss=0.1908, simple_loss=0.2831, pruned_loss=0.04928, over 16383.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2719, pruned_loss=0.04816, over 3199109.26 frames. ], batch size: 146, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:20:01,845 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:27,668 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 13, batch 5400, loss[loss=0.1983, simple_loss=0.2904, pruned_loss=0.05306, over 16933.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2747, pruned_loss=0.04886, over 3202942.37 frames. ], batch size: 109, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:22:20,883 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:22:26,302 INFO [train.py:904] (3/8) Epoch 13, batch 5450, loss[loss=0.2172, simple_loss=0.3051, pruned_loss=0.06468, over 16809.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2784, pruned_loss=0.05079, over 3201700.86 frames. ], batch size: 83, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:02,325 INFO [optim.py:368] (3/8) 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 19:23:25,732 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.33 vs. limit=5.0 2023-04-29 19:23:43,865 INFO [train.py:904] (3/8) Epoch 13, batch 5500, loss[loss=0.2127, simple_loss=0.3048, pruned_loss=0.06033, over 16874.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2856, pruned_loss=0.05508, over 3192373.66 frames. ], batch size: 96, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:57,071 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:24:07,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0633, 1.5704, 1.9150, 2.0951, 2.2361, 2.3921, 1.6721, 2.2847], device='cuda:3'), covar=tensor([0.0170, 0.0341, 0.0178, 0.0229, 0.0199, 0.0128, 0.0322, 0.0091], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0176, 0.0160, 0.0164, 0.0174, 0.0129, 0.0175, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:25:00,716 INFO [train.py:904] (3/8) Epoch 13, batch 5550, loss[loss=0.2205, simple_loss=0.3055, pruned_loss=0.06777, over 16379.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2937, pruned_loss=0.0612, over 3159302.33 frames. ], batch size: 146, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:25:36,619 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5711, 2.6059, 1.8238, 2.7277, 2.1209, 2.7349, 2.0418, 2.3039], device='cuda:3'), covar=tensor([0.0273, 0.0324, 0.1130, 0.0191, 0.0562, 0.0475, 0.1042, 0.0563], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0165, 0.0189, 0.0136, 0.0166, 0.0206, 0.0195, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 19:25:38,573 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.215e+02 3.416e+02 4.230e+02 5.078e+02 9.045e+02, threshold=8.460e+02, percent-clipped=18.0 2023-04-29 19:26:20,876 INFO [train.py:904] (3/8) Epoch 13, batch 5600, loss[loss=0.2926, simple_loss=0.344, pruned_loss=0.1206, over 11090.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2985, pruned_loss=0.06583, over 3108527.21 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:26:48,501 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4863, 4.5855, 4.7487, 4.5774, 4.5903, 5.1304, 4.7070, 4.4780], device='cuda:3'), covar=tensor([0.1271, 0.1922, 0.2082, 0.1780, 0.2470, 0.1127, 0.1590, 0.2418], device='cuda:3'), in_proj_covar=tensor([0.0362, 0.0506, 0.0544, 0.0430, 0.0586, 0.0575, 0.0437, 0.0586], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 19:27:02,226 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:27:41,559 INFO [train.py:904] (3/8) Epoch 13, batch 5650, loss[loss=0.2776, simple_loss=0.3297, pruned_loss=0.1127, over 11408.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3039, pruned_loss=0.07006, over 3104860.41 frames. ], batch size: 253, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:28:12,100 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:28:17,741 INFO [optim.py:368] (3/8) 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 19:28:36,916 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 19:28:58,366 INFO [train.py:904] (3/8) Epoch 13, batch 5700, loss[loss=0.2918, simple_loss=0.3432, pruned_loss=0.1203, over 11454.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3062, pruned_loss=0.07271, over 3080759.00 frames. ], batch size: 246, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:29:27,678 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:33,278 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6020, 1.7042, 2.2238, 2.5730, 2.6071, 2.8913, 1.8407, 2.8226], device='cuda:3'), covar=tensor([0.0165, 0.0408, 0.0229, 0.0219, 0.0194, 0.0162, 0.0380, 0.0105], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0162, 0.0173, 0.0128, 0.0173, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:29:41,979 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:48,673 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 19:30:18,041 INFO [train.py:904] (3/8) Epoch 13, batch 5750, loss[loss=0.2202, simple_loss=0.3117, pruned_loss=0.06435, over 16923.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3087, pruned_loss=0.0743, over 3058562.17 frames. ], batch size: 96, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:30:40,870 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7393, 1.3218, 1.6692, 1.6020, 1.8343, 1.8392, 1.5641, 1.7822], device='cuda:3'), covar=tensor([0.0193, 0.0298, 0.0147, 0.0184, 0.0201, 0.0145, 0.0332, 0.0092], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0176, 0.0159, 0.0162, 0.0173, 0.0129, 0.0174, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:30:56,349 INFO [optim.py:368] (3/8) 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 19:31:21,775 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:39,674 INFO [train.py:904] (3/8) Epoch 13, batch 5800, loss[loss=0.2111, simple_loss=0.2985, pruned_loss=0.06183, over 16349.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3082, pruned_loss=0.07214, over 3076169.81 frames. ], batch size: 146, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:31:44,371 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:32:57,618 INFO [train.py:904] (3/8) Epoch 13, batch 5850, loss[loss=0.1974, simple_loss=0.2727, pruned_loss=0.06109, over 11315.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3057, pruned_loss=0.07001, over 3088036.14 frames. ], batch size: 247, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:33:05,734 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2284, 5.5155, 5.2518, 5.2732, 4.9424, 4.9016, 4.9643, 5.6352], device='cuda:3'), covar=tensor([0.0923, 0.0728, 0.0960, 0.0703, 0.0754, 0.0748, 0.1009, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0563, 0.0698, 0.0573, 0.0497, 0.0441, 0.0450, 0.0584, 0.0539], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:33:15,137 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:33:22,569 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1962, 4.8853, 5.1304, 5.3219, 5.5343, 4.8526, 5.5451, 5.4899], device='cuda:3'), covar=tensor([0.1408, 0.1332, 0.1477, 0.0629, 0.0436, 0.0784, 0.0366, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0540, 0.0676, 0.0804, 0.0684, 0.0516, 0.0531, 0.0542, 0.0623], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:33:36,796 INFO [optim.py:368] (3/8) 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 19:34:07,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6558, 3.9164, 3.0506, 2.1890, 2.6549, 2.4331, 4.1382, 3.6041], device='cuda:3'), covar=tensor([0.2632, 0.0550, 0.1475, 0.2450, 0.2151, 0.1733, 0.0384, 0.0932], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0258, 0.0287, 0.0285, 0.0281, 0.0227, 0.0271, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:34:19,531 INFO [train.py:904] (3/8) Epoch 13, batch 5900, loss[loss=0.2085, simple_loss=0.2958, pruned_loss=0.06066, over 16515.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3046, pruned_loss=0.06937, over 3094992.32 frames. ], batch size: 68, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:34:57,945 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:35:42,724 INFO [train.py:904] (3/8) Epoch 13, batch 5950, loss[loss=0.2037, simple_loss=0.292, pruned_loss=0.05773, over 16722.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3055, pruned_loss=0.06809, over 3102755.51 frames. ], batch size: 76, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:36:21,828 INFO [optim.py:368] (3/8) 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 19:36:26,846 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:36:32,811 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:36:50,114 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9450, 5.2574, 5.0067, 4.9529, 4.7428, 4.6691, 4.7074, 5.3441], device='cuda:3'), covar=tensor([0.1170, 0.0852, 0.0947, 0.0732, 0.0778, 0.0942, 0.1035, 0.0895], device='cuda:3'), in_proj_covar=tensor([0.0569, 0.0705, 0.0579, 0.0503, 0.0444, 0.0455, 0.0591, 0.0546], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:36:52,786 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 19:37:04,077 INFO [train.py:904] (3/8) Epoch 13, batch 6000, loss[loss=0.2012, simple_loss=0.284, pruned_loss=0.05914, over 17216.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3036, pruned_loss=0.06724, over 3107340.64 frames. ], batch size: 44, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:37:04,077 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 19:37:14,215 INFO [train.py:938] (3/8) Epoch 13, validation: loss=0.1599, simple_loss=0.2726, pruned_loss=0.02359, over 944034.00 frames. 2023-04-29 19:37:14,215 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 19:37:49,172 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 19:38:13,251 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:32,213 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 6050, loss[loss=0.2218, simple_loss=0.3103, pruned_loss=0.06662, over 16638.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.302, pruned_loss=0.06651, over 3106057.67 frames. ], batch size: 68, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:39:07,416 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:14,937 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:40,847 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 19:39:55,362 INFO [train.py:904] (3/8) Epoch 13, batch 6100, loss[loss=0.2121, simple_loss=0.2983, pruned_loss=0.0629, over 16658.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3018, pruned_loss=0.06602, over 3110931.07 frames. ], batch size: 62, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:40:01,516 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:12,610 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:35,801 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7061, 3.9177, 2.9395, 2.2488, 2.7589, 2.5243, 4.1287, 3.6031], device='cuda:3'), covar=tensor([0.2668, 0.0740, 0.1684, 0.2430, 0.2347, 0.1696, 0.0453, 0.1003], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0259, 0.0287, 0.0285, 0.0283, 0.0226, 0.0272, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:40:47,010 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 6150, loss[loss=0.2027, simple_loss=0.2838, pruned_loss=0.06082, over 16597.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2995, pruned_loss=0.0657, over 3107088.33 frames. ], batch size: 62, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:41:17,425 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:29,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 19:41:56,814 INFO [optim.py:368] (3/8) 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 19:41:57,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0045, 3.9847, 3.9561, 3.2521, 3.9581, 1.7354, 3.7661, 3.5720], device='cuda:3'), covar=tensor([0.0107, 0.0083, 0.0139, 0.0290, 0.0080, 0.2549, 0.0122, 0.0213], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0125, 0.0171, 0.0161, 0.0143, 0.0184, 0.0159, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:42:39,407 INFO [train.py:904] (3/8) Epoch 13, batch 6200, loss[loss=0.2062, simple_loss=0.2929, pruned_loss=0.05975, over 16660.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2984, pruned_loss=0.06535, over 3099788.39 frames. ], batch size: 62, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:42:59,887 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5191, 4.8238, 4.5722, 4.5663, 4.3091, 4.3235, 4.3342, 4.8602], device='cuda:3'), covar=tensor([0.1205, 0.0834, 0.1055, 0.0804, 0.0897, 0.1236, 0.1043, 0.0933], device='cuda:3'), in_proj_covar=tensor([0.0567, 0.0702, 0.0578, 0.0504, 0.0443, 0.0453, 0.0588, 0.0542], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:43:00,165 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-29 19:43:05,152 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:43:37,003 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 19:43:57,679 INFO [train.py:904] (3/8) Epoch 13, batch 6250, loss[loss=0.1828, simple_loss=0.2803, pruned_loss=0.04267, over 16759.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2973, pruned_loss=0.0646, over 3108966.27 frames. ], batch size: 83, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:44:04,279 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8765, 4.8400, 4.7449, 4.0182, 4.7642, 1.7321, 4.5343, 4.4888], device='cuda:3'), covar=tensor([0.0089, 0.0082, 0.0150, 0.0347, 0.0102, 0.2386, 0.0144, 0.0181], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0125, 0.0171, 0.0162, 0.0143, 0.0184, 0.0159, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:44:37,187 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.817e+02 3.684e+02 4.282e+02 1.153e+03, threshold=7.367e+02, percent-clipped=4.0 2023-04-29 19:44:45,113 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:45:13,034 INFO [train.py:904] (3/8) Epoch 13, batch 6300, loss[loss=0.2031, simple_loss=0.2802, pruned_loss=0.063, over 17104.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.297, pruned_loss=0.06419, over 3114711.44 frames. ], batch size: 49, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:45:34,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2251, 4.2055, 4.0999, 3.4615, 4.1503, 1.6549, 3.9628, 3.7838], device='cuda:3'), covar=tensor([0.0085, 0.0079, 0.0141, 0.0293, 0.0078, 0.2408, 0.0116, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0125, 0.0171, 0.0161, 0.0143, 0.0184, 0.0159, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:46:02,725 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:46:02,870 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0330, 2.4969, 2.6241, 1.9327, 2.7148, 2.7843, 2.4180, 2.3495], device='cuda:3'), covar=tensor([0.0653, 0.0187, 0.0175, 0.0847, 0.0082, 0.0211, 0.0391, 0.0410], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0101, 0.0088, 0.0138, 0.0069, 0.0109, 0.0121, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 19:46:08,186 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:46:28,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6810, 1.7219, 2.2056, 2.5585, 2.6333, 2.9342, 1.6939, 2.9064], device='cuda:3'), covar=tensor([0.0151, 0.0440, 0.0271, 0.0243, 0.0225, 0.0147, 0.0457, 0.0098], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0175, 0.0159, 0.0162, 0.0173, 0.0130, 0.0175, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:46:33,870 INFO [train.py:904] (3/8) Epoch 13, batch 6350, loss[loss=0.2248, simple_loss=0.3143, pruned_loss=0.06767, over 16932.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2971, pruned_loss=0.06454, over 3116656.75 frames. ], batch size: 96, lr: 5.22e-03, grad_scale: 4.0 2023-04-29 19:46:49,583 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4062, 3.3194, 3.3913, 3.5100, 3.5346, 3.2767, 3.5180, 3.5609], device='cuda:3'), covar=tensor([0.1117, 0.0930, 0.1017, 0.0583, 0.0631, 0.1957, 0.0889, 0.0822], device='cuda:3'), in_proj_covar=tensor([0.0538, 0.0671, 0.0800, 0.0683, 0.0518, 0.0528, 0.0547, 0.0629], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:47:13,675 INFO [optim.py:368] (3/8) 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 19:47:24,813 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:47:49,925 INFO [train.py:904] (3/8) Epoch 13, batch 6400, loss[loss=0.1856, simple_loss=0.2603, pruned_loss=0.05546, over 16384.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2994, pruned_loss=0.06733, over 3077937.64 frames. ], batch size: 35, lr: 5.22e-03, grad_scale: 8.0 2023-04-29 19:47:56,272 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:29,358 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:36,343 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 6450, loss[loss=0.2238, simple_loss=0.3217, pruned_loss=0.06292, over 16508.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2993, pruned_loss=0.06664, over 3072061.50 frames. ], batch size: 68, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:49:48,883 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 6500, loss[loss=0.2305, simple_loss=0.3045, pruned_loss=0.07819, over 16501.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2977, pruned_loss=0.06623, over 3082737.25 frames. ], batch size: 75, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:50:45,317 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:50:47,690 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 6550, loss[loss=0.2181, simple_loss=0.3181, pruned_loss=0.05905, over 16626.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.3001, pruned_loss=0.06687, over 3088945.71 frames. ], batch size: 68, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:52:01,586 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:06,728 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7479, 1.7433, 1.5376, 1.4491, 1.8634, 1.5453, 1.5819, 1.8625], device='cuda:3'), covar=tensor([0.0167, 0.0221, 0.0311, 0.0296, 0.0163, 0.0217, 0.0188, 0.0171], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0208, 0.0201, 0.0204, 0.0208, 0.0208, 0.0214, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:52:22,963 INFO [optim.py:368] (3/8) 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 19:52:24,054 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 19:52:24,982 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 6600, loss[loss=0.2222, simple_loss=0.313, pruned_loss=0.06565, over 16423.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3017, pruned_loss=0.06666, over 3106903.04 frames. ], batch size: 68, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:53:12,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6967, 3.7233, 2.1122, 4.2986, 2.6621, 4.1690, 2.3600, 2.8733], device='cuda:3'), covar=tensor([0.0202, 0.0308, 0.1625, 0.0118, 0.0768, 0.0462, 0.1401, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0166, 0.0189, 0.0136, 0.0167, 0.0206, 0.0196, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 19:53:13,948 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:53:48,848 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:05,481 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 19:54:14,487 INFO [train.py:904] (3/8) Epoch 13, batch 6650, loss[loss=0.2591, simple_loss=0.3259, pruned_loss=0.09617, over 11756.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3022, pruned_loss=0.06788, over 3103276.06 frames. ], batch size: 248, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:54:24,724 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9948, 2.8063, 2.6728, 2.0867, 2.5861, 2.2511, 2.6419, 3.0242], device='cuda:3'), covar=tensor([0.0384, 0.0677, 0.0540, 0.1523, 0.0786, 0.0775, 0.0781, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0149, 0.0160, 0.0145, 0.0138, 0.0126, 0.0138, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 19:54:47,990 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.058e+02 3.713e+02 4.648e+02 8.477e+02, threshold=7.425e+02, percent-clipped=4.0 2023-04-29 19:55:02,267 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:08,756 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-29 19:55:30,391 INFO [train.py:904] (3/8) Epoch 13, batch 6700, loss[loss=0.1974, simple_loss=0.2887, pruned_loss=0.05304, over 16881.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3018, pruned_loss=0.06812, over 3098738.16 frames. ], batch size: 90, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:55:36,462 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:57,188 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2199, 4.1917, 4.1026, 3.4509, 4.1439, 1.6621, 3.9262, 3.7518], device='cuda:3'), covar=tensor([0.0086, 0.0074, 0.0142, 0.0272, 0.0073, 0.2457, 0.0116, 0.0197], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0124, 0.0170, 0.0160, 0.0142, 0.0184, 0.0159, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 19:56:10,002 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:13,781 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 19:56:46,043 INFO [train.py:904] (3/8) Epoch 13, batch 6750, loss[loss=0.2244, simple_loss=0.3076, pruned_loss=0.0706, over 15380.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3013, pruned_loss=0.06847, over 3095783.78 frames. ], batch size: 191, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:56:49,431 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:55,866 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 19:57:22,973 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.101e+02 3.923e+02 4.752e+02 1.055e+03, threshold=7.847e+02, percent-clipped=2.0 2023-04-29 19:57:36,990 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 19:58:01,820 INFO [train.py:904] (3/8) Epoch 13, batch 6800, loss[loss=0.2021, simple_loss=0.3021, pruned_loss=0.05109, over 16852.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.301, pruned_loss=0.06796, over 3101599.66 frames. ], batch size: 96, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:58:54,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5449, 1.6216, 2.1173, 2.4576, 2.5043, 2.7464, 1.7898, 2.7284], device='cuda:3'), covar=tensor([0.0172, 0.0435, 0.0253, 0.0247, 0.0239, 0.0160, 0.0411, 0.0105], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0161, 0.0172, 0.0129, 0.0175, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 19:59:09,042 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-29 19:59:19,100 INFO [train.py:904] (3/8) Epoch 13, batch 6850, loss[loss=0.1972, simple_loss=0.2984, pruned_loss=0.04802, over 16789.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2997, pruned_loss=0.0665, over 3128849.37 frames. ], batch size: 83, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:59:54,880 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:00:00,842 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.804e+02 2.799e+02 3.352e+02 4.100e+02 8.801e+02, threshold=6.704e+02, percent-clipped=1.0 2023-04-29 20:00:34,560 INFO [train.py:904] (3/8) Epoch 13, batch 6900, loss[loss=0.2239, simple_loss=0.305, pruned_loss=0.07136, over 16656.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3024, pruned_loss=0.06624, over 3132503.91 frames. ], batch size: 134, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:00:35,012 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6689, 4.6976, 4.5204, 4.2195, 4.1333, 4.5930, 4.4359, 4.2720], device='cuda:3'), covar=tensor([0.0708, 0.0743, 0.0312, 0.0323, 0.0992, 0.0567, 0.0517, 0.0793], device='cuda:3'), in_proj_covar=tensor([0.0246, 0.0336, 0.0295, 0.0275, 0.0313, 0.0319, 0.0200, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:00:38,585 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-04-29 20:01:52,386 INFO [train.py:904] (3/8) Epoch 13, batch 6950, loss[loss=0.2755, simple_loss=0.3292, pruned_loss=0.1108, over 11186.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3048, pruned_loss=0.06889, over 3093254.79 frames. ], batch size: 246, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:02:18,898 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:02:22,272 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-29 20:02:36,165 INFO [optim.py:368] (3/8) 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 20:02:38,359 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:02:53,897 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7568, 5.0324, 5.3507, 5.0459, 5.0612, 5.7052, 5.1120, 4.8554], device='cuda:3'), covar=tensor([0.1044, 0.1893, 0.2223, 0.2256, 0.2653, 0.0954, 0.1693, 0.2535], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0519, 0.0563, 0.0439, 0.0600, 0.0587, 0.0453, 0.0597], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 20:02:55,426 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 20:03:08,751 INFO [train.py:904] (3/8) Epoch 13, batch 7000, loss[loss=0.2251, simple_loss=0.3193, pruned_loss=0.06543, over 16876.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3048, pruned_loss=0.06808, over 3102486.79 frames. ], batch size: 90, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:04:01,193 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8812, 2.0461, 2.3518, 3.1175, 2.1144, 2.2802, 2.2165, 2.1192], device='cuda:3'), covar=tensor([0.1079, 0.2959, 0.1999, 0.0610, 0.3762, 0.2093, 0.2746, 0.2993], device='cuda:3'), in_proj_covar=tensor([0.0365, 0.0400, 0.0333, 0.0313, 0.0415, 0.0460, 0.0367, 0.0467], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:04:07,265 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 7050, loss[loss=0.2251, simple_loss=0.3088, pruned_loss=0.07074, over 16935.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3053, pruned_loss=0.06742, over 3113028.44 frames. ], batch size: 109, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:05:02,583 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.845e+02 3.584e+02 4.333e+02 9.135e+02, threshold=7.167e+02, percent-clipped=3.0 2023-04-29 20:05:03,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7434, 1.7468, 1.5173, 1.5089, 1.8670, 1.6341, 1.6834, 1.8517], device='cuda:3'), covar=tensor([0.0113, 0.0197, 0.0285, 0.0257, 0.0135, 0.0190, 0.0130, 0.0134], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0211, 0.0204, 0.0205, 0.0208, 0.0208, 0.0214, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:05:05,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9480, 5.5720, 5.8192, 5.4973, 5.5836, 6.0968, 5.6136, 5.4074], device='cuda:3'), covar=tensor([0.0873, 0.1785, 0.1734, 0.1840, 0.2261, 0.0937, 0.1382, 0.2204], device='cuda:3'), in_proj_covar=tensor([0.0368, 0.0520, 0.0563, 0.0439, 0.0600, 0.0587, 0.0453, 0.0597], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 20:05:19,793 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:05:33,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3809, 3.4587, 2.0307, 3.8739, 2.5479, 3.8303, 2.0357, 2.7019], device='cuda:3'), covar=tensor([0.0274, 0.0370, 0.1647, 0.0155, 0.0810, 0.0456, 0.1634, 0.0763], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0165, 0.0188, 0.0136, 0.0167, 0.0206, 0.0194, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 20:05:38,297 INFO [train.py:904] (3/8) Epoch 13, batch 7100, loss[loss=0.194, simple_loss=0.2857, pruned_loss=0.05117, over 16743.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3037, pruned_loss=0.06706, over 3106523.05 frames. ], batch size: 124, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:06:56,805 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:06:57,457 INFO [train.py:904] (3/8) Epoch 13, batch 7150, loss[loss=0.2614, simple_loss=0.3257, pruned_loss=0.09855, over 11892.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3014, pruned_loss=0.0666, over 3106400.54 frames. ], batch size: 246, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:07:33,974 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 20:08:12,087 INFO [train.py:904] (3/8) Epoch 13, batch 7200, loss[loss=0.1895, simple_loss=0.2763, pruned_loss=0.05131, over 17225.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2994, pruned_loss=0.06557, over 3076961.06 frames. ], batch size: 45, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:08:45,129 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:09:29,929 INFO [train.py:904] (3/8) Epoch 13, batch 7250, loss[loss=0.2311, simple_loss=0.2941, pruned_loss=0.08403, over 11126.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2963, pruned_loss=0.0637, over 3081916.00 frames. ], batch size: 248, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:09:56,888 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:10:12,180 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 7300, loss[loss=0.2074, simple_loss=0.2963, pruned_loss=0.05922, over 16713.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2961, pruned_loss=0.06408, over 3080302.61 frames. ], batch size: 89, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:11:09,071 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:29,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2333, 3.2446, 1.8780, 3.6497, 2.3896, 3.6114, 1.9176, 2.5182], device='cuda:3'), covar=tensor([0.0250, 0.0304, 0.1626, 0.0128, 0.0874, 0.0371, 0.1646, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0165, 0.0190, 0.0136, 0.0169, 0.0207, 0.0197, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 20:11:40,200 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:52,952 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-29 20:12:02,352 INFO [train.py:904] (3/8) Epoch 13, batch 7350, loss[loss=0.2618, simple_loss=0.3265, pruned_loss=0.09859, over 11418.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2968, pruned_loss=0.06485, over 3072945.34 frames. ], batch size: 246, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:12:28,591 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 20:12:46,223 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 2.924e+02 3.417e+02 4.096e+02 1.600e+03, threshold=6.834e+02, percent-clipped=4.0 2023-04-29 20:13:21,035 INFO [train.py:904] (3/8) Epoch 13, batch 7400, loss[loss=0.2181, simple_loss=0.3052, pruned_loss=0.06556, over 16242.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2976, pruned_loss=0.06483, over 3083638.44 frames. ], batch size: 165, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:14:15,530 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2875, 4.2489, 4.1490, 3.2700, 4.1850, 1.5821, 3.9659, 3.8130], device='cuda:3'), covar=tensor([0.0124, 0.0095, 0.0182, 0.0456, 0.0111, 0.2888, 0.0153, 0.0289], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0122, 0.0167, 0.0159, 0.0140, 0.0182, 0.0157, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:14:32,917 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:14:41,690 INFO [train.py:904] (3/8) Epoch 13, batch 7450, loss[loss=0.2025, simple_loss=0.2866, pruned_loss=0.05921, over 16607.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2983, pruned_loss=0.06552, over 3083434.65 frames. ], batch size: 62, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:15:30,917 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.914e+02 3.121e+02 3.842e+02 4.432e+02 7.351e+02, threshold=7.685e+02, percent-clipped=1.0 2023-04-29 20:16:03,372 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 20:16:05,625 INFO [train.py:904] (3/8) Epoch 13, batch 7500, loss[loss=0.1795, simple_loss=0.2655, pruned_loss=0.04678, over 17199.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2985, pruned_loss=0.06542, over 3057622.95 frames. ], batch size: 45, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:16:19,398 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 20:17:24,559 INFO [train.py:904] (3/8) Epoch 13, batch 7550, loss[loss=0.2009, simple_loss=0.2821, pruned_loss=0.05986, over 16830.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.298, pruned_loss=0.06638, over 3037913.61 frames. ], batch size: 83, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:17:43,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4459, 5.7815, 5.4558, 5.4995, 5.2187, 5.0852, 5.2350, 5.8624], device='cuda:3'), covar=tensor([0.1014, 0.0767, 0.1060, 0.0765, 0.0790, 0.0767, 0.0988, 0.0816], device='cuda:3'), in_proj_covar=tensor([0.0568, 0.0702, 0.0576, 0.0500, 0.0442, 0.0455, 0.0587, 0.0538], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:17:58,119 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-29 20:18:07,707 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 7600, loss[loss=0.2162, simple_loss=0.2892, pruned_loss=0.07162, over 15392.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2974, pruned_loss=0.06698, over 3037877.39 frames. ], batch size: 191, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:18:59,708 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:19:37,562 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:19:41,501 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-29 20:19:43,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.46 vs. limit=5.0 2023-04-29 20:20:00,020 INFO [train.py:904] (3/8) Epoch 13, batch 7650, loss[loss=0.2256, simple_loss=0.3085, pruned_loss=0.07136, over 16195.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.298, pruned_loss=0.06723, over 3060360.43 frames. ], batch size: 165, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:20:12,383 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 20:20:35,805 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:20:44,160 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.211e+02 3.428e+02 4.160e+02 4.831e+02 1.059e+03, threshold=8.320e+02, percent-clipped=3.0 2023-04-29 20:20:53,101 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:21:06,958 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0140, 5.3116, 5.0838, 5.0552, 4.7941, 4.7572, 4.7605, 5.4174], device='cuda:3'), covar=tensor([0.0994, 0.0804, 0.0955, 0.0760, 0.0795, 0.0851, 0.1039, 0.0820], device='cuda:3'), in_proj_covar=tensor([0.0568, 0.0702, 0.0576, 0.0499, 0.0441, 0.0455, 0.0587, 0.0536], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:21:18,576 INFO [train.py:904] (3/8) Epoch 13, batch 7700, loss[loss=0.198, simple_loss=0.286, pruned_loss=0.05495, over 16774.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2982, pruned_loss=0.06713, over 3074673.25 frames. ], batch size: 39, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:21:54,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0510, 1.5197, 1.8274, 2.0655, 2.2083, 2.3183, 1.6031, 2.2069], device='cuda:3'), covar=tensor([0.0188, 0.0383, 0.0226, 0.0228, 0.0207, 0.0138, 0.0392, 0.0108], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0162, 0.0173, 0.0128, 0.0175, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 20:22:26,967 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:22:35,876 INFO [train.py:904] (3/8) Epoch 13, batch 7750, loss[loss=0.2355, simple_loss=0.321, pruned_loss=0.07498, over 16341.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2984, pruned_loss=0.06687, over 3074392.52 frames. ], batch size: 35, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:23:20,357 INFO [optim.py:368] (3/8) 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 20:23:40,229 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:23:52,176 INFO [train.py:904] (3/8) Epoch 13, batch 7800, loss[loss=0.1894, simple_loss=0.2837, pruned_loss=0.04755, over 16832.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3003, pruned_loss=0.0678, over 3084059.91 frames. ], batch size: 96, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:24:09,273 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:24:48,721 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8986, 5.3374, 5.5867, 5.2998, 5.3938, 5.9499, 5.4455, 5.2659], device='cuda:3'), covar=tensor([0.0940, 0.1810, 0.1913, 0.1975, 0.2306, 0.0858, 0.1306, 0.2117], device='cuda:3'), in_proj_covar=tensor([0.0366, 0.0519, 0.0567, 0.0441, 0.0597, 0.0584, 0.0450, 0.0593], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 20:25:09,849 INFO [train.py:904] (3/8) Epoch 13, batch 7850, loss[loss=0.2287, simple_loss=0.3108, pruned_loss=0.07332, over 15235.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3007, pruned_loss=0.06718, over 3096684.49 frames. ], batch size: 190, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:25:26,620 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1199, 5.6415, 5.9020, 5.5745, 5.5708, 6.2037, 5.7243, 5.4640], device='cuda:3'), covar=tensor([0.0788, 0.1941, 0.2049, 0.1945, 0.2461, 0.0963, 0.1503, 0.2472], device='cuda:3'), in_proj_covar=tensor([0.0366, 0.0518, 0.0566, 0.0441, 0.0597, 0.0584, 0.0450, 0.0593], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 20:25:40,335 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:25:52,268 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 3.020e+02 3.492e+02 4.287e+02 1.158e+03, threshold=6.983e+02, percent-clipped=4.0 2023-04-29 20:26:02,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8614, 5.1508, 4.9260, 4.9293, 4.6654, 4.6119, 4.6177, 5.2543], device='cuda:3'), covar=tensor([0.1004, 0.0807, 0.0935, 0.0754, 0.0748, 0.0865, 0.1023, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0570, 0.0703, 0.0580, 0.0502, 0.0442, 0.0456, 0.0588, 0.0538], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:26:22,707 INFO [train.py:904] (3/8) Epoch 13, batch 7900, loss[loss=0.21, simple_loss=0.3009, pruned_loss=0.05953, over 16877.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3001, pruned_loss=0.06672, over 3103025.63 frames. ], batch size: 116, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:27:36,767 INFO [train.py:904] (3/8) Epoch 13, batch 7950, loss[loss=0.2272, simple_loss=0.3075, pruned_loss=0.0735, over 16765.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3009, pruned_loss=0.06734, over 3101377.92 frames. ], batch size: 124, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:28:02,252 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:28:18,208 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.009e+02 2.840e+02 3.431e+02 4.053e+02 8.396e+02, threshold=6.863e+02, percent-clipped=2.0 2023-04-29 20:28:33,887 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 8000, loss[loss=0.1997, simple_loss=0.291, pruned_loss=0.05414, over 16849.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3021, pruned_loss=0.0683, over 3082758.40 frames. ], batch size: 42, lr: 5.19e-03, grad_scale: 8.0 2023-04-29 20:30:02,333 INFO [train.py:904] (3/8) Epoch 13, batch 8050, loss[loss=0.2093, simple_loss=0.2934, pruned_loss=0.06257, over 15361.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3011, pruned_loss=0.06758, over 3098728.62 frames. ], batch size: 190, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:30:02,896 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:30:25,159 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 20:30:37,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 20:30:45,831 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.353e+02 3.065e+02 3.748e+02 4.862e+02 1.232e+03, threshold=7.497e+02, percent-clipped=5.0 2023-04-29 20:30:53,724 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3271, 3.2052, 3.5882, 1.6836, 3.7631, 3.8096, 2.7424, 2.7187], device='cuda:3'), covar=tensor([0.0820, 0.0241, 0.0164, 0.1282, 0.0058, 0.0140, 0.0450, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0138, 0.0069, 0.0109, 0.0121, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 20:31:15,197 INFO [train.py:904] (3/8) Epoch 13, batch 8100, loss[loss=0.2341, simple_loss=0.2989, pruned_loss=0.08463, over 11411.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.3005, pruned_loss=0.0672, over 3097713.99 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:29,508 INFO [train.py:904] (3/8) Epoch 13, batch 8150, loss[loss=0.1736, simple_loss=0.2616, pruned_loss=0.04279, over 16862.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2984, pruned_loss=0.06666, over 3094146.40 frames. ], batch size: 96, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:53,000 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:14,448 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.878e+02 3.437e+02 4.165e+02 9.532e+02, threshold=6.873e+02, percent-clipped=2.0 2023-04-29 20:33:46,565 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 8200, loss[loss=0.2004, simple_loss=0.2923, pruned_loss=0.05431, over 15488.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2957, pruned_loss=0.06603, over 3100494.92 frames. ], batch size: 191, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:33:49,156 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:34:58,956 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0196, 3.9326, 3.8827, 3.1576, 3.8657, 1.6256, 3.6938, 3.5480], device='cuda:3'), covar=tensor([0.0098, 0.0097, 0.0168, 0.0353, 0.0116, 0.2689, 0.0138, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0123, 0.0169, 0.0159, 0.0141, 0.0184, 0.0157, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:35:09,258 INFO [train.py:904] (3/8) Epoch 13, batch 8250, loss[loss=0.1851, simple_loss=0.2711, pruned_loss=0.04959, over 12070.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2951, pruned_loss=0.06389, over 3083576.20 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:35:23,973 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:35:27,814 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:35:37,694 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:35:42,533 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 20:35:55,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5691, 3.6064, 3.3719, 3.1334, 3.1629, 3.5036, 3.3026, 3.3517], device='cuda:3'), covar=tensor([0.0517, 0.0517, 0.0243, 0.0211, 0.0502, 0.0399, 0.1248, 0.0472], device='cuda:3'), in_proj_covar=tensor([0.0245, 0.0337, 0.0295, 0.0273, 0.0308, 0.0316, 0.0201, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:35:57,080 INFO [optim.py:368] (3/8) 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 20:35:57,721 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7355, 1.2773, 1.6414, 1.6235, 1.8077, 1.8973, 1.5661, 1.8194], device='cuda:3'), covar=tensor([0.0204, 0.0312, 0.0164, 0.0207, 0.0217, 0.0129, 0.0295, 0.0093], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0173, 0.0156, 0.0159, 0.0171, 0.0127, 0.0173, 0.0117], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 20:36:29,984 INFO [train.py:904] (3/8) Epoch 13, batch 8300, loss[loss=0.1974, simple_loss=0.2892, pruned_loss=0.05275, over 16778.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2916, pruned_loss=0.06068, over 3056163.55 frames. ], batch size: 124, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:36:34,452 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 20:36:55,889 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:37:43,810 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:37:51,985 INFO [train.py:904] (3/8) Epoch 13, batch 8350, loss[loss=0.2299, simple_loss=0.3149, pruned_loss=0.07243, over 16880.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.291, pruned_loss=0.05877, over 3068867.15 frames. ], batch size: 116, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:38:39,963 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.345e+02 2.793e+02 3.218e+02 5.548e+02, threshold=5.587e+02, percent-clipped=0.0 2023-04-29 20:39:12,412 INFO [train.py:904] (3/8) Epoch 13, batch 8400, loss[loss=0.1762, simple_loss=0.2622, pruned_loss=0.04511, over 12447.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2876, pruned_loss=0.05623, over 3068448.92 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:29,224 INFO [train.py:904] (3/8) Epoch 13, batch 8450, loss[loss=0.1744, simple_loss=0.2738, pruned_loss=0.03744, over 16798.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2861, pruned_loss=0.05469, over 3060889.48 frames. ], batch size: 102, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:32,070 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3936, 3.0533, 2.7145, 2.1871, 2.1989, 2.1792, 2.9889, 2.8902], device='cuda:3'), covar=tensor([0.2434, 0.0744, 0.1429, 0.2448, 0.2531, 0.2014, 0.0470, 0.1171], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0252, 0.0283, 0.0280, 0.0274, 0.0225, 0.0267, 0.0297], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:40:55,917 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:41:09,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0365, 3.9093, 4.1174, 4.2444, 4.3687, 3.9749, 4.3199, 4.3762], device='cuda:3'), covar=tensor([0.1674, 0.1152, 0.1393, 0.0680, 0.0573, 0.1262, 0.0640, 0.0672], device='cuda:3'), in_proj_covar=tensor([0.0532, 0.0657, 0.0784, 0.0674, 0.0514, 0.0521, 0.0539, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:41:14,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4022, 4.6685, 4.5045, 4.4780, 4.1949, 4.1705, 4.2119, 4.7076], device='cuda:3'), covar=tensor([0.1025, 0.0868, 0.0967, 0.0697, 0.0780, 0.1332, 0.0860, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0558, 0.0689, 0.0569, 0.0494, 0.0435, 0.0449, 0.0579, 0.0528], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:41:17,432 INFO [optim.py:368] (3/8) 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 20:41:19,921 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5296, 3.6813, 3.6736, 2.7540, 3.3407, 3.6682, 3.5008, 2.1757], device='cuda:3'), covar=tensor([0.0318, 0.0038, 0.0036, 0.0251, 0.0078, 0.0075, 0.0061, 0.0376], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0068, 0.0071, 0.0126, 0.0082, 0.0091, 0.0080, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 20:41:20,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6071, 3.7623, 2.9394, 2.1131, 2.4516, 2.3527, 3.9870, 3.3324], device='cuda:3'), covar=tensor([0.2793, 0.0672, 0.1542, 0.2603, 0.2721, 0.1952, 0.0391, 0.1176], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0251, 0.0282, 0.0279, 0.0273, 0.0224, 0.0266, 0.0295], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:41:49,390 INFO [train.py:904] (3/8) Epoch 13, batch 8500, loss[loss=0.1631, simple_loss=0.2559, pruned_loss=0.03517, over 16715.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2822, pruned_loss=0.052, over 3071271.75 frames. ], batch size: 134, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:42:12,543 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:42:45,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3916, 3.7288, 3.7098, 2.5573, 3.4105, 3.7504, 3.5387, 2.1513], device='cuda:3'), covar=tensor([0.0404, 0.0033, 0.0031, 0.0307, 0.0068, 0.0060, 0.0053, 0.0391], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0068, 0.0070, 0.0125, 0.0082, 0.0090, 0.0080, 0.0119], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 20:42:47,071 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1173, 2.4222, 2.5460, 1.9103, 2.7254, 2.8204, 2.4916, 2.4559], device='cuda:3'), covar=tensor([0.0630, 0.0223, 0.0207, 0.0909, 0.0075, 0.0159, 0.0369, 0.0380], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0097, 0.0085, 0.0133, 0.0066, 0.0104, 0.0116, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 20:43:10,837 INFO [train.py:904] (3/8) Epoch 13, batch 8550, loss[loss=0.1845, simple_loss=0.2721, pruned_loss=0.04847, over 16683.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2799, pruned_loss=0.0511, over 3057311.50 frames. ], batch size: 57, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:43:19,677 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:22,070 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7085, 2.7511, 2.4732, 3.9581, 2.5457, 3.9030, 1.4999, 2.7526], device='cuda:3'), covar=tensor([0.1289, 0.0612, 0.1050, 0.0140, 0.0121, 0.0475, 0.1462, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0163, 0.0182, 0.0155, 0.0201, 0.0207, 0.0186, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 20:43:23,659 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:43:29,285 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-29 20:44:07,712 INFO [optim.py:368] (3/8) 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 20:44:10,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8965, 4.8223, 4.6866, 4.1874, 4.6912, 1.6456, 4.5041, 4.5165], device='cuda:3'), covar=tensor([0.0066, 0.0064, 0.0131, 0.0252, 0.0083, 0.2474, 0.0106, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0121, 0.0165, 0.0154, 0.0138, 0.0182, 0.0154, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:44:12,599 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:44:24,090 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1131, 1.9040, 2.0772, 3.7406, 1.9004, 2.2736, 2.0873, 2.0459], device='cuda:3'), covar=tensor([0.1102, 0.4095, 0.2728, 0.0470, 0.4731, 0.2775, 0.3615, 0.3909], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0396, 0.0330, 0.0309, 0.0409, 0.0451, 0.0360, 0.0459], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:44:50,596 INFO [train.py:904] (3/8) Epoch 13, batch 8600, loss[loss=0.1693, simple_loss=0.2678, pruned_loss=0.03542, over 16641.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2806, pruned_loss=0.05038, over 3060430.96 frames. ], batch size: 62, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:45:24,801 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 20:45:51,027 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:04,164 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:16,959 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:20,480 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 8650, loss[loss=0.1797, simple_loss=0.2714, pruned_loss=0.044, over 16806.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2783, pruned_loss=0.04863, over 3046619.26 frames. ], batch size: 124, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:47:40,992 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.301e+02 2.678e+02 3.280e+02 8.282e+02, threshold=5.356e+02, percent-clipped=3.0 2023-04-29 20:47:47,947 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6638, 2.4942, 2.1388, 3.4707, 1.7080, 3.6207, 1.5326, 2.6269], device='cuda:3'), covar=tensor([0.1526, 0.0783, 0.1337, 0.0221, 0.0094, 0.0373, 0.1776, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0162, 0.0182, 0.0154, 0.0198, 0.0205, 0.0185, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 20:47:59,465 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0141, 4.0768, 3.9072, 3.6672, 3.6413, 3.9927, 3.7058, 3.7801], device='cuda:3'), covar=tensor([0.0572, 0.0618, 0.0276, 0.0254, 0.0667, 0.0554, 0.0934, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0331, 0.0291, 0.0270, 0.0302, 0.0312, 0.0199, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 20:48:01,189 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:05,352 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:12,910 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 8700, loss[loss=0.1832, simple_loss=0.2779, pruned_loss=0.04422, over 15376.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2753, pruned_loss=0.04702, over 3043201.15 frames. ], batch size: 191, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:48:31,907 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5747, 3.6705, 2.1333, 4.0717, 2.7299, 4.0285, 2.1823, 2.8715], device='cuda:3'), covar=tensor([0.0231, 0.0292, 0.1511, 0.0162, 0.0736, 0.0427, 0.1518, 0.0704], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0158, 0.0183, 0.0130, 0.0161, 0.0198, 0.0190, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-29 20:49:34,327 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 20:49:46,651 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:49:54,310 INFO [train.py:904] (3/8) Epoch 13, batch 8750, loss[loss=0.1806, simple_loss=0.2788, pruned_loss=0.04119, over 16528.00 frames. ], tot_loss[loss=0.184, simple_loss=0.275, pruned_loss=0.04647, over 3043603.49 frames. ], batch size: 68, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:50:08,142 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 20:50:30,258 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:51:07,895 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.672e+02 2.544e+02 3.252e+02 4.020e+02 9.168e+02, threshold=6.504e+02, percent-clipped=9.0 2023-04-29 20:51:48,202 INFO [train.py:904] (3/8) Epoch 13, batch 8800, loss[loss=0.1863, simple_loss=0.2772, pruned_loss=0.04772, over 15609.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2734, pruned_loss=0.04485, over 3069431.08 frames. ], batch size: 194, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:52:02,478 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:52:39,117 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 8850, loss[loss=0.1536, simple_loss=0.2457, pruned_loss=0.03074, over 12373.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2754, pruned_loss=0.04419, over 3056906.11 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:53:41,391 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:45,816 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:54:37,528 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 8900, loss[loss=0.1639, simple_loss=0.2567, pruned_loss=0.03556, over 16287.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2754, pruned_loss=0.04369, over 3046394.37 frames. ], batch size: 35, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:55:22,704 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:26,793 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:56:54,227 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:57:21,721 INFO [train.py:904] (3/8) Epoch 13, batch 8950, loss[loss=0.1595, simple_loss=0.2531, pruned_loss=0.03295, over 16376.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2751, pruned_loss=0.04419, over 3046522.33 frames. ], batch size: 146, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:58:29,463 INFO [optim.py:368] (3/8) 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 20:58:41,702 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:58:49,742 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6934, 3.2155, 2.7112, 5.0957, 3.7223, 4.4522, 1.8026, 3.1641], device='cuda:3'), covar=tensor([0.1524, 0.0698, 0.1229, 0.0109, 0.0245, 0.0345, 0.1587, 0.0788], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0162, 0.0183, 0.0153, 0.0195, 0.0205, 0.0186, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 20:58:54,556 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 9000, loss[loss=0.1618, simple_loss=0.2536, pruned_loss=0.03497, over 16664.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2717, pruned_loss=0.04301, over 3034082.06 frames. ], batch size: 134, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:59:11,360 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 20:59:22,056 INFO [train.py:938] (3/8) Epoch 13, validation: loss=0.1517, simple_loss=0.2561, pruned_loss=0.02371, over 944034.00 frames. 2023-04-29 20:59:22,056 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 21:01:06,010 INFO [train.py:904] (3/8) Epoch 13, batch 9050, loss[loss=0.1859, simple_loss=0.2718, pruned_loss=0.05001, over 16687.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2727, pruned_loss=0.04358, over 3065555.99 frames. ], batch size: 134, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:07,080 INFO [optim.py:368] (3/8) 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 21:02:34,844 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 21:02:46,082 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 21:02:52,476 INFO [train.py:904] (3/8) Epoch 13, batch 9100, loss[loss=0.1883, simple_loss=0.2811, pruned_loss=0.04774, over 16614.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2724, pruned_loss=0.04393, over 3062005.29 frames. ], batch size: 134, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:58,111 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:03:33,445 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 9150, loss[loss=0.1822, simple_loss=0.2702, pruned_loss=0.0471, over 12319.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2735, pruned_loss=0.04349, over 3082589.86 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:05:29,362 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0435, 3.4163, 3.5258, 3.4911, 3.5005, 3.3794, 3.1701, 3.4460], device='cuda:3'), covar=tensor([0.0717, 0.0784, 0.0686, 0.0809, 0.0700, 0.0711, 0.1182, 0.0541], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0347, 0.0353, 0.0338, 0.0395, 0.0375, 0.0458, 0.0298], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 21:05:44,668 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-29 21:05:52,920 INFO [optim.py:368] (3/8) 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 21:06:25,381 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 21:06:31,913 INFO [train.py:904] (3/8) Epoch 13, batch 9200, loss[loss=0.1459, simple_loss=0.2298, pruned_loss=0.03101, over 11901.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2684, pruned_loss=0.04193, over 3081722.70 frames. ], batch size: 247, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:07:43,509 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:07:56,466 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3967, 2.9749, 2.7278, 2.1577, 2.1739, 2.2064, 3.0640, 2.8900], device='cuda:3'), covar=tensor([0.2263, 0.0724, 0.1361, 0.2302, 0.2229, 0.1845, 0.0411, 0.1145], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0250, 0.0280, 0.0276, 0.0262, 0.0222, 0.0263, 0.0290], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:08:07,048 INFO [train.py:904] (3/8) Epoch 13, batch 9250, loss[loss=0.1804, simple_loss=0.2773, pruned_loss=0.04179, over 15238.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2683, pruned_loss=0.04209, over 3075284.25 frames. ], batch size: 190, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:08:23,932 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 21:09:12,849 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 2.375e+02 2.887e+02 3.471e+02 9.563e+02, threshold=5.774e+02, percent-clipped=3.0 2023-04-29 21:09:24,455 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:24,536 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:39,403 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:55,048 INFO [train.py:904] (3/8) Epoch 13, batch 9300, loss[loss=0.1726, simple_loss=0.2649, pruned_loss=0.0401, over 15289.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2663, pruned_loss=0.04173, over 3034830.70 frames. ], batch size: 192, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:11:11,074 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:23,562 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:40,549 INFO [train.py:904] (3/8) Epoch 13, batch 9350, loss[loss=0.1808, simple_loss=0.2643, pruned_loss=0.04863, over 12468.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2663, pruned_loss=0.04161, over 3034127.99 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:12:25,655 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131174.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:12:41,634 INFO [optim.py:368] (3/8) 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 21:12:44,779 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 21:13:09,782 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9790, 4.2393, 4.0714, 4.0983, 3.7078, 3.8311, 3.8771, 4.2169], device='cuda:3'), covar=tensor([0.0929, 0.0900, 0.0935, 0.0772, 0.0787, 0.1698, 0.0843, 0.0968], device='cuda:3'), in_proj_covar=tensor([0.0543, 0.0679, 0.0549, 0.0482, 0.0426, 0.0439, 0.0564, 0.0516], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:13:20,691 INFO [train.py:904] (3/8) Epoch 13, batch 9400, loss[loss=0.1615, simple_loss=0.2484, pruned_loss=0.0373, over 12681.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2664, pruned_loss=0.04111, over 3053047.12 frames. ], batch size: 248, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:13:25,652 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:13:48,252 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 21:13:59,504 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:14:27,987 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:14:32,338 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 21:14:59,708 INFO [train.py:904] (3/8) Epoch 13, batch 9450, loss[loss=0.1496, simple_loss=0.2495, pruned_loss=0.02487, over 16883.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2681, pruned_loss=0.04165, over 3038609.72 frames. ], batch size: 102, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:15:00,864 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:15:43,432 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7291, 3.7733, 4.1244, 4.0892, 4.0762, 3.8609, 3.8767, 3.8877], device='cuda:3'), covar=tensor([0.0413, 0.0828, 0.0482, 0.0518, 0.0617, 0.0502, 0.0751, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0341, 0.0344, 0.0331, 0.0388, 0.0368, 0.0446, 0.0293], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-29 21:16:03,420 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 9500, loss[loss=0.1692, simple_loss=0.2555, pruned_loss=0.04148, over 12661.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2671, pruned_loss=0.04103, over 3056113.75 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:17:19,353 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 21:18:25,268 INFO [train.py:904] (3/8) Epoch 13, batch 9550, loss[loss=0.1625, simple_loss=0.2499, pruned_loss=0.03751, over 12647.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2668, pruned_loss=0.04107, over 3066297.55 frames. ], batch size: 246, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:19:29,575 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.317e+02 2.711e+02 3.253e+02 6.482e+02, threshold=5.422e+02, percent-clipped=4.0 2023-04-29 21:20:03,674 INFO [train.py:904] (3/8) Epoch 13, batch 9600, loss[loss=0.1804, simple_loss=0.2721, pruned_loss=0.04432, over 16956.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2684, pruned_loss=0.04203, over 3079211.78 frames. ], batch size: 109, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:20:17,991 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:20:29,751 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 9650, loss[loss=0.1833, simple_loss=0.2634, pruned_loss=0.05156, over 12163.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2703, pruned_loss=0.04249, over 3075614.97 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:22:10,995 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3229, 4.6355, 4.4696, 4.4679, 4.0938, 4.0731, 4.1471, 4.6899], device='cuda:3'), covar=tensor([0.1094, 0.1010, 0.1065, 0.0710, 0.0770, 0.1506, 0.1109, 0.0968], device='cuda:3'), in_proj_covar=tensor([0.0549, 0.0688, 0.0552, 0.0486, 0.0429, 0.0441, 0.0570, 0.0523], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:22:34,310 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:22:47,912 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:23:16,269 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 21:23:18,256 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 9700, loss[loss=0.1804, simple_loss=0.2774, pruned_loss=0.04175, over 16819.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2698, pruned_loss=0.04233, over 3080891.45 frames. ], batch size: 102, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:24:36,254 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 13, batch 9750, loss[loss=0.166, simple_loss=0.2463, pruned_loss=0.04285, over 12181.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2687, pruned_loss=0.04252, over 3059082.89 frames. ], batch size: 247, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:25:22,343 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:26:22,846 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 9800, loss[loss=0.1625, simple_loss=0.2658, pruned_loss=0.02957, over 16360.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2692, pruned_loss=0.04162, over 3078381.34 frames. ], batch size: 68, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:27:41,011 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-04-29 21:28:15,709 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5478, 2.0516, 1.7823, 1.7821, 2.3043, 2.0289, 2.1864, 2.4377], device='cuda:3'), covar=tensor([0.0111, 0.0346, 0.0424, 0.0408, 0.0239, 0.0329, 0.0151, 0.0220], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0207, 0.0202, 0.0201, 0.0205, 0.0205, 0.0202, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:28:39,800 INFO [train.py:904] (3/8) Epoch 13, batch 9850, loss[loss=0.1922, simple_loss=0.2835, pruned_loss=0.05041, over 16903.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2702, pruned_loss=0.04116, over 3081762.53 frames. ], batch size: 109, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:29:32,973 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7927, 1.2314, 1.6579, 1.7048, 1.8004, 1.8921, 1.5751, 1.8280], device='cuda:3'), covar=tensor([0.0200, 0.0358, 0.0171, 0.0227, 0.0227, 0.0152, 0.0374, 0.0102], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0172, 0.0154, 0.0156, 0.0169, 0.0125, 0.0173, 0.0116], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 21:29:46,604 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 13, batch 9900, loss[loss=0.1764, simple_loss=0.2598, pruned_loss=0.04647, over 12510.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2703, pruned_loss=0.04098, over 3069320.41 frames. ], batch size: 250, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:32:30,669 INFO [train.py:904] (3/8) Epoch 13, batch 9950, loss[loss=0.1826, simple_loss=0.2824, pruned_loss=0.04147, over 15484.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.272, pruned_loss=0.0416, over 3052013.36 frames. ], batch size: 192, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:33:02,011 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:17,881 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 21:33:20,073 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:42,483 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8991, 3.0929, 3.1025, 2.1985, 2.9605, 3.2087, 3.1632, 1.9191], device='cuda:3'), covar=tensor([0.0479, 0.0041, 0.0043, 0.0304, 0.0079, 0.0064, 0.0051, 0.0423], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0068, 0.0070, 0.0126, 0.0082, 0.0089, 0.0079, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 21:33:47,201 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.347e+02 2.831e+02 3.741e+02 5.862e+02, threshold=5.662e+02, percent-clipped=1.0 2023-04-29 21:34:31,155 INFO [train.py:904] (3/8) Epoch 13, batch 10000, loss[loss=0.1737, simple_loss=0.276, pruned_loss=0.03574, over 15350.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2706, pruned_loss=0.0409, over 3075741.20 frames. ], batch size: 192, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:35:27,631 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:06,299 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:11,639 INFO [train.py:904] (3/8) Epoch 13, batch 10050, loss[loss=0.1746, simple_loss=0.2704, pruned_loss=0.03942, over 16622.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2711, pruned_loss=0.04121, over 3089486.84 frames. ], batch size: 89, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:36:23,305 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:04,199 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:14,088 INFO [optim.py:368] (3/8) 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 21:37:18,463 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 21:37:45,821 INFO [train.py:904] (3/8) Epoch 13, batch 10100, loss[loss=0.1677, simple_loss=0.2625, pruned_loss=0.03649, over 16845.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2712, pruned_loss=0.04137, over 3092847.97 frames. ], batch size: 90, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:38:16,889 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:39:29,407 INFO [train.py:904] (3/8) Epoch 14, batch 0, loss[loss=0.1828, simple_loss=0.2607, pruned_loss=0.05246, over 15895.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2607, pruned_loss=0.05246, over 15895.00 frames. ], batch size: 35, lr: 4.96e-03, grad_scale: 8.0 2023-04-29 21:39:29,407 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 21:39:36,900 INFO [train.py:938] (3/8) Epoch 14, validation: loss=0.1515, simple_loss=0.2551, pruned_loss=0.024, over 944034.00 frames. 2023-04-29 21:39:36,901 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 21:40:19,402 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2728, 3.4971, 3.3758, 2.1996, 2.9746, 2.3960, 3.7503, 3.6754], device='cuda:3'), covar=tensor([0.0246, 0.0767, 0.0637, 0.1758, 0.0789, 0.0950, 0.0493, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0141, 0.0157, 0.0145, 0.0136, 0.0125, 0.0135, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 21:40:22,370 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 50, loss[loss=0.1782, simple_loss=0.2745, pruned_loss=0.04099, over 17235.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2825, pruned_loss=0.06082, over 753294.78 frames. ], batch size: 52, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:27,124 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 100, loss[loss=0.1906, simple_loss=0.2738, pruned_loss=0.05366, over 16713.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2765, pruned_loss=0.05709, over 1323939.41 frames. ], batch size: 89, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:58,408 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:13,333 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:21,198 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 21:42:23,911 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:44,827 INFO [optim.py:368] (3/8) 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 21:42:51,054 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 150, loss[loss=0.172, simple_loss=0.268, pruned_loss=0.03801, over 17046.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2749, pruned_loss=0.05459, over 1756182.02 frames. ], batch size: 53, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:43:04,670 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2692, 4.2763, 4.7257, 4.7149, 4.7327, 4.3701, 4.4414, 4.2896], device='cuda:3'), covar=tensor([0.0341, 0.0555, 0.0393, 0.0431, 0.0479, 0.0412, 0.0786, 0.0503], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0352, 0.0353, 0.0340, 0.0397, 0.0380, 0.0464, 0.0302], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 21:43:20,790 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:24,005 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:43:30,561 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:31,337 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 21:43:51,824 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 21:44:08,561 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:44:15,117 INFO [train.py:904] (3/8) Epoch 14, batch 200, loss[loss=0.1685, simple_loss=0.2544, pruned_loss=0.04134, over 16991.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2742, pruned_loss=0.0542, over 2106923.31 frames. ], batch size: 41, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:44:49,259 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5509, 2.5294, 2.1079, 2.3740, 2.8872, 2.7322, 3.4194, 3.1501], device='cuda:3'), covar=tensor([0.0109, 0.0359, 0.0457, 0.0372, 0.0231, 0.0329, 0.0180, 0.0208], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0216, 0.0209, 0.0207, 0.0213, 0.0213, 0.0215, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:45:02,144 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.381e+02 2.781e+02 3.514e+02 5.258e+02, threshold=5.562e+02, percent-clipped=0.0 2023-04-29 21:45:15,516 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:45:23,095 INFO [train.py:904] (3/8) Epoch 14, batch 250, loss[loss=0.1789, simple_loss=0.2664, pruned_loss=0.04572, over 17064.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2717, pruned_loss=0.05226, over 2389662.09 frames. ], batch size: 55, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:45:38,785 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:46:33,336 INFO [train.py:904] (3/8) Epoch 14, batch 300, loss[loss=0.1399, simple_loss=0.2237, pruned_loss=0.02802, over 16775.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2687, pruned_loss=0.05093, over 2596548.24 frames. ], batch size: 39, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:47:16,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4423, 2.4798, 1.9862, 2.2815, 2.9341, 2.6115, 3.2210, 3.1359], device='cuda:3'), covar=tensor([0.0138, 0.0384, 0.0554, 0.0431, 0.0233, 0.0337, 0.0283, 0.0235], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0216, 0.0209, 0.0208, 0.0214, 0.0215, 0.0216, 0.0203], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:47:22,675 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.346e+02 2.777e+02 3.394e+02 6.651e+02, threshold=5.554e+02, percent-clipped=2.0 2023-04-29 21:47:43,453 INFO [train.py:904] (3/8) Epoch 14, batch 350, loss[loss=0.195, simple_loss=0.2695, pruned_loss=0.06023, over 16200.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2665, pruned_loss=0.05001, over 2764209.78 frames. ], batch size: 165, lr: 4.95e-03, grad_scale: 1.0 2023-04-29 21:48:00,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1048, 2.0650, 2.5666, 3.0951, 2.8179, 3.3984, 2.4014, 3.4016], device='cuda:3'), covar=tensor([0.0185, 0.0371, 0.0255, 0.0204, 0.0246, 0.0155, 0.0315, 0.0141], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0176, 0.0159, 0.0162, 0.0173, 0.0129, 0.0177, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-29 21:48:51,145 INFO [train.py:904] (3/8) Epoch 14, batch 400, loss[loss=0.1642, simple_loss=0.2418, pruned_loss=0.04329, over 16811.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2645, pruned_loss=0.04985, over 2883537.42 frames. ], batch size: 39, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:49:38,426 INFO [optim.py:368] (3/8) 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 21:49:39,469 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 450, loss[loss=0.212, simple_loss=0.3033, pruned_loss=0.0603, over 17052.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2629, pruned_loss=0.04925, over 2986232.06 frames. ], batch size: 53, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:50:12,003 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 21:50:50,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6948, 3.7315, 2.9018, 2.1616, 2.4985, 2.2751, 3.8629, 3.3784], device='cuda:3'), covar=tensor([0.2442, 0.0663, 0.1504, 0.2629, 0.2466, 0.1912, 0.0458, 0.1164], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0256, 0.0288, 0.0283, 0.0272, 0.0228, 0.0270, 0.0302], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:51:11,603 INFO [train.py:904] (3/8) Epoch 14, batch 500, loss[loss=0.1649, simple_loss=0.2597, pruned_loss=0.03505, over 17096.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2618, pruned_loss=0.04873, over 3058461.75 frames. ], batch size: 47, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:51:20,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5397, 2.2640, 2.3892, 4.3164, 2.2807, 2.6953, 2.3065, 2.5237], device='cuda:3'), covar=tensor([0.1078, 0.3327, 0.2501, 0.0476, 0.3783, 0.2225, 0.3329, 0.3085], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0406, 0.0341, 0.0320, 0.0419, 0.0464, 0.0371, 0.0474], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:51:35,531 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7004, 2.6667, 2.2470, 2.6815, 2.9967, 2.8948, 3.5174, 3.2385], device='cuda:3'), covar=tensor([0.0104, 0.0346, 0.0430, 0.0321, 0.0224, 0.0290, 0.0182, 0.0218], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0216, 0.0211, 0.0209, 0.0215, 0.0215, 0.0218, 0.0205], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:51:58,582 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 550, loss[loss=0.1899, simple_loss=0.2862, pruned_loss=0.04677, over 17260.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2615, pruned_loss=0.04828, over 3115912.96 frames. ], batch size: 52, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:52:34,275 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:52:48,596 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9332, 4.5952, 4.9291, 5.1044, 5.3139, 4.7138, 5.3055, 5.2514], device='cuda:3'), covar=tensor([0.1442, 0.1223, 0.1572, 0.0675, 0.0469, 0.0794, 0.0487, 0.0609], device='cuda:3'), in_proj_covar=tensor([0.0576, 0.0704, 0.0848, 0.0723, 0.0548, 0.0552, 0.0572, 0.0667], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:52:51,095 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5107, 2.8354, 2.9421, 1.9775, 2.6836, 2.1802, 3.1041, 3.0641], device='cuda:3'), covar=tensor([0.0248, 0.0802, 0.0583, 0.1754, 0.0765, 0.0899, 0.0563, 0.0904], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0144, 0.0157, 0.0144, 0.0136, 0.0124, 0.0134, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 21:53:14,758 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1964, 5.8761, 6.0211, 5.7398, 5.8634, 6.3273, 5.9292, 5.6353], device='cuda:3'), covar=tensor([0.0882, 0.1729, 0.2020, 0.1959, 0.2380, 0.0826, 0.1396, 0.2493], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0522, 0.0574, 0.0444, 0.0603, 0.0597, 0.0453, 0.0592], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 21:53:28,149 INFO [train.py:904] (3/8) Epoch 14, batch 600, loss[loss=0.1631, simple_loss=0.2378, pruned_loss=0.04415, over 15940.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2609, pruned_loss=0.04826, over 3166206.64 frames. ], batch size: 35, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:53:41,853 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:54:14,704 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3254, 4.4449, 4.7387, 4.7803, 4.8244, 4.5352, 4.3802, 4.3191], device='cuda:3'), covar=tensor([0.0557, 0.0844, 0.0610, 0.0567, 0.0643, 0.0571, 0.1150, 0.0627], device='cuda:3'), in_proj_covar=tensor([0.0358, 0.0374, 0.0374, 0.0358, 0.0417, 0.0400, 0.0492, 0.0319], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 21:54:17,819 INFO [optim.py:368] (3/8) 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 21:54:39,144 INFO [train.py:904] (3/8) Epoch 14, batch 650, loss[loss=0.1779, simple_loss=0.2679, pruned_loss=0.0439, over 17100.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2599, pruned_loss=0.04836, over 3193660.87 frames. ], batch size: 53, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:54:39,597 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8408, 3.9921, 2.3070, 4.5356, 3.0360, 4.5772, 2.5977, 3.2373], device='cuda:3'), covar=tensor([0.0246, 0.0323, 0.1523, 0.0267, 0.0750, 0.0378, 0.1352, 0.0653], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0166, 0.0190, 0.0141, 0.0169, 0.0207, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 21:54:57,995 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9112, 4.1880, 4.0437, 4.0696, 3.7256, 3.8299, 3.8243, 4.1861], device='cuda:3'), covar=tensor([0.1084, 0.1019, 0.0990, 0.0708, 0.0715, 0.1432, 0.0930, 0.1021], device='cuda:3'), in_proj_covar=tensor([0.0593, 0.0741, 0.0603, 0.0527, 0.0472, 0.0472, 0.0621, 0.0565], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:55:49,508 INFO [train.py:904] (3/8) Epoch 14, batch 700, loss[loss=0.2189, simple_loss=0.2894, pruned_loss=0.07419, over 15480.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2603, pruned_loss=0.04823, over 3208741.01 frames. ], batch size: 190, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:56:01,318 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:38,038 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.258e+02 2.690e+02 3.168e+02 6.172e+02, threshold=5.381e+02, percent-clipped=1.0 2023-04-29 21:56:39,020 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 750, loss[loss=0.169, simple_loss=0.2518, pruned_loss=0.04313, over 16724.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2601, pruned_loss=0.04782, over 3234341.80 frames. ], batch size: 89, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:57:05,554 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 21:57:09,268 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:57:24,376 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:43,135 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:59,778 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:07,424 INFO [train.py:904] (3/8) Epoch 14, batch 800, loss[loss=0.2192, simple_loss=0.2843, pruned_loss=0.07706, over 16364.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2596, pruned_loss=0.04789, over 3239066.56 frames. ], batch size: 165, lr: 4.95e-03, grad_scale: 4.0 2023-04-29 21:58:16,055 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:18,702 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.42 vs. limit=5.0 2023-04-29 21:58:54,909 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.382e+02 2.716e+02 3.142e+02 6.467e+02, threshold=5.432e+02, percent-clipped=2.0 2023-04-29 21:59:08,473 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7018, 5.0609, 4.8108, 4.8660, 4.5486, 4.5269, 4.5005, 5.1437], device='cuda:3'), covar=tensor([0.1188, 0.0932, 0.1085, 0.0703, 0.0868, 0.1175, 0.1203, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0596, 0.0744, 0.0603, 0.0529, 0.0474, 0.0474, 0.0623, 0.0568], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 21:59:15,848 INFO [train.py:904] (3/8) Epoch 14, batch 850, loss[loss=0.153, simple_loss=0.2402, pruned_loss=0.03292, over 16982.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2584, pruned_loss=0.04683, over 3260628.90 frames. ], batch size: 41, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 21:59:23,489 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:59:58,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1932, 2.1266, 1.7519, 1.9325, 2.4097, 2.2341, 2.3270, 2.5244], device='cuda:3'), covar=tensor([0.0243, 0.0314, 0.0421, 0.0375, 0.0192, 0.0266, 0.0206, 0.0229], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0217, 0.0211, 0.0210, 0.0217, 0.0218, 0.0222, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:00:24,478 INFO [train.py:904] (3/8) Epoch 14, batch 900, loss[loss=0.1664, simple_loss=0.2511, pruned_loss=0.04086, over 16942.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2577, pruned_loss=0.04604, over 3283555.63 frames. ], batch size: 41, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:01:14,368 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.740e+02 2.299e+02 2.623e+02 3.116e+02 6.405e+02, threshold=5.245e+02, percent-clipped=2.0 2023-04-29 22:01:15,044 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 22:01:34,126 INFO [train.py:904] (3/8) Epoch 14, batch 950, loss[loss=0.1763, simple_loss=0.2733, pruned_loss=0.03968, over 17064.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2582, pruned_loss=0.04639, over 3290543.49 frames. ], batch size: 50, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:01:52,518 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-29 22:02:36,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0917, 4.5784, 4.5374, 3.2846, 3.9265, 4.5240, 4.1600, 2.7122], device='cuda:3'), covar=tensor([0.0369, 0.0044, 0.0028, 0.0291, 0.0085, 0.0060, 0.0059, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0129, 0.0085, 0.0094, 0.0083, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:02:41,755 INFO [train.py:904] (3/8) Epoch 14, batch 1000, loss[loss=0.1495, simple_loss=0.2338, pruned_loss=0.03265, over 17215.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2572, pruned_loss=0.04571, over 3302049.79 frames. ], batch size: 44, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:03:29,371 INFO [optim.py:368] (3/8) 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 22:03:49,998 INFO [train.py:904] (3/8) Epoch 14, batch 1050, loss[loss=0.1809, simple_loss=0.2724, pruned_loss=0.04469, over 16770.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2575, pruned_loss=0.04592, over 3309259.14 frames. ], batch size: 62, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:04:10,798 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 1100, loss[loss=0.1824, simple_loss=0.2662, pruned_loss=0.0493, over 16502.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2566, pruned_loss=0.04542, over 3311198.12 frames. ], batch size: 68, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:05:39,978 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2771, 3.5947, 3.4408, 2.2276, 2.8777, 2.5285, 3.6016, 3.7702], device='cuda:3'), covar=tensor([0.0329, 0.0822, 0.0620, 0.1776, 0.0922, 0.0894, 0.0752, 0.0951], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0147, 0.0157, 0.0145, 0.0137, 0.0125, 0.0135, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:05:47,408 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.639e+02 2.252e+02 2.587e+02 3.047e+02 6.829e+02, threshold=5.174e+02, percent-clipped=2.0 2023-04-29 22:06:05,375 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9895, 4.1266, 2.4733, 4.7002, 3.1156, 4.6855, 2.9400, 3.5997], device='cuda:3'), covar=tensor([0.0238, 0.0328, 0.1403, 0.0176, 0.0746, 0.0373, 0.1174, 0.0503], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0170, 0.0192, 0.0145, 0.0170, 0.0211, 0.0200, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:06:08,348 INFO [train.py:904] (3/8) Epoch 14, batch 1150, loss[loss=0.1661, simple_loss=0.2415, pruned_loss=0.04537, over 16931.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2556, pruned_loss=0.04505, over 3316487.50 frames. ], batch size: 96, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:06:08,669 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:06:55,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0142, 5.4155, 5.1815, 5.1754, 4.8823, 4.7836, 4.8106, 5.5037], device='cuda:3'), covar=tensor([0.1208, 0.0968, 0.1062, 0.0823, 0.0868, 0.0945, 0.1112, 0.1026], device='cuda:3'), in_proj_covar=tensor([0.0600, 0.0749, 0.0608, 0.0535, 0.0477, 0.0480, 0.0628, 0.0573], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:07:06,235 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5969, 3.0277, 2.6790, 5.1028, 4.5186, 4.4021, 1.4806, 3.2688], device='cuda:3'), covar=tensor([0.1327, 0.0596, 0.1090, 0.0111, 0.0189, 0.0371, 0.1464, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0163, 0.0183, 0.0160, 0.0196, 0.0209, 0.0187, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:07:16,495 INFO [train.py:904] (3/8) Epoch 14, batch 1200, loss[loss=0.1871, simple_loss=0.258, pruned_loss=0.05805, over 12216.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2553, pruned_loss=0.04491, over 3321084.65 frames. ], batch size: 246, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:08:05,977 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 1250, loss[loss=0.1702, simple_loss=0.2443, pruned_loss=0.04806, over 16664.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2559, pruned_loss=0.04524, over 3318356.42 frames. ], batch size: 134, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:09:33,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0017, 2.1669, 2.4685, 2.8737, 2.7206, 3.2637, 2.3422, 3.1687], device='cuda:3'), covar=tensor([0.0182, 0.0366, 0.0266, 0.0240, 0.0248, 0.0157, 0.0349, 0.0168], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0178, 0.0163, 0.0166, 0.0175, 0.0133, 0.0178, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:09:37,985 INFO [train.py:904] (3/8) Epoch 14, batch 1300, loss[loss=0.1886, simple_loss=0.2624, pruned_loss=0.05744, over 16699.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2554, pruned_loss=0.04526, over 3328401.42 frames. ], batch size: 134, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:10:27,126 INFO [optim.py:368] (3/8) 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 22:10:48,347 INFO [train.py:904] (3/8) Epoch 14, batch 1350, loss[loss=0.153, simple_loss=0.2449, pruned_loss=0.03051, over 17227.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2553, pruned_loss=0.0447, over 3313409.52 frames. ], batch size: 45, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:11:06,800 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:11:29,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9271, 4.6351, 4.9021, 5.1386, 5.3194, 4.6203, 5.2849, 5.2532], device='cuda:3'), covar=tensor([0.1599, 0.1264, 0.1652, 0.0675, 0.0498, 0.0920, 0.0544, 0.0599], device='cuda:3'), in_proj_covar=tensor([0.0581, 0.0717, 0.0861, 0.0735, 0.0552, 0.0561, 0.0577, 0.0676], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:11:33,319 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7820, 3.8340, 4.0921, 4.1029, 4.1061, 3.8633, 3.8931, 3.8378], device='cuda:3'), covar=tensor([0.0372, 0.0718, 0.0462, 0.0431, 0.0494, 0.0499, 0.0807, 0.0565], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0389, 0.0388, 0.0370, 0.0431, 0.0415, 0.0506, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 22:11:56,451 INFO [train.py:904] (3/8) Epoch 14, batch 1400, loss[loss=0.1802, simple_loss=0.2716, pruned_loss=0.04441, over 17133.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2564, pruned_loss=0.04494, over 3319403.31 frames. ], batch size: 49, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:12:13,482 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:26,521 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:12:29,451 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:13:06,263 INFO [train.py:904] (3/8) Epoch 14, batch 1450, loss[loss=0.1819, simple_loss=0.2715, pruned_loss=0.0462, over 17077.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2549, pruned_loss=0.04467, over 3314419.48 frames. ], batch size: 53, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:13:06,531 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:13:17,358 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8041, 4.4234, 4.3728, 3.1916, 3.8518, 4.3075, 4.0983, 2.4496], device='cuda:3'), covar=tensor([0.0441, 0.0048, 0.0037, 0.0306, 0.0092, 0.0087, 0.0065, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0072, 0.0073, 0.0128, 0.0084, 0.0094, 0.0082, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:13:52,473 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:13:55,207 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:12,864 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:14,808 INFO [train.py:904] (3/8) Epoch 14, batch 1500, loss[loss=0.1856, simple_loss=0.2651, pruned_loss=0.053, over 17009.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2545, pruned_loss=0.04502, over 3321311.70 frames. ], batch size: 41, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:14:18,445 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-29 22:14:36,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1412, 3.9509, 4.1830, 4.3230, 4.3740, 3.9787, 4.1704, 4.3897], device='cuda:3'), covar=tensor([0.1318, 0.1025, 0.1210, 0.0596, 0.0548, 0.1336, 0.1863, 0.0609], device='cuda:3'), in_proj_covar=tensor([0.0588, 0.0727, 0.0873, 0.0743, 0.0560, 0.0572, 0.0586, 0.0687], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:14:46,641 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:15:03,755 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.267e+02 2.734e+02 3.162e+02 6.055e+02, threshold=5.467e+02, percent-clipped=3.0 2023-04-29 22:15:22,991 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 22:15:23,303 INFO [train.py:904] (3/8) Epoch 14, batch 1550, loss[loss=0.1549, simple_loss=0.2406, pruned_loss=0.0346, over 17222.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2562, pruned_loss=0.04561, over 3323632.49 frames. ], batch size: 45, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:16:10,846 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:16:33,557 INFO [train.py:904] (3/8) Epoch 14, batch 1600, loss[loss=0.1609, simple_loss=0.2471, pruned_loss=0.03729, over 16707.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2578, pruned_loss=0.04642, over 3328287.39 frames. ], batch size: 37, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:17:21,876 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 1650, loss[loss=0.1602, simple_loss=0.2512, pruned_loss=0.03461, over 17184.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2594, pruned_loss=0.04716, over 3325549.62 frames. ], batch size: 44, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:37,800 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 22:18:52,510 INFO [train.py:904] (3/8) Epoch 14, batch 1700, loss[loss=0.2032, simple_loss=0.2769, pruned_loss=0.06472, over 16478.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2623, pruned_loss=0.0483, over 3314257.58 frames. ], batch size: 146, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:53,046 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:20:03,031 INFO [train.py:904] (3/8) Epoch 14, batch 1750, loss[loss=0.1917, simple_loss=0.2686, pruned_loss=0.05742, over 16405.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2629, pruned_loss=0.04844, over 3309944.28 frames. ], batch size: 146, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:20:08,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-29 22:20:19,322 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2623, 4.0648, 4.3333, 4.4795, 4.5604, 4.1423, 4.3642, 4.5787], device='cuda:3'), covar=tensor([0.1582, 0.1145, 0.1279, 0.0686, 0.0603, 0.1132, 0.2157, 0.0641], device='cuda:3'), in_proj_covar=tensor([0.0591, 0.0731, 0.0877, 0.0747, 0.0562, 0.0577, 0.0587, 0.0690], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:20:19,376 INFO [zipformer.py:625] (3/8) 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] (3/8) attn_weights_entropy = tensor([4.9664, 4.9868, 5.5150, 5.4681, 5.4633, 5.0640, 5.0549, 4.7319], device='cuda:3'), covar=tensor([0.0307, 0.0444, 0.0307, 0.0383, 0.0397, 0.0339, 0.0912, 0.0463], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0391, 0.0391, 0.0373, 0.0433, 0.0418, 0.0511, 0.0332], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 22:20:42,477 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 22:20:44,897 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 1800, loss[loss=0.2005, simple_loss=0.2897, pruned_loss=0.0557, over 16662.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2639, pruned_loss=0.04826, over 3317114.23 frames. ], batch size: 57, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:21:37,193 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 22:22:00,849 INFO [optim.py:368] (3/8) 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 22:22:22,387 INFO [train.py:904] (3/8) Epoch 14, batch 1850, loss[loss=0.1887, simple_loss=0.2693, pruned_loss=0.05408, over 16451.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2646, pruned_loss=0.04814, over 3316325.03 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:02,957 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 1900, loss[loss=0.181, simple_loss=0.2625, pruned_loss=0.0498, over 16528.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2642, pruned_loss=0.04772, over 3309022.06 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:44,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4031, 5.2795, 5.2019, 4.7411, 4.7979, 5.2494, 5.2264, 4.8548], device='cuda:3'), covar=tensor([0.0533, 0.0502, 0.0254, 0.0283, 0.0988, 0.0423, 0.0283, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0372, 0.0328, 0.0308, 0.0343, 0.0357, 0.0222, 0.0382], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:24:15,907 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5630, 3.6246, 3.3062, 2.9696, 3.2112, 3.4591, 3.2980, 3.2871], device='cuda:3'), covar=tensor([0.0595, 0.0480, 0.0254, 0.0255, 0.0609, 0.0352, 0.1475, 0.0489], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0373, 0.0328, 0.0308, 0.0344, 0.0358, 0.0222, 0.0383], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:24:23,479 INFO [optim.py:368] (3/8) 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 22:24:42,151 INFO [train.py:904] (3/8) Epoch 14, batch 1950, loss[loss=0.1752, simple_loss=0.2695, pruned_loss=0.04041, over 17059.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2643, pruned_loss=0.04715, over 3305174.95 frames. ], batch size: 55, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:25:27,295 INFO [scaling.py:679] (3/8) 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] (3/8) Epoch 14, batch 2000, loss[loss=0.1681, simple_loss=0.2469, pruned_loss=0.0447, over 16975.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2633, pruned_loss=0.04661, over 3305087.93 frames. ], batch size: 41, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:25:53,762 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-29 22:26:18,089 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 22:26:29,279 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-04-29 22:26:43,618 INFO [optim.py:368] (3/8) 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 22:26:46,373 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3276, 3.2641, 3.3509, 3.4935, 3.5136, 3.2790, 3.4322, 3.5779], device='cuda:3'), covar=tensor([0.1071, 0.0884, 0.0953, 0.0603, 0.0557, 0.2131, 0.1232, 0.0650], device='cuda:3'), in_proj_covar=tensor([0.0597, 0.0739, 0.0887, 0.0757, 0.0568, 0.0586, 0.0597, 0.0700], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:26:57,062 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 2050, loss[loss=0.172, simple_loss=0.2632, pruned_loss=0.04039, over 17142.00 frames. ], tot_loss[loss=0.179, simple_loss=0.264, pruned_loss=0.04704, over 3307965.61 frames. ], batch size: 47, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:27:15,545 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:21,597 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0618, 5.0191, 4.8124, 4.2491, 4.8793, 1.8973, 4.6332, 4.7473], device='cuda:3'), covar=tensor([0.0094, 0.0089, 0.0192, 0.0427, 0.0105, 0.2565, 0.0149, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0136, 0.0184, 0.0170, 0.0154, 0.0196, 0.0173, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:27:28,540 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 22:27:46,384 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 22:27:49,261 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 2100, loss[loss=0.1979, simple_loss=0.2848, pruned_loss=0.05547, over 17072.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.265, pruned_loss=0.04831, over 3317182.15 frames. ], batch size: 55, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:28:25,589 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:28:54,052 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 22:28:56,312 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:29:09,012 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.284e+02 2.721e+02 3.276e+02 7.722e+02, threshold=5.442e+02, percent-clipped=1.0 2023-04-29 22:29:26,684 INFO [train.py:904] (3/8) Epoch 14, batch 2150, loss[loss=0.1855, simple_loss=0.2624, pruned_loss=0.05429, over 16867.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2656, pruned_loss=0.04885, over 3322957.74 frames. ], batch size: 96, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:29:59,849 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 22:30:05,049 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8983, 3.0067, 2.5732, 4.6329, 3.8549, 4.2305, 1.5945, 3.0469], device='cuda:3'), covar=tensor([0.1257, 0.0664, 0.1130, 0.0186, 0.0229, 0.0392, 0.1510, 0.0787], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0163, 0.0181, 0.0161, 0.0198, 0.0210, 0.0186, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:30:06,129 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:30:30,227 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9064, 2.2862, 2.3659, 4.7407, 2.3140, 2.7017, 2.3794, 2.5298], device='cuda:3'), covar=tensor([0.0991, 0.3581, 0.2686, 0.0353, 0.4138, 0.2527, 0.3160, 0.3536], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0412, 0.0345, 0.0328, 0.0421, 0.0476, 0.0378, 0.0484], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:30:35,961 INFO [train.py:904] (3/8) Epoch 14, batch 2200, loss[loss=0.1544, simple_loss=0.239, pruned_loss=0.03487, over 16751.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2659, pruned_loss=0.04871, over 3324611.27 frames. ], batch size: 39, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:31:12,754 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:31:27,739 INFO [optim.py:368] (3/8) 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 22:31:41,765 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 22:31:45,970 INFO [train.py:904] (3/8) Epoch 14, batch 2250, loss[loss=0.2154, simple_loss=0.2985, pruned_loss=0.06619, over 15644.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2656, pruned_loss=0.04879, over 3328026.48 frames. ], batch size: 191, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:32:54,465 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8301, 3.7449, 3.9023, 4.0251, 4.0813, 3.6363, 3.9178, 4.1139], device='cuda:3'), covar=tensor([0.1473, 0.0986, 0.1205, 0.0630, 0.0558, 0.1991, 0.1800, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0601, 0.0743, 0.0890, 0.0764, 0.0572, 0.0588, 0.0602, 0.0702], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:32:56,452 INFO [train.py:904] (3/8) Epoch 14, batch 2300, loss[loss=0.181, simple_loss=0.2605, pruned_loss=0.05077, over 16885.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2649, pruned_loss=0.04848, over 3327979.93 frames. ], batch size: 96, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:33:48,102 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.438e+02 2.937e+02 3.292e+02 7.337e+02, threshold=5.875e+02, percent-clipped=1.0 2023-04-29 22:34:06,356 INFO [train.py:904] (3/8) Epoch 14, batch 2350, loss[loss=0.2021, simple_loss=0.2781, pruned_loss=0.063, over 16855.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.265, pruned_loss=0.04909, over 3321863.18 frames. ], batch size: 116, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:34:09,643 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1092, 5.1203, 5.6201, 5.5646, 5.6209, 5.2215, 5.2079, 4.9221], device='cuda:3'), covar=tensor([0.0285, 0.0499, 0.0347, 0.0438, 0.0420, 0.0356, 0.0864, 0.0414], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0395, 0.0397, 0.0378, 0.0442, 0.0422, 0.0515, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 22:34:09,686 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:15,198 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:17,035 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:57,559 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8786, 4.8117, 4.7641, 4.4506, 4.4655, 4.8223, 4.7376, 4.4870], device='cuda:3'), covar=tensor([0.0642, 0.0731, 0.0278, 0.0290, 0.0823, 0.0462, 0.0384, 0.0684], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0370, 0.0328, 0.0308, 0.0342, 0.0354, 0.0223, 0.0381], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:35:17,012 INFO [train.py:904] (3/8) Epoch 14, batch 2400, loss[loss=0.1971, simple_loss=0.2743, pruned_loss=0.05992, over 16687.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2656, pruned_loss=0.04909, over 3324602.87 frames. ], batch size: 134, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:35:19,203 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:23,354 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:35,559 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:42,620 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:36:08,842 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 2450, loss[loss=0.1502, simple_loss=0.2444, pruned_loss=0.028, over 17239.00 frames. ], tot_loss[loss=0.182, simple_loss=0.266, pruned_loss=0.04899, over 3329927.08 frames. ], batch size: 45, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:37:35,054 INFO [train.py:904] (3/8) Epoch 14, batch 2500, loss[loss=0.1633, simple_loss=0.2576, pruned_loss=0.0345, over 17042.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2664, pruned_loss=0.04839, over 3330254.70 frames. ], batch size: 50, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:38:05,093 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 22:38:28,206 INFO [optim.py:368] (3/8) 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 22:38:37,238 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-29 22:38:41,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2109, 1.5169, 1.9586, 2.0601, 2.2459, 2.2440, 1.6457, 2.2756], device='cuda:3'), covar=tensor([0.0185, 0.0370, 0.0220, 0.0238, 0.0210, 0.0215, 0.0365, 0.0109], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0181, 0.0166, 0.0171, 0.0180, 0.0136, 0.0181, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:38:45,442 INFO [train.py:904] (3/8) Epoch 14, batch 2550, loss[loss=0.1678, simple_loss=0.2498, pruned_loss=0.04297, over 16892.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2663, pruned_loss=0.04877, over 3326861.32 frames. ], batch size: 96, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:38:49,917 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9386, 4.6650, 4.9658, 5.1787, 5.3675, 4.6093, 5.3432, 5.3277], device='cuda:3'), covar=tensor([0.1716, 0.1383, 0.1766, 0.0720, 0.0538, 0.0948, 0.0465, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0607, 0.0749, 0.0902, 0.0770, 0.0576, 0.0595, 0.0608, 0.0708], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:39:06,150 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7228, 4.6997, 4.6428, 4.0997, 4.6957, 1.9149, 4.4384, 4.4423], device='cuda:3'), covar=tensor([0.0102, 0.0089, 0.0164, 0.0316, 0.0082, 0.2311, 0.0143, 0.0163], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0135, 0.0182, 0.0169, 0.0153, 0.0194, 0.0172, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:39:55,149 INFO [train.py:904] (3/8) Epoch 14, batch 2600, loss[loss=0.1892, simple_loss=0.2798, pruned_loss=0.0493, over 17103.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2655, pruned_loss=0.04795, over 3329127.06 frames. ], batch size: 53, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:40:19,780 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-29 22:40:46,904 INFO [optim.py:368] (3/8) 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 22:41:03,701 INFO [train.py:904] (3/8) Epoch 14, batch 2650, loss[loss=0.1571, simple_loss=0.2525, pruned_loss=0.03086, over 17247.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2662, pruned_loss=0.04749, over 3334927.94 frames. ], batch size: 45, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:12,178 INFO [train.py:904] (3/8) Epoch 14, batch 2700, loss[loss=0.1865, simple_loss=0.2625, pruned_loss=0.05524, over 16514.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2664, pruned_loss=0.04707, over 3329415.56 frames. ], batch size: 146, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:13,620 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:16,938 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5423, 3.9399, 4.1121, 2.3661, 3.4084, 2.7785, 3.9644, 4.0764], device='cuda:3'), covar=tensor([0.0297, 0.0782, 0.0445, 0.1676, 0.0684, 0.0833, 0.0622, 0.0922], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0152, 0.0160, 0.0147, 0.0139, 0.0126, 0.0138, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:42:23,624 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:31,294 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:04,243 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.219e+02 2.559e+02 3.067e+02 4.567e+02, threshold=5.119e+02, percent-clipped=0.0 2023-04-29 22:43:05,838 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5079, 3.6421, 2.2143, 3.9184, 2.7872, 3.9015, 2.1843, 2.8989], device='cuda:3'), covar=tensor([0.0241, 0.0322, 0.1422, 0.0242, 0.0765, 0.0500, 0.1385, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0172, 0.0191, 0.0148, 0.0170, 0.0214, 0.0200, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:43:20,427 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:21,290 INFO [train.py:904] (3/8) Epoch 14, batch 2750, loss[loss=0.2005, simple_loss=0.274, pruned_loss=0.06349, over 16858.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2669, pruned_loss=0.04691, over 3324958.99 frames. ], batch size: 109, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:43:27,364 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9761, 4.4168, 3.0722, 2.3570, 3.0026, 2.6808, 4.7315, 3.7925], device='cuda:3'), covar=tensor([0.2550, 0.0586, 0.1721, 0.2511, 0.2550, 0.1696, 0.0403, 0.1184], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0265, 0.0293, 0.0292, 0.0287, 0.0234, 0.0279, 0.0317], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:43:48,148 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 22:43:52,762 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 22:44:04,354 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 22:44:29,010 INFO [train.py:904] (3/8) Epoch 14, batch 2800, loss[loss=0.1887, simple_loss=0.2676, pruned_loss=0.05488, over 16755.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2664, pruned_loss=0.04696, over 3331425.88 frames. ], batch size: 124, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:44:29,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7703, 4.8397, 5.0095, 4.8046, 4.8310, 5.4695, 4.9463, 4.6417], device='cuda:3'), covar=tensor([0.1355, 0.2017, 0.2044, 0.2227, 0.2633, 0.1027, 0.1553, 0.2638], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0536, 0.0590, 0.0458, 0.0618, 0.0614, 0.0468, 0.0613], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:45:20,154 INFO [optim.py:368] (3/8) 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 22:45:37,180 INFO [train.py:904] (3/8) Epoch 14, batch 2850, loss[loss=0.1668, simple_loss=0.2585, pruned_loss=0.03761, over 17102.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2657, pruned_loss=0.04682, over 3339411.39 frames. ], batch size: 47, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:04,053 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-29 22:46:22,318 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8245, 3.9351, 2.3185, 4.5572, 2.9223, 4.5132, 2.5312, 3.1924], device='cuda:3'), covar=tensor([0.0250, 0.0364, 0.1550, 0.0199, 0.0815, 0.0376, 0.1369, 0.0653], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0172, 0.0191, 0.0149, 0.0170, 0.0215, 0.0200, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:46:45,307 INFO [train.py:904] (3/8) Epoch 14, batch 2900, loss[loss=0.1841, simple_loss=0.2808, pruned_loss=0.04367, over 17256.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2645, pruned_loss=0.04732, over 3336189.94 frames. ], batch size: 52, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:58,840 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:47:28,196 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4062, 3.6963, 3.2391, 2.0619, 2.8754, 2.2924, 3.8656, 3.9141], device='cuda:3'), covar=tensor([0.0228, 0.0625, 0.0679, 0.1891, 0.0859, 0.1051, 0.0461, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0153, 0.0162, 0.0148, 0.0140, 0.0127, 0.0140, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:47:34,490 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2285, 3.4176, 3.4143, 2.2229, 3.0022, 2.4768, 3.7496, 3.6514], device='cuda:3'), covar=tensor([0.0211, 0.0738, 0.0564, 0.1620, 0.0705, 0.0911, 0.0457, 0.0781], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0153, 0.0162, 0.0147, 0.0140, 0.0127, 0.0140, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:47:36,338 INFO [optim.py:368] (3/8) 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 22:47:45,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7597, 1.8137, 1.5744, 1.5098, 1.9483, 1.6260, 1.6839, 1.9203], device='cuda:3'), covar=tensor([0.0143, 0.0293, 0.0361, 0.0383, 0.0213, 0.0314, 0.0216, 0.0261], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0221, 0.0215, 0.0214, 0.0221, 0.0222, 0.0231, 0.0216], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:47:53,602 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 22:47:54,145 INFO [train.py:904] (3/8) Epoch 14, batch 2950, loss[loss=0.2125, simple_loss=0.281, pruned_loss=0.07198, over 15660.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2645, pruned_loss=0.04803, over 3330789.78 frames. ], batch size: 191, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:47:56,061 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 22:48:07,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0772, 4.0272, 3.9679, 3.4867, 3.9916, 1.7885, 3.7601, 3.5347], device='cuda:3'), covar=tensor([0.0113, 0.0105, 0.0171, 0.0266, 0.0083, 0.2412, 0.0127, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0136, 0.0183, 0.0171, 0.0154, 0.0195, 0.0173, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:48:24,035 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:02,685 INFO [train.py:904] (3/8) Epoch 14, batch 3000, loss[loss=0.1605, simple_loss=0.2478, pruned_loss=0.03658, over 16844.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2646, pruned_loss=0.04836, over 3330129.73 frames. ], batch size: 42, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:49:02,686 INFO [train.py:929] (3/8) Computing validation loss 2023-04-29 22:49:12,424 INFO [train.py:938] (3/8) Epoch 14, validation: loss=0.1382, simple_loss=0.244, pruned_loss=0.01621, over 944034.00 frames. 2023-04-29 22:49:12,424 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-29 22:49:24,935 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:31,423 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:06,980 INFO [optim.py:368] (3/8) 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 22:50:24,239 INFO [train.py:904] (3/8) Epoch 14, batch 3050, loss[loss=0.1592, simple_loss=0.2575, pruned_loss=0.03039, over 17124.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2646, pruned_loss=0.04861, over 3327042.55 frames. ], batch size: 48, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:50:25,007 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 22:50:33,224 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:40,116 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:51:32,217 INFO [train.py:904] (3/8) Epoch 14, batch 3100, loss[loss=0.1943, simple_loss=0.2807, pruned_loss=0.05394, over 17058.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.264, pruned_loss=0.04839, over 3324812.44 frames. ], batch size: 53, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:52:06,922 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9577, 4.9608, 5.4287, 5.4440, 5.4506, 5.0868, 5.0408, 4.7672], device='cuda:3'), covar=tensor([0.0320, 0.0492, 0.0450, 0.0434, 0.0439, 0.0386, 0.0966, 0.0445], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0395, 0.0397, 0.0376, 0.0439, 0.0420, 0.0515, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 22:52:18,122 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7531, 2.9687, 2.6960, 5.0529, 4.2554, 4.4564, 1.5528, 3.3276], device='cuda:3'), covar=tensor([0.1390, 0.0755, 0.1118, 0.0154, 0.0226, 0.0402, 0.1633, 0.0717], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0181, 0.0162, 0.0200, 0.0210, 0.0186, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:52:24,972 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 3150, loss[loss=0.1981, simple_loss=0.3, pruned_loss=0.04813, over 17070.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2631, pruned_loss=0.0476, over 3326219.08 frames. ], batch size: 55, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:53:50,739 INFO [train.py:904] (3/8) Epoch 14, batch 3200, loss[loss=0.1778, simple_loss=0.2565, pruned_loss=0.04951, over 15479.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2623, pruned_loss=0.04685, over 3328959.20 frames. ], batch size: 190, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:53:52,802 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 22:54:17,402 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:54:33,125 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-29 22:54:41,146 INFO [optim.py:368] (3/8) 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 22:54:59,017 INFO [train.py:904] (3/8) Epoch 14, batch 3250, loss[loss=0.1592, simple_loss=0.2446, pruned_loss=0.03694, over 17217.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2623, pruned_loss=0.04666, over 3318210.47 frames. ], batch size: 44, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:55:22,124 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:55:41,122 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 3300, loss[loss=0.2245, simple_loss=0.2872, pruned_loss=0.08096, over 16875.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2649, pruned_loss=0.04838, over 3313708.13 frames. ], batch size: 109, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:56:21,195 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0064, 5.0937, 5.5230, 5.5060, 5.5324, 5.1581, 5.0997, 4.8171], device='cuda:3'), covar=tensor([0.0304, 0.0424, 0.0395, 0.0421, 0.0412, 0.0338, 0.0938, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0394, 0.0393, 0.0372, 0.0439, 0.0418, 0.0512, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-29 22:57:01,615 INFO [optim.py:368] (3/8) 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 22:57:16,993 INFO [train.py:904] (3/8) Epoch 14, batch 3350, loss[loss=0.1847, simple_loss=0.2575, pruned_loss=0.05597, over 16694.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2654, pruned_loss=0.04831, over 3308910.77 frames. ], batch size: 89, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:57:46,415 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5359, 2.2603, 2.2442, 4.2613, 2.1585, 2.7357, 2.3343, 2.4705], device='cuda:3'), covar=tensor([0.0997, 0.3377, 0.2619, 0.0464, 0.3956, 0.2282, 0.3282, 0.3263], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0413, 0.0345, 0.0329, 0.0421, 0.0477, 0.0378, 0.0485], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:58:24,246 INFO [train.py:904] (3/8) Epoch 14, batch 3400, loss[loss=0.1763, simple_loss=0.2665, pruned_loss=0.04302, over 16765.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2654, pruned_loss=0.04828, over 3314250.76 frames. ], batch size: 57, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:59:01,410 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9429, 2.6320, 2.6138, 1.9822, 2.5397, 2.7394, 2.5412, 1.8997], device='cuda:3'), covar=tensor([0.0317, 0.0075, 0.0053, 0.0292, 0.0102, 0.0089, 0.0094, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0073, 0.0073, 0.0127, 0.0085, 0.0094, 0.0083, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:59:16,983 INFO [optim.py:368] (3/8) 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 22:59:20,384 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7244, 4.5293, 4.6738, 4.9087, 5.0574, 4.5376, 4.9943, 5.0329], device='cuda:3'), covar=tensor([0.1771, 0.1243, 0.1821, 0.1004, 0.0743, 0.0986, 0.1179, 0.1124], device='cuda:3'), in_proj_covar=tensor([0.0601, 0.0745, 0.0899, 0.0765, 0.0576, 0.0588, 0.0602, 0.0704], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:59:32,673 INFO [train.py:904] (3/8) Epoch 14, batch 3450, loss[loss=0.1951, simple_loss=0.2662, pruned_loss=0.062, over 16825.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2635, pruned_loss=0.04738, over 3318192.58 frames. ], batch size: 116, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:59:35,413 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0308, 2.4440, 2.3272, 2.7858, 2.4243, 3.2666, 1.8260, 2.6518], device='cuda:3'), covar=tensor([0.0906, 0.0537, 0.0873, 0.0171, 0.0144, 0.0415, 0.1062, 0.0603], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0164, 0.0183, 0.0164, 0.0202, 0.0212, 0.0188, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 22:59:43,672 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8658, 3.6137, 3.8223, 3.5785, 3.7646, 4.2144, 3.9587, 3.5421], device='cuda:3'), covar=tensor([0.1942, 0.2594, 0.1938, 0.2970, 0.2942, 0.1939, 0.1330, 0.2650], device='cuda:3'), in_proj_covar=tensor([0.0382, 0.0545, 0.0595, 0.0463, 0.0627, 0.0624, 0.0472, 0.0617], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 22:59:48,825 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3453, 4.6780, 4.4637, 4.4984, 4.1879, 4.2101, 4.2470, 4.6982], device='cuda:3'), covar=tensor([0.1185, 0.0811, 0.1014, 0.0711, 0.0793, 0.1352, 0.0984, 0.0941], device='cuda:3'), in_proj_covar=tensor([0.0609, 0.0759, 0.0622, 0.0547, 0.0485, 0.0481, 0.0632, 0.0584], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 22:59:57,257 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 23:00:41,080 INFO [train.py:904] (3/8) Epoch 14, batch 3500, loss[loss=0.1818, simple_loss=0.2719, pruned_loss=0.0459, over 16762.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2626, pruned_loss=0.04693, over 3311755.02 frames. ], batch size: 62, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:00:47,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1618, 3.2450, 3.3222, 2.1133, 2.9002, 2.4441, 3.6788, 3.5672], device='cuda:3'), covar=tensor([0.0237, 0.0907, 0.0592, 0.1809, 0.0784, 0.0968, 0.0493, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0156, 0.0163, 0.0148, 0.0140, 0.0128, 0.0141, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 23:01:04,949 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 23:01:33,347 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3425, 4.1282, 4.5228, 2.1404, 4.7540, 4.8099, 3.5348, 3.7144], device='cuda:3'), covar=tensor([0.0602, 0.0219, 0.0208, 0.1126, 0.0061, 0.0146, 0.0333, 0.0362], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0104, 0.0090, 0.0137, 0.0072, 0.0115, 0.0122, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-29 23:01:37,084 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 2.212e+02 2.743e+02 3.344e+02 6.080e+02, threshold=5.486e+02, percent-clipped=2.0 2023-04-29 23:01:51,741 INFO [train.py:904] (3/8) Epoch 14, batch 3550, loss[loss=0.1612, simple_loss=0.2537, pruned_loss=0.0344, over 17124.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2618, pruned_loss=0.04654, over 3311036.51 frames. ], batch size: 48, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:02:15,165 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:27,912 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:03:02,250 INFO [train.py:904] (3/8) Epoch 14, batch 3600, loss[loss=0.1423, simple_loss=0.2219, pruned_loss=0.03134, over 16898.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2605, pruned_loss=0.04628, over 3309607.79 frames. ], batch size: 96, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:03:22,564 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:03:58,636 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.279e+02 2.683e+02 3.136e+02 7.106e+02, threshold=5.366e+02, percent-clipped=1.0 2023-04-29 23:04:14,253 INFO [train.py:904] (3/8) Epoch 14, batch 3650, loss[loss=0.1484, simple_loss=0.2251, pruned_loss=0.03581, over 15925.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2593, pruned_loss=0.04667, over 3310715.50 frames. ], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:04:20,163 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 23:04:57,244 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:05:31,032 INFO [train.py:904] (3/8) Epoch 14, batch 3700, loss[loss=0.169, simple_loss=0.2417, pruned_loss=0.04817, over 16593.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2581, pruned_loss=0.04801, over 3291072.45 frames. ], batch size: 68, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:06:30,616 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:06:32,146 INFO [optim.py:368] (3/8) 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 23:06:46,315 INFO [train.py:904] (3/8) Epoch 14, batch 3750, loss[loss=0.1752, simple_loss=0.2434, pruned_loss=0.05346, over 16818.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2591, pruned_loss=0.04965, over 3268017.79 frames. ], batch size: 102, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:07:30,536 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6041, 3.6324, 2.8117, 2.1020, 2.3100, 2.1982, 3.6292, 3.1868], device='cuda:3'), covar=tensor([0.2495, 0.0608, 0.1535, 0.2848, 0.2836, 0.1993, 0.0485, 0.1387], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0264, 0.0291, 0.0291, 0.0289, 0.0234, 0.0277, 0.0316], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 23:08:01,576 INFO [train.py:904] (3/8) Epoch 14, batch 3800, loss[loss=0.1965, simple_loss=0.275, pruned_loss=0.05903, over 16571.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2608, pruned_loss=0.05102, over 3256676.57 frames. ], batch size: 62, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:00,855 INFO [optim.py:368] (3/8) 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 23:09:05,815 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-04-29 23:09:16,026 INFO [train.py:904] (3/8) Epoch 14, batch 3850, loss[loss=0.1741, simple_loss=0.2496, pruned_loss=0.04929, over 16853.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.261, pruned_loss=0.05174, over 3259964.49 frames. ], batch size: 96, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:30,729 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7507, 4.7108, 4.6892, 4.1245, 4.7118, 1.7322, 4.4834, 4.3917], device='cuda:3'), covar=tensor([0.0105, 0.0089, 0.0172, 0.0327, 0.0087, 0.2497, 0.0134, 0.0170], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0135, 0.0182, 0.0170, 0.0154, 0.0193, 0.0173, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:09:43,642 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1884, 2.0401, 2.2247, 3.9243, 2.0950, 2.3976, 2.1400, 2.2465], device='cuda:3'), covar=tensor([0.1172, 0.3322, 0.2452, 0.0476, 0.3555, 0.2439, 0.3487, 0.2763], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0417, 0.0347, 0.0329, 0.0423, 0.0481, 0.0381, 0.0490], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:09:53,693 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:10:30,026 INFO [train.py:904] (3/8) Epoch 14, batch 3900, loss[loss=0.1724, simple_loss=0.2491, pruned_loss=0.04789, over 16741.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2595, pruned_loss=0.05159, over 3277860.88 frames. ], batch size: 83, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:05,160 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.144e+02 2.448e+02 3.272e+02 6.091e+02, threshold=4.896e+02, percent-clipped=2.0 2023-04-29 23:11:40,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9344, 5.2099, 5.3966, 5.2493, 5.2679, 5.8384, 5.3083, 5.0499], device='cuda:3'), covar=tensor([0.1172, 0.1927, 0.2093, 0.1914, 0.2468, 0.0934, 0.1438, 0.2103], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0541, 0.0585, 0.0457, 0.0618, 0.0612, 0.0467, 0.0608], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 23:11:45,241 INFO [train.py:904] (3/8) Epoch 14, batch 3950, loss[loss=0.1809, simple_loss=0.2519, pruned_loss=0.05494, over 16704.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2595, pruned_loss=0.05229, over 3274517.47 frames. ], batch size: 134, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:12:58,425 INFO [train.py:904] (3/8) Epoch 14, batch 4000, loss[loss=0.1729, simple_loss=0.2615, pruned_loss=0.04212, over 17004.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2597, pruned_loss=0.05284, over 3279601.25 frames. ], batch size: 41, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:13:35,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3697, 3.5413, 1.9912, 3.7484, 2.6184, 3.8584, 2.2287, 2.7420], device='cuda:3'), covar=tensor([0.0269, 0.0353, 0.1731, 0.0209, 0.0833, 0.0417, 0.1469, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0191, 0.0149, 0.0171, 0.0215, 0.0201, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 23:13:41,758 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:47,673 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:55,964 INFO [optim.py:368] (3/8) 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 23:13:58,955 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:14:13,530 INFO [train.py:904] (3/8) Epoch 14, batch 4050, loss[loss=0.1524, simple_loss=0.2432, pruned_loss=0.0308, over 16695.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2604, pruned_loss=0.05191, over 3286004.92 frames. ], batch size: 89, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:11,249 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 23:15:12,377 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:15:24,937 INFO [train.py:904] (3/8) Epoch 14, batch 4100, loss[loss=0.168, simple_loss=0.2535, pruned_loss=0.04124, over 17271.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2612, pruned_loss=0.05103, over 3284782.75 frames. ], batch size: 52, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:30,442 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:16:22,082 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 23:16:24,202 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 4150, loss[loss=0.2288, simple_loss=0.3223, pruned_loss=0.06769, over 16703.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2694, pruned_loss=0.05419, over 3238252.27 frames. ], batch size: 134, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:17:02,278 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3515, 4.1514, 3.8790, 4.4683, 4.5498, 4.2762, 4.5373, 4.6538], device='cuda:3'), covar=tensor([0.1363, 0.1204, 0.2907, 0.1187, 0.0971, 0.1280, 0.1101, 0.1078], device='cuda:3'), in_proj_covar=tensor([0.0583, 0.0721, 0.0861, 0.0740, 0.0557, 0.0568, 0.0577, 0.0682], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:17:41,235 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:17:55,762 INFO [train.py:904] (3/8) Epoch 14, batch 4200, loss[loss=0.2286, simple_loss=0.315, pruned_loss=0.07106, over 16665.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2768, pruned_loss=0.05598, over 3205919.35 frames. ], batch size: 57, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:18:55,501 INFO [optim.py:368] (3/8) 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 23:19:09,991 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 23:19:10,105 INFO [train.py:904] (3/8) Epoch 14, batch 4250, loss[loss=0.1944, simple_loss=0.2886, pruned_loss=0.05012, over 17126.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2796, pruned_loss=0.05537, over 3197833.35 frames. ], batch size: 48, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:19:12,566 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 4300, loss[loss=0.1877, simple_loss=0.2837, pruned_loss=0.04584, over 16757.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2809, pruned_loss=0.0541, over 3218515.42 frames. ], batch size: 89, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:21:14,588 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:21:23,663 INFO [optim.py:368] (3/8) 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 23:21:38,134 INFO [train.py:904] (3/8) Epoch 14, batch 4350, loss[loss=0.2238, simple_loss=0.3081, pruned_loss=0.06972, over 16785.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2837, pruned_loss=0.05488, over 3216232.93 frames. ], batch size: 39, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:21:41,040 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4606, 4.5078, 4.3265, 4.0614, 4.0234, 4.4356, 4.1254, 4.0993], device='cuda:3'), covar=tensor([0.0523, 0.0306, 0.0238, 0.0243, 0.0793, 0.0301, 0.0576, 0.0559], device='cuda:3'), in_proj_covar=tensor([0.0258, 0.0356, 0.0312, 0.0292, 0.0329, 0.0340, 0.0212, 0.0363], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:21:42,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2148, 1.6321, 1.9411, 2.2634, 2.3401, 2.5998, 1.6084, 2.3645], device='cuda:3'), covar=tensor([0.0177, 0.0388, 0.0225, 0.0254, 0.0227, 0.0132, 0.0410, 0.0107], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0163, 0.0171, 0.0179, 0.0135, 0.0179, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:21:50,991 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5105, 2.9676, 3.0115, 1.9472, 2.6546, 2.2202, 2.8936, 3.1635], device='cuda:3'), covar=tensor([0.0344, 0.0824, 0.0527, 0.1845, 0.0867, 0.0885, 0.0829, 0.0892], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0153, 0.0161, 0.0147, 0.0139, 0.0126, 0.0139, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 23:22:26,305 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:29,178 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:33,160 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:51,135 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:51,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4344, 2.5984, 2.1354, 2.3342, 2.9936, 2.6231, 3.1457, 3.1486], device='cuda:3'), covar=tensor([0.0064, 0.0311, 0.0427, 0.0358, 0.0176, 0.0287, 0.0178, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0214, 0.0210, 0.0209, 0.0216, 0.0216, 0.0223, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:22:53,919 INFO [train.py:904] (3/8) Epoch 14, batch 4400, loss[loss=0.1874, simple_loss=0.2856, pruned_loss=0.04455, over 16897.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2853, pruned_loss=0.0558, over 3210279.19 frames. ], batch size: 96, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:23:46,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6893, 1.8055, 1.6446, 1.5204, 1.9082, 1.6133, 1.6849, 1.9488], device='cuda:3'), covar=tensor([0.0118, 0.0222, 0.0314, 0.0274, 0.0146, 0.0202, 0.0125, 0.0150], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0213, 0.0209, 0.0208, 0.0215, 0.0216, 0.0221, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:23:51,148 INFO [optim.py:368] (3/8) 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 23:23:57,355 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:24:06,500 INFO [train.py:904] (3/8) Epoch 14, batch 4450, loss[loss=0.2248, simple_loss=0.3115, pruned_loss=0.06901, over 16502.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2889, pruned_loss=0.05704, over 3223320.78 frames. ], batch size: 35, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:18,607 INFO [train.py:904] (3/8) Epoch 14, batch 4500, loss[loss=0.2199, simple_loss=0.3062, pruned_loss=0.06679, over 16737.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2892, pruned_loss=0.05776, over 3216872.67 frames. ], batch size: 89, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:52,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2762, 5.5174, 5.2942, 5.3261, 5.0181, 4.8127, 4.9658, 5.6399], device='cuda:3'), covar=tensor([0.0890, 0.0720, 0.0860, 0.0663, 0.0702, 0.0726, 0.0822, 0.0703], device='cuda:3'), in_proj_covar=tensor([0.0581, 0.0719, 0.0588, 0.0522, 0.0463, 0.0463, 0.0604, 0.0558], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:26:18,140 INFO [optim.py:368] (3/8) 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 23:26:27,078 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 4550, loss[loss=0.2045, simple_loss=0.2947, pruned_loss=0.05709, over 16504.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2901, pruned_loss=0.05895, over 3207598.06 frames. ], batch size: 68, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:26:48,558 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9607, 5.4890, 5.7091, 5.4557, 5.3613, 6.0901, 5.5209, 5.2644], device='cuda:3'), covar=tensor([0.0759, 0.1611, 0.1600, 0.1751, 0.2639, 0.0828, 0.1203, 0.2305], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0524, 0.0564, 0.0443, 0.0597, 0.0593, 0.0453, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-29 23:27:24,859 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 23:27:45,695 INFO [train.py:904] (3/8) Epoch 14, batch 4600, loss[loss=0.1845, simple_loss=0.2677, pruned_loss=0.0507, over 16666.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2905, pruned_loss=0.05888, over 3209323.17 frames. ], batch size: 57, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:28:42,903 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.145e+02 2.405e+02 2.956e+02 4.611e+02, threshold=4.810e+02, percent-clipped=0.0 2023-04-29 23:28:57,101 INFO [train.py:904] (3/8) Epoch 14, batch 4650, loss[loss=0.1987, simple_loss=0.2903, pruned_loss=0.05357, over 16487.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2897, pruned_loss=0.05886, over 3208716.17 frames. ], batch size: 68, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:29:52,214 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:29:58,853 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:08,005 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:10,584 INFO [train.py:904] (3/8) Epoch 14, batch 4700, loss[loss=0.1967, simple_loss=0.2765, pruned_loss=0.05848, over 17233.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2868, pruned_loss=0.05766, over 3217373.41 frames. ], batch size: 45, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:01,303 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:09,101 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:31:14,782 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5672, 3.6732, 2.8682, 2.2047, 2.5834, 2.3862, 4.0191, 3.3912], device='cuda:3'), covar=tensor([0.2840, 0.0790, 0.1663, 0.2235, 0.2307, 0.1765, 0.0473, 0.1039], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0260, 0.0290, 0.0289, 0.0285, 0.0231, 0.0276, 0.0311], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-29 23:31:18,047 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:23,575 INFO [train.py:904] (3/8) Epoch 14, batch 4750, loss[loss=0.1765, simple_loss=0.2624, pruned_loss=0.04531, over 16373.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.283, pruned_loss=0.0558, over 3216488.38 frames. ], batch size: 35, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:27,801 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:32:36,916 INFO [train.py:904] (3/8) Epoch 14, batch 4800, loss[loss=0.1903, simple_loss=0.2882, pruned_loss=0.04615, over 15399.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2795, pruned_loss=0.05353, over 3214243.70 frames. ], batch size: 190, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:03,258 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0937, 1.5118, 1.8588, 2.0215, 2.1665, 2.3060, 1.5532, 2.2348], device='cuda:3'), covar=tensor([0.0194, 0.0393, 0.0243, 0.0268, 0.0249, 0.0164, 0.0435, 0.0106], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0177, 0.0162, 0.0170, 0.0178, 0.0134, 0.0179, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:33:36,432 INFO [optim.py:368] (3/8) 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 23:33:46,293 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:33:52,184 INFO [train.py:904] (3/8) Epoch 14, batch 4850, loss[loss=0.1923, simple_loss=0.2885, pruned_loss=0.04806, over 15336.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2808, pruned_loss=0.05297, over 3205909.57 frames. ], batch size: 190, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:58,262 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:35:05,044 INFO [train.py:904] (3/8) Epoch 14, batch 4900, loss[loss=0.1964, simple_loss=0.2906, pruned_loss=0.05106, over 16207.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2803, pruned_loss=0.05212, over 3181427.99 frames. ], batch size: 165, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:35:28,833 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:36:02,817 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.076e+02 2.406e+02 2.771e+02 5.710e+02, threshold=4.811e+02, percent-clipped=1.0 2023-04-29 23:36:18,496 INFO [train.py:904] (3/8) Epoch 14, batch 4950, loss[loss=0.2042, simple_loss=0.2953, pruned_loss=0.05656, over 17221.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2797, pruned_loss=0.0517, over 3193750.35 frames. ], batch size: 44, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:37:30,153 INFO [train.py:904] (3/8) Epoch 14, batch 5000, loss[loss=0.1673, simple_loss=0.2698, pruned_loss=0.03239, over 16821.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2817, pruned_loss=0.05213, over 3199208.59 frames. ], batch size: 102, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:37:33,298 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 23:38:16,725 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2410, 3.6070, 3.6052, 2.0381, 2.9478, 2.4760, 3.5879, 3.7023], device='cuda:3'), covar=tensor([0.0257, 0.0700, 0.0605, 0.1868, 0.0871, 0.0890, 0.0584, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0153, 0.0162, 0.0148, 0.0140, 0.0127, 0.0139, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 23:38:24,300 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.271e+02 2.523e+02 3.075e+02 5.093e+02, threshold=5.046e+02, percent-clipped=1.0 2023-04-29 23:38:32,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0024, 2.0013, 2.1226, 3.5991, 1.9181, 2.3660, 2.1170, 2.1828], device='cuda:3'), covar=tensor([0.1284, 0.3464, 0.2579, 0.0511, 0.4001, 0.2323, 0.3373, 0.3049], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0414, 0.0343, 0.0322, 0.0418, 0.0475, 0.0374, 0.0482], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:38:36,266 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 5050, loss[loss=0.194, simple_loss=0.2857, pruned_loss=0.05116, over 16565.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.282, pruned_loss=0.05181, over 3203841.95 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:38:44,054 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 23:39:31,297 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:39:49,986 INFO [train.py:904] (3/8) Epoch 14, batch 5100, loss[loss=0.1874, simple_loss=0.2853, pruned_loss=0.0447, over 16873.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2808, pruned_loss=0.05146, over 3191647.43 frames. ], batch size: 116, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:40:03,772 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 23:40:45,948 INFO [optim.py:368] (3/8) 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 23:41:00,262 INFO [train.py:904] (3/8) Epoch 14, batch 5150, loss[loss=0.1896, simple_loss=0.2712, pruned_loss=0.05399, over 16406.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2815, pruned_loss=0.05139, over 3176387.54 frames. ], batch size: 35, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:08,569 INFO [train.py:904] (3/8) Epoch 14, batch 5200, loss[loss=0.1749, simple_loss=0.2595, pruned_loss=0.04508, over 16595.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2797, pruned_loss=0.0507, over 3192632.08 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:22,362 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:42:36,648 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-04-29 23:43:03,327 INFO [optim.py:368] (3/8) 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 23:43:18,145 INFO [train.py:904] (3/8) Epoch 14, batch 5250, loss[loss=0.1608, simple_loss=0.2572, pruned_loss=0.03219, over 16714.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2773, pruned_loss=0.05022, over 3210273.67 frames. ], batch size: 89, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:44:26,543 INFO [train.py:904] (3/8) Epoch 14, batch 5300, loss[loss=0.1607, simple_loss=0.2366, pruned_loss=0.04243, over 16347.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2731, pruned_loss=0.04881, over 3215154.48 frames. ], batch size: 35, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:44:32,782 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7112, 4.5668, 4.5705, 3.2547, 4.5638, 1.6271, 4.2090, 4.2777], device='cuda:3'), covar=tensor([0.0120, 0.0127, 0.0185, 0.0771, 0.0140, 0.3236, 0.0195, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0130, 0.0175, 0.0165, 0.0148, 0.0189, 0.0165, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:45:23,602 INFO [optim.py:368] (3/8) 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 23:45:33,749 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:45:36,799 INFO [train.py:904] (3/8) Epoch 14, batch 5350, loss[loss=0.2034, simple_loss=0.2835, pruned_loss=0.06168, over 12101.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2715, pruned_loss=0.04823, over 3211205.76 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:46:40,458 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:46:46,694 INFO [train.py:904] (3/8) Epoch 14, batch 5400, loss[loss=0.1846, simple_loss=0.2692, pruned_loss=0.04998, over 17015.00 frames. ], tot_loss[loss=0.186, simple_loss=0.274, pruned_loss=0.04898, over 3194712.88 frames. ], batch size: 41, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:47:27,685 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-29 23:47:37,501 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7783, 2.8040, 2.5891, 4.5283, 3.4760, 4.2136, 1.5927, 3.0573], device='cuda:3'), covar=tensor([0.1286, 0.0715, 0.1148, 0.0124, 0.0228, 0.0324, 0.1514, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0160, 0.0200, 0.0208, 0.0187, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-29 23:47:46,867 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.156e+02 2.492e+02 3.031e+02 4.895e+02, threshold=4.984e+02, percent-clipped=1.0 2023-04-29 23:48:02,464 INFO [train.py:904] (3/8) Epoch 14, batch 5450, loss[loss=0.2233, simple_loss=0.3109, pruned_loss=0.06788, over 16376.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2774, pruned_loss=0.05071, over 3177171.38 frames. ], batch size: 68, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:19,960 INFO [train.py:904] (3/8) Epoch 14, batch 5500, loss[loss=0.268, simple_loss=0.3433, pruned_loss=0.09635, over 11621.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.285, pruned_loss=0.0557, over 3157896.81 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:32,388 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0304, 4.1772, 3.9756, 3.7215, 3.5392, 4.0982, 3.8217, 3.7555], device='cuda:3'), covar=tensor([0.0644, 0.0591, 0.0323, 0.0336, 0.1034, 0.0517, 0.0872, 0.0660], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0353, 0.0306, 0.0288, 0.0325, 0.0340, 0.0207, 0.0361], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:49:35,232 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:49:36,901 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4312, 2.9764, 3.0072, 1.9010, 2.6479, 2.1559, 3.0821, 3.2512], device='cuda:3'), covar=tensor([0.0261, 0.0635, 0.0595, 0.1833, 0.0814, 0.0938, 0.0587, 0.0691], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0139, 0.0127, 0.0139, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-29 23:50:23,833 INFO [optim.py:368] (3/8) 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 23:50:36,992 INFO [train.py:904] (3/8) Epoch 14, batch 5550, loss[loss=0.3268, simple_loss=0.3708, pruned_loss=0.1414, over 10930.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2926, pruned_loss=0.06149, over 3130658.17 frames. ], batch size: 246, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:50:51,278 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:51:55,242 INFO [train.py:904] (3/8) Epoch 14, batch 5600, loss[loss=0.2292, simple_loss=0.3084, pruned_loss=0.075, over 15260.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.297, pruned_loss=0.06537, over 3108107.72 frames. ], batch size: 191, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:53:03,059 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.471e+02 3.513e+02 4.095e+02 5.435e+02 1.324e+03, threshold=8.190e+02, percent-clipped=8.0 2023-04-29 23:53:17,474 INFO [train.py:904] (3/8) Epoch 14, batch 5650, loss[loss=0.2433, simple_loss=0.3173, pruned_loss=0.08469, over 16677.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3022, pruned_loss=0.06946, over 3081598.96 frames. ], batch size: 134, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:54:25,228 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 23:54:38,719 INFO [train.py:904] (3/8) Epoch 14, batch 5700, loss[loss=0.2131, simple_loss=0.2968, pruned_loss=0.06476, over 17146.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.3035, pruned_loss=0.07074, over 3085101.19 frames. ], batch size: 46, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:55:45,156 INFO [optim.py:368] (3/8) 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 23:55:59,942 INFO [train.py:904] (3/8) Epoch 14, batch 5750, loss[loss=0.2751, simple_loss=0.3365, pruned_loss=0.1069, over 11299.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3066, pruned_loss=0.07283, over 3040047.56 frames. ], batch size: 247, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:56:56,371 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 23:57:21,149 INFO [train.py:904] (3/8) Epoch 14, batch 5800, loss[loss=0.2567, simple_loss=0.3202, pruned_loss=0.09663, over 12179.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3066, pruned_loss=0.07234, over 3023543.73 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:57:25,488 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1952, 4.1467, 4.0748, 3.4318, 4.1066, 1.6608, 3.9295, 3.7787], device='cuda:3'), covar=tensor([0.0107, 0.0094, 0.0168, 0.0319, 0.0102, 0.2619, 0.0129, 0.0221], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0175, 0.0164, 0.0148, 0.0188, 0.0164, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-29 23:58:26,263 INFO [optim.py:368] (3/8) 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 23:58:41,172 INFO [train.py:904] (3/8) Epoch 14, batch 5850, loss[loss=0.2379, simple_loss=0.3063, pruned_loss=0.08476, over 11630.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3038, pruned_loss=0.06997, over 3044350.29 frames. ], batch size: 248, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:00:03,509 INFO [train.py:904] (3/8) Epoch 14, batch 5900, loss[loss=0.196, simple_loss=0.2865, pruned_loss=0.05278, over 16684.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3027, pruned_loss=0.06911, over 3049483.22 frames. ], batch size: 134, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:01:10,285 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:01:15,088 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:16,307 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1523, 4.1005, 4.0516, 3.2499, 4.0587, 1.5183, 3.8350, 3.6221], device='cuda:3'), covar=tensor([0.0122, 0.0113, 0.0165, 0.0384, 0.0114, 0.2894, 0.0154, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0130, 0.0175, 0.0165, 0.0148, 0.0188, 0.0164, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:01:19,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 00:01:23,363 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-30 00:01:24,299 INFO [train.py:904] (3/8) Epoch 14, batch 5950, loss[loss=0.1979, simple_loss=0.2882, pruned_loss=0.05381, over 17221.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3026, pruned_loss=0.06745, over 3069804.88 frames. ], batch size: 44, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,551 INFO [train.py:904] (3/8) Epoch 14, batch 6000, loss[loss=0.2262, simple_loss=0.3026, pruned_loss=0.0749, over 15343.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3014, pruned_loss=0.06647, over 3087063.06 frames. ], batch size: 190, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,552 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 00:02:55,351 INFO [train.py:938] (3/8) Epoch 14, validation: loss=0.1574, simple_loss=0.2706, pruned_loss=0.0221, over 944034.00 frames. 2023-04-30 00:02:55,352 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-30 00:02:56,944 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:01,566 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:40,119 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8178, 2.8784, 2.7456, 4.6500, 3.5507, 4.2138, 1.5909, 3.1717], device='cuda:3'), covar=tensor([0.1250, 0.0703, 0.1047, 0.0140, 0.0344, 0.0356, 0.1513, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0184, 0.0160, 0.0201, 0.0208, 0.0187, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 00:03:59,082 INFO [optim.py:368] (3/8) 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 00:04:09,568 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7292, 1.3245, 1.7095, 1.6216, 1.7568, 1.9198, 1.5579, 1.7890], device='cuda:3'), covar=tensor([0.0200, 0.0287, 0.0151, 0.0213, 0.0197, 0.0130, 0.0314, 0.0093], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0175, 0.0159, 0.0165, 0.0174, 0.0132, 0.0177, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-30 00:04:18,772 INFO [train.py:904] (3/8) Epoch 14, batch 6050, loss[loss=0.2387, simple_loss=0.3068, pruned_loss=0.08527, over 11972.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3002, pruned_loss=0.06632, over 3076654.63 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:04:38,561 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8948, 4.9405, 4.7137, 4.4075, 4.3388, 4.8284, 4.7236, 4.4832], device='cuda:3'), covar=tensor([0.0949, 0.1072, 0.0359, 0.0389, 0.1155, 0.0665, 0.0515, 0.1071], device='cuda:3'), in_proj_covar=tensor([0.0256, 0.0351, 0.0304, 0.0285, 0.0323, 0.0335, 0.0207, 0.0361], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:04:45,585 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2915, 4.2975, 4.7214, 4.6980, 4.6626, 4.3810, 4.3629, 4.2461], device='cuda:3'), covar=tensor([0.0334, 0.0634, 0.0373, 0.0390, 0.0465, 0.0421, 0.0961, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0353, 0.0376, 0.0373, 0.0357, 0.0422, 0.0400, 0.0490, 0.0318], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 00:05:40,154 INFO [train.py:904] (3/8) Epoch 14, batch 6100, loss[loss=0.2066, simple_loss=0.2988, pruned_loss=0.05726, over 16918.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2998, pruned_loss=0.06525, over 3086907.43 frames. ], batch size: 116, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:06:45,097 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 6150, loss[loss=0.1962, simple_loss=0.2814, pruned_loss=0.05555, over 16863.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2973, pruned_loss=0.06421, over 3097033.48 frames. ], batch size: 116, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:08:16,180 INFO [train.py:904] (3/8) Epoch 14, batch 6200, loss[loss=0.1839, simple_loss=0.2767, pruned_loss=0.04551, over 16705.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2957, pruned_loss=0.06352, over 3095221.12 frames. ], batch size: 76, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:09:17,888 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 6250, loss[loss=0.1933, simple_loss=0.2811, pruned_loss=0.05275, over 16638.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2948, pruned_loss=0.06233, over 3117328.87 frames. ], batch size: 62, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:10:42,267 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:10:46,667 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 6300, loss[loss=0.2188, simple_loss=0.3015, pruned_loss=0.06802, over 16413.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2949, pruned_loss=0.06235, over 3108308.89 frames. ], batch size: 146, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:11:17,779 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:11:52,436 INFO [optim.py:368] (3/8) 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 00:12:05,893 INFO [train.py:904] (3/8) Epoch 14, batch 6350, loss[loss=0.2209, simple_loss=0.2999, pruned_loss=0.07092, over 16925.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2965, pruned_loss=0.06402, over 3101137.10 frames. ], batch size: 109, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:12:37,617 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:12:50,049 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:13:08,055 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9315, 1.7893, 2.4193, 2.7476, 2.6864, 3.1709, 1.9647, 3.0035], device='cuda:3'), covar=tensor([0.0141, 0.0413, 0.0264, 0.0224, 0.0220, 0.0127, 0.0409, 0.0117], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0177, 0.0162, 0.0168, 0.0176, 0.0134, 0.0180, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:13:09,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4452, 3.5043, 3.3021, 3.0348, 3.1096, 3.3957, 3.2953, 3.2668], device='cuda:3'), covar=tensor([0.0592, 0.0511, 0.0240, 0.0258, 0.0463, 0.0363, 0.0980, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0349, 0.0302, 0.0283, 0.0320, 0.0331, 0.0206, 0.0357], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:13:22,146 INFO [train.py:904] (3/8) Epoch 14, batch 6400, loss[loss=0.2033, simple_loss=0.2865, pruned_loss=0.06007, over 16248.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2971, pruned_loss=0.06556, over 3091248.07 frames. ], batch size: 165, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:13:23,650 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-30 00:13:24,497 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5121, 5.5061, 5.3208, 4.7132, 5.4300, 2.3010, 5.1984, 5.2086], device='cuda:3'), covar=tensor([0.0066, 0.0055, 0.0133, 0.0322, 0.0072, 0.2043, 0.0095, 0.0147], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0174, 0.0163, 0.0147, 0.0187, 0.0163, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:13:50,257 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5264, 2.2577, 2.4264, 4.4658, 2.2126, 2.5856, 2.4040, 2.4120], device='cuda:3'), covar=tensor([0.1042, 0.3346, 0.2428, 0.0387, 0.3993, 0.2522, 0.3164, 0.3331], device='cuda:3'), in_proj_covar=tensor([0.0368, 0.0404, 0.0337, 0.0315, 0.0413, 0.0466, 0.0370, 0.0472], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:14:09,312 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 3.125e+02 3.740e+02 4.766e+02 1.292e+03, threshold=7.481e+02, percent-clipped=6.0 2023-04-30 00:14:25,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0461, 2.5666, 2.6795, 1.8901, 2.7400, 2.8567, 2.5268, 2.3900], device='cuda:3'), covar=tensor([0.0752, 0.0217, 0.0224, 0.1006, 0.0112, 0.0254, 0.0378, 0.0429], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0105, 0.0089, 0.0139, 0.0073, 0.0114, 0.0123, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 00:14:37,173 INFO [train.py:904] (3/8) Epoch 14, batch 6450, loss[loss=0.2068, simple_loss=0.2939, pruned_loss=0.05988, over 16967.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2985, pruned_loss=0.0662, over 3058762.63 frames. ], batch size: 109, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:14:37,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5233, 3.5961, 3.3008, 3.0308, 3.1485, 3.4438, 3.2652, 3.3156], device='cuda:3'), covar=tensor([0.0473, 0.0428, 0.0240, 0.0252, 0.0459, 0.0369, 0.1349, 0.0453], device='cuda:3'), in_proj_covar=tensor([0.0252, 0.0348, 0.0301, 0.0282, 0.0319, 0.0330, 0.0204, 0.0357], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:15:46,309 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8393, 1.3390, 1.6938, 1.6294, 1.7947, 1.9215, 1.5948, 1.7554], device='cuda:3'), covar=tensor([0.0178, 0.0308, 0.0156, 0.0232, 0.0204, 0.0142, 0.0290, 0.0092], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0176, 0.0160, 0.0166, 0.0175, 0.0133, 0.0178, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:15:54,740 INFO [train.py:904] (3/8) Epoch 14, batch 6500, loss[loss=0.1992, simple_loss=0.2827, pruned_loss=0.05784, over 16698.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2966, pruned_loss=0.06563, over 3073591.79 frames. ], batch size: 134, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:15:59,654 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0145, 1.9867, 2.4553, 2.9029, 2.7963, 3.3927, 2.0511, 3.2246], device='cuda:3'), covar=tensor([0.0143, 0.0361, 0.0243, 0.0214, 0.0222, 0.0118, 0.0380, 0.0109], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0175, 0.0160, 0.0165, 0.0174, 0.0133, 0.0178, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-30 00:16:06,207 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 00:16:30,462 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-30 00:16:59,451 INFO [optim.py:368] (3/8) 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 00:17:12,091 INFO [train.py:904] (3/8) Epoch 14, batch 6550, loss[loss=0.2693, simple_loss=0.3289, pruned_loss=0.1049, over 11825.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2993, pruned_loss=0.06585, over 3087331.64 frames. ], batch size: 248, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:18:22,246 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:25,702 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:26,459 INFO [train.py:904] (3/8) Epoch 14, batch 6600, loss[loss=0.2556, simple_loss=0.3194, pruned_loss=0.09596, over 11295.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3018, pruned_loss=0.06696, over 3072726.33 frames. ], batch size: 246, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:29,570 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 2.801e+02 3.547e+02 4.436e+02 1.538e+03, threshold=7.095e+02, percent-clipped=5.0 2023-04-30 00:19:33,988 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:38,206 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 6650, loss[loss=0.1852, simple_loss=0.2693, pruned_loss=0.05055, over 16687.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3014, pruned_loss=0.0672, over 3080051.09 frames. ], batch size: 124, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:51,248 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5218, 3.4876, 3.4682, 2.8645, 3.3937, 2.0398, 3.2258, 2.9049], device='cuda:3'), covar=tensor([0.0121, 0.0097, 0.0140, 0.0211, 0.0093, 0.1998, 0.0116, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0173, 0.0162, 0.0147, 0.0187, 0.0162, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:20:19,655 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:20:58,165 INFO [train.py:904] (3/8) Epoch 14, batch 6700, loss[loss=0.2143, simple_loss=0.3011, pruned_loss=0.06375, over 16580.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2994, pruned_loss=0.06665, over 3092028.82 frames. ], batch size: 134, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:21:40,469 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:44,880 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:52,993 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:03,784 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.354e+02 4.045e+02 5.286e+02 1.404e+03, threshold=8.090e+02, percent-clipped=8.0 2023-04-30 00:22:09,133 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 6750, loss[loss=0.1864, simple_loss=0.2657, pruned_loss=0.05355, over 17113.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2971, pruned_loss=0.06621, over 3105756.48 frames. ], batch size: 48, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:23:07,193 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6226, 3.5797, 4.0707, 1.8287, 4.1679, 4.2990, 2.9900, 2.9707], device='cuda:3'), covar=tensor([0.0736, 0.0207, 0.0156, 0.1173, 0.0056, 0.0109, 0.0403, 0.0424], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0138, 0.0072, 0.0113, 0.0123, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 00:23:14,922 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:22,342 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:26,056 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4796, 3.6021, 2.6926, 2.1406, 2.4038, 2.2041, 3.7278, 3.2772], device='cuda:3'), covar=tensor([0.2863, 0.0687, 0.1712, 0.2435, 0.2411, 0.1955, 0.0504, 0.1173], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0263, 0.0292, 0.0292, 0.0286, 0.0232, 0.0279, 0.0311], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 00:23:28,409 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6486, 4.5742, 4.4104, 3.7624, 4.5125, 1.5652, 4.2789, 4.2527], device='cuda:3'), covar=tensor([0.0081, 0.0077, 0.0158, 0.0366, 0.0079, 0.2667, 0.0118, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0129, 0.0174, 0.0163, 0.0147, 0.0188, 0.0163, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:23:29,044 INFO [train.py:904] (3/8) Epoch 14, batch 6800, loss[loss=0.2211, simple_loss=0.3081, pruned_loss=0.06708, over 16899.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2973, pruned_loss=0.06579, over 3120370.77 frames. ], batch size: 109, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:23:38,388 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:24:34,694 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 6850, loss[loss=0.2039, simple_loss=0.3098, pruned_loss=0.04903, over 16727.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2989, pruned_loss=0.06676, over 3107922.16 frames. ], batch size: 76, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:25:52,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8272, 5.1961, 5.3883, 5.1759, 5.1955, 5.7858, 5.2490, 4.9703], device='cuda:3'), covar=tensor([0.1038, 0.1643, 0.2226, 0.1937, 0.2433, 0.0923, 0.1452, 0.2510], device='cuda:3'), in_proj_covar=tensor([0.0371, 0.0522, 0.0573, 0.0442, 0.0591, 0.0591, 0.0453, 0.0600], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 00:26:02,477 INFO [train.py:904] (3/8) Epoch 14, batch 6900, loss[loss=0.2238, simple_loss=0.3077, pruned_loss=0.06996, over 16286.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3016, pruned_loss=0.06638, over 3113475.72 frames. ], batch size: 165, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:08,128 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-30 00:27:10,346 INFO [optim.py:368] (3/8) 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 00:27:22,848 INFO [train.py:904] (3/8) Epoch 14, batch 6950, loss[loss=0.1986, simple_loss=0.2876, pruned_loss=0.05477, over 16520.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3037, pruned_loss=0.06855, over 3084464.53 frames. ], batch size: 75, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:32,485 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5028, 4.7887, 4.5819, 4.5399, 4.2625, 4.3079, 4.2302, 4.8925], device='cuda:3'), covar=tensor([0.1057, 0.0924, 0.0993, 0.0860, 0.0762, 0.1162, 0.1166, 0.0877], device='cuda:3'), in_proj_covar=tensor([0.0583, 0.0716, 0.0595, 0.0521, 0.0454, 0.0465, 0.0603, 0.0551], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:27:33,774 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:27:56,333 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 00:27:59,931 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:38,216 INFO [train.py:904] (3/8) Epoch 14, batch 7000, loss[loss=0.2248, simple_loss=0.2965, pruned_loss=0.07659, over 11730.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3034, pruned_loss=0.06747, over 3087143.69 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:29:06,860 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:29:12,849 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:18,623 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:42,506 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 7050, loss[loss=0.1845, simple_loss=0.2828, pruned_loss=0.04311, over 16864.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3041, pruned_loss=0.06734, over 3077148.78 frames. ], batch size: 102, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:30:09,238 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0819, 5.7021, 5.9192, 5.6040, 5.6102, 6.2246, 5.6779, 5.4354], device='cuda:3'), covar=tensor([0.0831, 0.1626, 0.1923, 0.1846, 0.2284, 0.0812, 0.1376, 0.2291], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0531, 0.0584, 0.0448, 0.0605, 0.0605, 0.0459, 0.0611], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 00:30:33,407 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:49,914 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:58,268 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 7100, loss[loss=0.2285, simple_loss=0.296, pruned_loss=0.0805, over 11213.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3033, pruned_loss=0.06772, over 3052232.19 frames. ], batch size: 250, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:31:15,009 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:32:18,352 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 7150, loss[loss=0.2064, simple_loss=0.3006, pruned_loss=0.05616, over 16700.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.302, pruned_loss=0.06759, over 3066322.65 frames. ], batch size: 83, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:35,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-04-30 00:33:47,875 INFO [train.py:904] (3/8) Epoch 14, batch 7200, loss[loss=0.1982, simple_loss=0.2894, pruned_loss=0.05351, over 16771.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2994, pruned_loss=0.06567, over 3069163.08 frames. ], batch size: 39, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:34:49,519 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8523, 1.3178, 1.6449, 1.6399, 1.7804, 1.9042, 1.5051, 1.7451], device='cuda:3'), covar=tensor([0.0179, 0.0313, 0.0180, 0.0244, 0.0207, 0.0158, 0.0347, 0.0112], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0159, 0.0165, 0.0173, 0.0130, 0.0177, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-30 00:34:55,353 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 7250, loss[loss=0.1751, simple_loss=0.263, pruned_loss=0.04364, over 17252.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2968, pruned_loss=0.06434, over 3069622.34 frames. ], batch size: 52, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:35:51,386 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6738, 4.9752, 5.1723, 4.9407, 4.9508, 5.5465, 4.9187, 4.7017], device='cuda:3'), covar=tensor([0.1052, 0.1716, 0.1819, 0.1770, 0.2535, 0.0986, 0.1765, 0.2712], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0523, 0.0571, 0.0442, 0.0592, 0.0595, 0.0451, 0.0600], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 00:36:01,955 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:36:22,192 INFO [train.py:904] (3/8) Epoch 14, batch 7300, loss[loss=0.2065, simple_loss=0.2946, pruned_loss=0.0592, over 16610.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2957, pruned_loss=0.06393, over 3069679.77 frames. ], batch size: 62, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:42,737 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:37:00,598 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:20,391 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:37:37,043 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:40,532 INFO [train.py:904] (3/8) Epoch 14, batch 7350, loss[loss=0.2508, simple_loss=0.3104, pruned_loss=0.0956, over 10918.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2964, pruned_loss=0.06474, over 3037224.61 frames. ], batch size: 247, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:37:48,835 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3416, 1.4853, 1.9521, 2.1689, 2.3062, 2.6012, 1.5908, 2.4287], device='cuda:3'), covar=tensor([0.0194, 0.0456, 0.0258, 0.0310, 0.0253, 0.0157, 0.0442, 0.0112], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0174, 0.0158, 0.0164, 0.0173, 0.0131, 0.0176, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-30 00:37:50,739 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 00:38:28,543 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2590, 2.2714, 2.2439, 4.0479, 1.9955, 2.6739, 2.3112, 2.3985], device='cuda:3'), covar=tensor([0.1128, 0.3252, 0.2498, 0.0453, 0.4021, 0.2061, 0.3240, 0.2992], device='cuda:3'), in_proj_covar=tensor([0.0372, 0.0411, 0.0342, 0.0320, 0.0420, 0.0472, 0.0377, 0.0480], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:38:36,841 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:36,874 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:38:44,960 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:57,314 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:59,353 INFO [train.py:904] (3/8) Epoch 14, batch 7400, loss[loss=0.1983, simple_loss=0.2871, pruned_loss=0.05477, over 16824.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2976, pruned_loss=0.06586, over 3022989.55 frames. ], batch size: 102, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:39:01,778 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:39:22,433 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6332, 4.1904, 4.2086, 2.7818, 3.7064, 4.2265, 3.7209, 2.4997], device='cuda:3'), covar=tensor([0.0407, 0.0034, 0.0037, 0.0321, 0.0082, 0.0082, 0.0074, 0.0340], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0072, 0.0073, 0.0129, 0.0085, 0.0095, 0.0083, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 00:39:52,862 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 00:40:08,879 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9534, 5.2352, 4.9765, 4.9037, 4.6714, 4.6221, 4.6382, 5.2967], device='cuda:3'), covar=tensor([0.1008, 0.0805, 0.0925, 0.0876, 0.0866, 0.0854, 0.1061, 0.0887], device='cuda:3'), in_proj_covar=tensor([0.0582, 0.0714, 0.0594, 0.0520, 0.0454, 0.0468, 0.0602, 0.0551], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:40:18,576 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:19,476 INFO [train.py:904] (3/8) Epoch 14, batch 7450, loss[loss=0.2461, simple_loss=0.3103, pruned_loss=0.09097, over 11641.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2984, pruned_loss=0.06631, over 3058444.40 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:41:43,385 INFO [train.py:904] (3/8) Epoch 14, batch 7500, loss[loss=0.1914, simple_loss=0.2799, pruned_loss=0.05146, over 16917.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2997, pruned_loss=0.06605, over 3070830.26 frames. ], batch size: 109, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:42:53,285 INFO [optim.py:368] (3/8) 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 00:43:02,954 INFO [train.py:904] (3/8) Epoch 14, batch 7550, loss[loss=0.2007, simple_loss=0.2874, pruned_loss=0.057, over 16429.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2987, pruned_loss=0.06608, over 3080181.47 frames. ], batch size: 68, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:44:19,041 INFO [train.py:904] (3/8) Epoch 14, batch 7600, loss[loss=0.2085, simple_loss=0.3013, pruned_loss=0.0578, over 17126.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2975, pruned_loss=0.06557, over 3091860.71 frames. ], batch size: 49, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:44:39,710 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:44:48,983 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:06,335 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:23,448 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:45:34,159 INFO [train.py:904] (3/8) Epoch 14, batch 7650, loss[loss=0.2211, simple_loss=0.301, pruned_loss=0.07062, over 16786.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2981, pruned_loss=0.06586, over 3104187.19 frames. ], batch size: 124, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:45:50,717 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:19,133 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:46:19,243 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:24,892 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0576, 1.9281, 2.1209, 3.6921, 1.9269, 2.1884, 2.0942, 2.0462], device='cuda:3'), covar=tensor([0.1334, 0.3903, 0.2755, 0.0622, 0.4842, 0.2933, 0.3543, 0.4042], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0408, 0.0339, 0.0318, 0.0417, 0.0468, 0.0373, 0.0474], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:46:37,202 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:49,447 INFO [train.py:904] (3/8) Epoch 14, batch 7700, loss[loss=0.1867, simple_loss=0.2817, pruned_loss=0.04589, over 16857.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2981, pruned_loss=0.06603, over 3107609.52 frames. ], batch size: 96, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:47:57,376 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.201e+02 3.391e+02 3.997e+02 4.581e+02 8.153e+02, threshold=7.994e+02, percent-clipped=6.0 2023-04-30 00:48:06,809 INFO [train.py:904] (3/8) Epoch 14, batch 7750, loss[loss=0.2053, simple_loss=0.3016, pruned_loss=0.05444, over 16773.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2976, pruned_loss=0.065, over 3128842.51 frames. ], batch size: 83, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:24,616 INFO [train.py:904] (3/8) Epoch 14, batch 7800, loss[loss=0.1951, simple_loss=0.2901, pruned_loss=0.05003, over 16946.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2987, pruned_loss=0.06612, over 3111706.97 frames. ], batch size: 90, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:51,661 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:22,511 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 14, batch 7850, loss[loss=0.209, simple_loss=0.3104, pruned_loss=0.0538, over 17024.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2997, pruned_loss=0.0658, over 3103044.10 frames. ], batch size: 50, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:50:46,025 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:47,144 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3578, 4.3262, 4.2245, 3.5232, 4.2446, 1.6782, 4.0021, 3.9002], device='cuda:3'), covar=tensor([0.0084, 0.0076, 0.0137, 0.0308, 0.0085, 0.2458, 0.0121, 0.0173], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0126, 0.0172, 0.0160, 0.0145, 0.0187, 0.0160, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:50:48,442 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:24,569 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:31,469 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 00:51:53,572 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,614 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:56,203 INFO [train.py:904] (3/8) Epoch 14, batch 7900, loss[loss=0.2143, simple_loss=0.304, pruned_loss=0.06233, over 16678.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.299, pruned_loss=0.06545, over 3096194.96 frames. ], batch size: 57, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:52:15,821 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:18,323 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:37,573 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 00:53:01,516 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 14, batch 7950, loss[loss=0.205, simple_loss=0.2878, pruned_loss=0.06106, over 16191.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2999, pruned_loss=0.06669, over 3074233.96 frames. ], batch size: 165, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:53:27,994 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:33,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2035, 4.1753, 4.0907, 3.3792, 4.1069, 1.5720, 3.8876, 3.7643], device='cuda:3'), covar=tensor([0.0112, 0.0091, 0.0152, 0.0356, 0.0105, 0.2723, 0.0141, 0.0222], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0127, 0.0173, 0.0162, 0.0146, 0.0188, 0.0161, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:53:37,124 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8896, 2.6614, 2.5678, 1.8070, 2.4648, 2.6911, 2.5564, 1.7489], device='cuda:3'), covar=tensor([0.0409, 0.0070, 0.0070, 0.0327, 0.0122, 0.0111, 0.0091, 0.0420], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0072, 0.0073, 0.0129, 0.0085, 0.0095, 0.0083, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 00:53:51,344 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:59,012 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:54:08,356 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:13,648 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:18,368 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 8000, loss[loss=0.201, simple_loss=0.2879, pruned_loss=0.05706, over 16542.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.301, pruned_loss=0.06841, over 3035172.02 frames. ], batch size: 68, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:55:10,469 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:34,104 INFO [optim.py:368] (3/8) 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 00:55:46,024 INFO [train.py:904] (3/8) Epoch 14, batch 8050, loss[loss=0.2066, simple_loss=0.3014, pruned_loss=0.0559, over 16816.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3004, pruned_loss=0.06758, over 3048645.45 frames. ], batch size: 102, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:55:47,216 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6607, 3.9898, 4.1097, 2.3097, 3.4295, 2.8883, 3.9508, 4.2579], device='cuda:3'), covar=tensor([0.0279, 0.0683, 0.0544, 0.1948, 0.0811, 0.0845, 0.0680, 0.0908], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0152, 0.0161, 0.0146, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 00:56:59,040 INFO [train.py:904] (3/8) Epoch 14, batch 8100, loss[loss=0.2402, simple_loss=0.3102, pruned_loss=0.08509, over 11271.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2991, pruned_loss=0.06624, over 3058471.56 frames. ], batch size: 248, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:57:27,236 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 00:57:42,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3591, 4.6382, 4.4429, 4.3678, 4.1170, 4.1250, 4.1242, 4.6763], device='cuda:3'), covar=tensor([0.1124, 0.0780, 0.0927, 0.0845, 0.0846, 0.1456, 0.1095, 0.0886], device='cuda:3'), in_proj_covar=tensor([0.0581, 0.0713, 0.0591, 0.0514, 0.0453, 0.0465, 0.0598, 0.0546], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:58:02,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7962, 4.7813, 4.5985, 3.9194, 4.6785, 1.6420, 4.4377, 4.4061], device='cuda:3'), covar=tensor([0.0077, 0.0064, 0.0145, 0.0344, 0.0074, 0.2571, 0.0113, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0128, 0.0175, 0.0163, 0.0147, 0.0189, 0.0162, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 00:58:06,569 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.001e+02 3.052e+02 3.743e+02 5.058e+02 1.059e+03, threshold=7.487e+02, percent-clipped=6.0 2023-04-30 00:58:17,058 INFO [train.py:904] (3/8) Epoch 14, batch 8150, loss[loss=0.2018, simple_loss=0.2807, pruned_loss=0.06141, over 16193.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2961, pruned_loss=0.06474, over 3079174.94 frames. ], batch size: 165, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:58:18,890 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:29,417 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:50,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8035, 2.9082, 2.5650, 4.6273, 3.6293, 4.1101, 1.6852, 2.9910], device='cuda:3'), covar=tensor([0.1316, 0.0704, 0.1184, 0.0145, 0.0311, 0.0382, 0.1519, 0.0839], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0164, 0.0206, 0.0213, 0.0190, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 00:58:52,072 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:34,234 INFO [train.py:904] (3/8) Epoch 14, batch 8200, loss[loss=0.1965, simple_loss=0.2851, pruned_loss=0.05395, over 16710.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2936, pruned_loss=0.06421, over 3069542.26 frames. ], batch size: 134, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:59:47,416 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:53,806 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:57,253 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8501, 4.0966, 3.1911, 2.2979, 2.8408, 2.6157, 4.3058, 3.6582], device='cuda:3'), covar=tensor([0.2710, 0.0621, 0.1547, 0.2713, 0.2487, 0.1843, 0.0436, 0.1119], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0261, 0.0289, 0.0291, 0.0285, 0.0232, 0.0274, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:00:02,734 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 01:00:05,605 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:20,713 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3623, 2.9746, 3.1406, 1.7814, 3.2539, 3.2929, 2.8248, 2.7613], device='cuda:3'), covar=tensor([0.0665, 0.0216, 0.0214, 0.1175, 0.0087, 0.0197, 0.0415, 0.0375], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0104, 0.0091, 0.0138, 0.0072, 0.0113, 0.0124, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 01:00:44,842 INFO [optim.py:368] (3/8) 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 01:00:50,799 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5082, 3.3524, 3.6157, 1.8064, 3.7294, 3.7411, 2.9910, 2.9937], device='cuda:3'), covar=tensor([0.0645, 0.0182, 0.0146, 0.1148, 0.0057, 0.0154, 0.0359, 0.0348], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0104, 0.0090, 0.0138, 0.0072, 0.0112, 0.0123, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 01:00:55,629 INFO [train.py:904] (3/8) Epoch 14, batch 8250, loss[loss=0.1822, simple_loss=0.269, pruned_loss=0.04766, over 12309.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2932, pruned_loss=0.06202, over 3066124.36 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:01:02,324 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:37,188 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:55,148 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 8300, loss[loss=0.183, simple_loss=0.2844, pruned_loss=0.04079, over 16503.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2898, pruned_loss=0.05911, over 3033838.07 frames. ], batch size: 75, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:02:18,954 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6772, 2.1321, 1.7937, 1.9856, 2.4123, 2.2019, 2.4365, 2.6490], device='cuda:3'), covar=tensor([0.0134, 0.0341, 0.0430, 0.0364, 0.0236, 0.0314, 0.0191, 0.0212], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0207, 0.0203, 0.0203, 0.0208, 0.0208, 0.0210, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:02:28,936 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:59,484 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4723, 3.0512, 2.6637, 2.1806, 2.1564, 2.2408, 2.9133, 2.9059], device='cuda:3'), covar=tensor([0.2355, 0.0762, 0.1521, 0.2407, 0.2313, 0.1993, 0.0476, 0.1204], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0256, 0.0285, 0.0286, 0.0279, 0.0229, 0.0270, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:03:12,772 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:27,883 INFO [optim.py:368] (3/8) 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 01:03:30,284 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5349, 3.5304, 3.4489, 2.7586, 3.3826, 2.0932, 3.2278, 2.9088], device='cuda:3'), covar=tensor([0.0123, 0.0105, 0.0147, 0.0205, 0.0097, 0.2130, 0.0114, 0.0179], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0126, 0.0173, 0.0159, 0.0144, 0.0187, 0.0160, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:03:32,944 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7536, 2.3166, 2.3438, 4.5138, 2.2189, 2.7913, 2.3402, 2.5254], device='cuda:3'), covar=tensor([0.0812, 0.3291, 0.2535, 0.0299, 0.3902, 0.2283, 0.3367, 0.3191], device='cuda:3'), in_proj_covar=tensor([0.0365, 0.0403, 0.0336, 0.0311, 0.0413, 0.0462, 0.0368, 0.0470], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:03:36,677 INFO [train.py:904] (3/8) Epoch 14, batch 8350, loss[loss=0.1939, simple_loss=0.2913, pruned_loss=0.04828, over 16762.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2885, pruned_loss=0.057, over 3008583.10 frames. ], batch size: 124, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:03:44,815 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.51 vs. limit=2.0 2023-04-30 01:04:07,899 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:04:25,895 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-30 01:04:57,945 INFO [train.py:904] (3/8) Epoch 14, batch 8400, loss[loss=0.1718, simple_loss=0.273, pruned_loss=0.03527, over 16845.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2858, pruned_loss=0.05477, over 3005370.05 frames. ], batch size: 102, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:05:54,609 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.349e+02 2.839e+02 3.572e+02 6.227e+02, threshold=5.678e+02, percent-clipped=3.0 2023-04-30 01:06:18,701 INFO [train.py:904] (3/8) Epoch 14, batch 8450, loss[loss=0.1629, simple_loss=0.264, pruned_loss=0.03087, over 16851.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2839, pruned_loss=0.0529, over 3018924.07 frames. ], batch size: 102, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:06:55,810 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:27,240 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:32,341 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:32,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-30 01:07:37,964 INFO [train.py:904] (3/8) Epoch 14, batch 8500, loss[loss=0.1634, simple_loss=0.2485, pruned_loss=0.03917, over 12079.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2801, pruned_loss=0.05026, over 3028798.56 frames. ], batch size: 247, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:07:50,015 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:51,359 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:54,681 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:01,436 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:13,346 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:38,538 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:51,329 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 8550, loss[loss=0.1762, simple_loss=0.2597, pruned_loss=0.04634, over 12042.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2778, pruned_loss=0.04941, over 3023465.44 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:09:11,902 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:15,277 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:17,274 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:19,130 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:39,024 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:41,515 INFO [train.py:904] (3/8) Epoch 14, batch 8600, loss[loss=0.1829, simple_loss=0.2766, pruned_loss=0.0446, over 16676.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2782, pruned_loss=0.0482, over 3037889.38 frames. ], batch size: 76, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:10:46,487 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:19,562 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:12:06,768 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 8650, loss[loss=0.1702, simple_loss=0.2672, pruned_loss=0.03663, over 15336.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2765, pruned_loss=0.04706, over 3030635.82 frames. ], batch size: 191, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:12:38,367 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0968, 3.3357, 3.7295, 2.1021, 2.9816, 2.3061, 3.5393, 3.5032], device='cuda:3'), covar=tensor([0.0226, 0.0745, 0.0429, 0.1804, 0.0698, 0.0938, 0.0555, 0.0878], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0146, 0.0155, 0.0142, 0.0134, 0.0123, 0.0134, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 01:12:52,331 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:14:05,318 INFO [train.py:904] (3/8) Epoch 14, batch 8700, loss[loss=0.2001, simple_loss=0.283, pruned_loss=0.05863, over 16886.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2736, pruned_loss=0.04601, over 3036395.36 frames. ], batch size: 116, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:14:53,873 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1141, 2.7203, 2.9804, 1.9788, 2.7418, 2.1491, 2.8139, 2.8876], device='cuda:3'), covar=tensor([0.0252, 0.0809, 0.0438, 0.1747, 0.0655, 0.0874, 0.0549, 0.0746], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0144, 0.0154, 0.0141, 0.0133, 0.0121, 0.0133, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 01:15:24,088 INFO [optim.py:368] (3/8) 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 01:15:38,741 INFO [train.py:904] (3/8) Epoch 14, batch 8750, loss[loss=0.1803, simple_loss=0.2772, pruned_loss=0.04169, over 15243.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2731, pruned_loss=0.04551, over 3031904.54 frames. ], batch size: 191, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:16:30,801 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:16:31,233 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 01:16:43,729 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1586, 5.5054, 5.3157, 5.3133, 4.9803, 4.9897, 4.8552, 5.5888], device='cuda:3'), covar=tensor([0.1038, 0.0705, 0.0677, 0.0619, 0.0656, 0.0608, 0.0933, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0695, 0.0570, 0.0502, 0.0440, 0.0453, 0.0583, 0.0528], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:16:45,816 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7715, 3.1553, 3.3672, 1.8149, 2.8564, 2.3055, 3.2561, 3.2900], device='cuda:3'), covar=tensor([0.0287, 0.0764, 0.0530, 0.2051, 0.0789, 0.0908, 0.0637, 0.0785], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0144, 0.0154, 0.0141, 0.0133, 0.0122, 0.0133, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 01:16:58,571 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-30 01:17:11,076 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:27,839 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1991, 4.0562, 4.2891, 4.4039, 4.5471, 4.1169, 4.5171, 4.5654], device='cuda:3'), covar=tensor([0.1650, 0.1200, 0.1526, 0.0730, 0.0545, 0.1089, 0.0556, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0536, 0.0658, 0.0784, 0.0681, 0.0514, 0.0531, 0.0535, 0.0629], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:17:28,573 INFO [train.py:904] (3/8) Epoch 14, batch 8800, loss[loss=0.1867, simple_loss=0.2804, pruned_loss=0.04656, over 16772.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.271, pruned_loss=0.04394, over 3048823.72 frames. ], batch size: 124, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:17:35,502 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-30 01:17:43,085 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:56,950 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:18:31,185 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4694, 4.4324, 4.2749, 3.7472, 4.3206, 1.5299, 4.1491, 4.0318], device='cuda:3'), covar=tensor([0.0063, 0.0066, 0.0137, 0.0236, 0.0080, 0.2546, 0.0101, 0.0200], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0124, 0.0170, 0.0154, 0.0142, 0.0186, 0.0157, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:18:35,743 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:00,060 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 8850, loss[loss=0.1726, simple_loss=0.2737, pruned_loss=0.03577, over 15325.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2738, pruned_loss=0.04344, over 3056837.56 frames. ], batch size: 190, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:19:23,455 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:27,224 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:37,089 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:27,422 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1511, 3.5500, 3.4713, 2.4800, 3.1820, 3.5686, 3.3787, 1.9709], device='cuda:3'), covar=tensor([0.0427, 0.0030, 0.0036, 0.0305, 0.0078, 0.0050, 0.0057, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0127, 0.0069, 0.0070, 0.0126, 0.0083, 0.0092, 0.0081, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 01:20:44,016 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:58,216 INFO [train.py:904] (3/8) Epoch 14, batch 8900, loss[loss=0.1831, simple_loss=0.2738, pruned_loss=0.04619, over 16832.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2744, pruned_loss=0.04296, over 3051985.94 frames. ], batch size: 116, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:21:25,585 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:21:34,673 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:22:03,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3986, 3.2668, 3.5354, 1.7670, 3.6865, 3.7244, 2.9523, 2.8082], device='cuda:3'), covar=tensor([0.0697, 0.0178, 0.0126, 0.1134, 0.0048, 0.0105, 0.0351, 0.0404], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0100, 0.0085, 0.0133, 0.0068, 0.0107, 0.0119, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 01:22:19,991 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4095, 4.7488, 5.0125, 5.0068, 4.9664, 4.6721, 4.2822, 4.5093], device='cuda:3'), covar=tensor([0.0583, 0.0640, 0.0569, 0.0584, 0.0791, 0.0553, 0.1577, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0354, 0.0355, 0.0338, 0.0399, 0.0377, 0.0459, 0.0300], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 01:22:46,890 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 8950, loss[loss=0.1637, simple_loss=0.2563, pruned_loss=0.03557, over 16789.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2739, pruned_loss=0.0431, over 3065402.82 frames. ], batch size: 124, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:23:08,808 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 01:23:29,132 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:23:32,806 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:38,411 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 9000, loss[loss=0.1707, simple_loss=0.2582, pruned_loss=0.04155, over 12027.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2705, pruned_loss=0.04196, over 3038604.76 frames. ], batch size: 248, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:24:48,129 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 01:24:58,095 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17808MB 2023-04-30 01:25:21,238 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:25:52,429 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:26:35,360 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8763, 3.7586, 3.9611, 3.7446, 3.8571, 4.2991, 4.0230, 3.6985], device='cuda:3'), covar=tensor([0.1908, 0.2361, 0.1967, 0.2491, 0.2963, 0.1612, 0.1386, 0.2618], device='cuda:3'), in_proj_covar=tensor([0.0352, 0.0495, 0.0543, 0.0420, 0.0562, 0.0574, 0.0433, 0.0570], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 01:26:39,921 INFO [train.py:904] (3/8) Epoch 14, batch 9050, loss[loss=0.1671, simple_loss=0.2546, pruned_loss=0.03981, over 17007.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2717, pruned_loss=0.04288, over 3051642.25 frames. ], batch size: 41, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:26:55,155 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:08,221 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:25,839 INFO [train.py:904] (3/8) Epoch 14, batch 9100, loss[loss=0.1863, simple_loss=0.281, pruned_loss=0.04585, over 16801.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2717, pruned_loss=0.04335, over 3062011.82 frames. ], batch size: 124, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:28:36,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9403, 1.9353, 2.5126, 2.8909, 2.6509, 3.2276, 2.1578, 3.2431], device='cuda:3'), covar=tensor([0.0156, 0.0412, 0.0241, 0.0224, 0.0252, 0.0137, 0.0375, 0.0092], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0156, 0.0161, 0.0172, 0.0129, 0.0175, 0.0118], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:3') 2023-04-30 01:29:30,788 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:29:58,704 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:30:13,177 INFO [optim.py:368] (3/8) 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 01:30:22,841 INFO [train.py:904] (3/8) Epoch 14, batch 9150, loss[loss=0.1712, simple_loss=0.2672, pruned_loss=0.03756, over 16739.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2729, pruned_loss=0.04317, over 3066788.60 frames. ], batch size: 83, lr: 4.80e-03, grad_scale: 4.0 2023-04-30 01:31:51,326 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:31:54,375 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 9200, loss[loss=0.1883, simple_loss=0.2764, pruned_loss=0.05012, over 17104.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2683, pruned_loss=0.04225, over 3064787.52 frames. ], batch size: 53, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:32:18,848 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 01:32:28,726 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:28,821 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:50,721 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4684, 3.4489, 3.7581, 1.8490, 3.9007, 3.9961, 3.0461, 2.8856], device='cuda:3'), covar=tensor([0.0720, 0.0196, 0.0151, 0.1152, 0.0047, 0.0090, 0.0358, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0100, 0.0085, 0.0134, 0.0069, 0.0108, 0.0120, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 01:33:23,826 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:31,525 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 9250, loss[loss=0.168, simple_loss=0.2578, pruned_loss=0.03905, over 16528.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.268, pruned_loss=0.04222, over 3053065.68 frames. ], batch size: 62, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:33:46,104 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141204.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:34:03,409 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:34:49,807 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 01:35:30,116 INFO [train.py:904] (3/8) Epoch 14, batch 9300, loss[loss=0.1755, simple_loss=0.258, pruned_loss=0.04654, over 12224.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2663, pruned_loss=0.04167, over 3046235.99 frames. ], batch size: 248, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:35:46,407 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 01:36:20,196 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:36:33,403 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 01:37:05,690 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.032e+02 2.564e+02 3.502e+02 9.246e+02, threshold=5.129e+02, percent-clipped=2.0 2023-04-30 01:37:14,521 INFO [train.py:904] (3/8) Epoch 14, batch 9350, loss[loss=0.1816, simple_loss=0.271, pruned_loss=0.04613, over 17016.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2659, pruned_loss=0.04165, over 3031254.99 frames. ], batch size: 109, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:37:16,940 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:08,721 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:54,829 INFO [train.py:904] (3/8) Epoch 14, batch 9400, loss[loss=0.1677, simple_loss=0.269, pruned_loss=0.03322, over 15365.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2656, pruned_loss=0.04119, over 3028522.90 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:38:55,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7606, 2.5472, 2.3373, 3.8152, 2.3136, 3.9266, 1.3163, 2.9240], device='cuda:3'), covar=tensor([0.1383, 0.0724, 0.1181, 0.0133, 0.0111, 0.0354, 0.1706, 0.0708], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0160, 0.0180, 0.0156, 0.0189, 0.0204, 0.0185, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-30 01:39:17,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9627, 4.2125, 4.0757, 4.0713, 3.7122, 3.8014, 3.8968, 4.2119], device='cuda:3'), covar=tensor([0.1103, 0.1027, 0.0977, 0.0748, 0.0813, 0.1594, 0.0902, 0.1023], device='cuda:3'), in_proj_covar=tensor([0.0556, 0.0685, 0.0560, 0.0493, 0.0436, 0.0448, 0.0576, 0.0523], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:39:49,377 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:12,081 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:25,322 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 14, batch 9450, loss[loss=0.1668, simple_loss=0.2561, pruned_loss=0.0387, over 12412.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2669, pruned_loss=0.04109, over 3036901.36 frames. ], batch size: 248, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:40:38,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8956, 2.7705, 2.5895, 1.9716, 2.4405, 2.8056, 2.6743, 1.9271], device='cuda:3'), covar=tensor([0.0378, 0.0053, 0.0058, 0.0310, 0.0130, 0.0081, 0.0096, 0.0358], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0070, 0.0072, 0.0128, 0.0084, 0.0093, 0.0082, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 01:41:03,099 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 01:41:24,365 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:41:29,126 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 01:41:40,220 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 01:41:54,100 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:14,326 INFO [train.py:904] (3/8) Epoch 14, batch 9500, loss[loss=0.1502, simple_loss=0.2392, pruned_loss=0.03064, over 13095.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2662, pruned_loss=0.04059, over 3038193.95 frames. ], batch size: 246, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:42:28,373 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 01:42:43,594 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:59,930 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 01:43:08,415 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 01:43:21,772 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6174, 3.7113, 2.1734, 4.1129, 2.6741, 4.0212, 2.4002, 2.9333], device='cuda:3'), covar=tensor([0.0242, 0.0348, 0.1575, 0.0218, 0.0844, 0.0495, 0.1448, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0161, 0.0184, 0.0132, 0.0163, 0.0198, 0.0191, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 01:43:47,717 INFO [optim.py:368] (3/8) 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 01:43:55,180 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 9550, loss[loss=0.1858, simple_loss=0.2868, pruned_loss=0.0424, over 15427.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2659, pruned_loss=0.04043, over 3046326.13 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:44:01,086 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:44:23,849 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:45:41,072 INFO [train.py:904] (3/8) Epoch 14, batch 9600, loss[loss=0.1995, simple_loss=0.3068, pruned_loss=0.0461, over 15515.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2671, pruned_loss=0.0407, over 3050036.68 frames. ], batch size: 191, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:45:41,659 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3016, 3.0352, 2.5631, 2.1647, 2.1843, 2.0509, 3.0154, 2.8512], device='cuda:3'), covar=tensor([0.2476, 0.0822, 0.1632, 0.2477, 0.2618, 0.2479, 0.0561, 0.1347], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0252, 0.0281, 0.0280, 0.0264, 0.0226, 0.0264, 0.0294], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:45:42,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4380, 5.4425, 5.1859, 4.8066, 5.1956, 1.9828, 4.9970, 5.1490], device='cuda:3'), covar=tensor([0.0077, 0.0084, 0.0183, 0.0250, 0.0087, 0.2317, 0.0117, 0.0146], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0123, 0.0165, 0.0150, 0.0140, 0.0184, 0.0155, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:45:46,874 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0761, 3.4887, 3.5563, 1.9525, 2.8952, 2.3373, 3.3785, 3.5669], device='cuda:3'), covar=tensor([0.0246, 0.0630, 0.0533, 0.1859, 0.0743, 0.0915, 0.0653, 0.0758], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0143, 0.0156, 0.0144, 0.0135, 0.0124, 0.0135, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 01:46:06,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2412, 3.3817, 3.6238, 3.5949, 3.6181, 3.4378, 3.4659, 3.4869], device='cuda:3'), covar=tensor([0.0375, 0.0685, 0.0457, 0.0499, 0.0517, 0.0499, 0.0770, 0.0501], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0350, 0.0351, 0.0335, 0.0395, 0.0373, 0.0450, 0.0297], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:3') 2023-04-30 01:46:22,421 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 14, batch 9650, loss[loss=0.1575, simple_loss=0.2526, pruned_loss=0.03118, over 16602.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2686, pruned_loss=0.041, over 3046059.15 frames. ], batch size: 68, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:47:34,613 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:47:35,047 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-30 01:48:16,937 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:48:39,781 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6098, 2.2491, 2.3437, 4.3689, 2.1868, 2.6572, 2.3868, 2.5250], device='cuda:3'), covar=tensor([0.0972, 0.3590, 0.2629, 0.0372, 0.4054, 0.2368, 0.3328, 0.3133], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0397, 0.0334, 0.0309, 0.0408, 0.0451, 0.0362, 0.0460], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:49:16,994 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 9700, loss[loss=0.1591, simple_loss=0.2574, pruned_loss=0.03041, over 16695.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2677, pruned_loss=0.04082, over 3043796.76 frames. ], batch size: 83, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:49:45,675 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 01:50:27,993 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:51:00,300 INFO [train.py:904] (3/8) Epoch 14, batch 9750, loss[loss=0.1618, simple_loss=0.2602, pruned_loss=0.03168, over 16597.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2666, pruned_loss=0.04087, over 3056059.85 frames. ], batch size: 62, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:52:38,206 INFO [train.py:904] (3/8) Epoch 14, batch 9800, loss[loss=0.1709, simple_loss=0.2729, pruned_loss=0.03449, over 16793.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2672, pruned_loss=0.04039, over 3054999.76 frames. ], batch size: 124, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:52:38,753 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5135, 4.8373, 4.6625, 4.6565, 4.3584, 4.3598, 4.2457, 4.9125], device='cuda:3'), covar=tensor([0.1077, 0.0775, 0.0839, 0.0719, 0.0763, 0.1027, 0.1058, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0694, 0.0569, 0.0503, 0.0443, 0.0454, 0.0583, 0.0528], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:52:48,824 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3140, 2.1757, 2.1354, 4.2423, 1.9639, 2.4950, 2.2621, 2.3314], device='cuda:3'), covar=tensor([0.1105, 0.3665, 0.2868, 0.0397, 0.4409, 0.2459, 0.3574, 0.3268], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0397, 0.0334, 0.0309, 0.0410, 0.0452, 0.0362, 0.0461], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:52:55,433 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0894, 5.0746, 4.9118, 4.5207, 4.5874, 5.0182, 4.9384, 4.6603], device='cuda:3'), covar=tensor([0.0561, 0.0549, 0.0273, 0.0280, 0.0956, 0.0437, 0.0295, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0328, 0.0289, 0.0268, 0.0297, 0.0312, 0.0199, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-30 01:53:18,123 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 01:53:30,510 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2923, 2.1341, 2.1739, 3.9839, 2.0803, 2.4676, 2.2729, 2.2785], device='cuda:3'), covar=tensor([0.1025, 0.3570, 0.2725, 0.0421, 0.4102, 0.2401, 0.3310, 0.3317], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0397, 0.0334, 0.0309, 0.0409, 0.0452, 0.0363, 0.0460], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:54:13,232 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:54:17,223 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 9850, loss[loss=0.1865, simple_loss=0.288, pruned_loss=0.04251, over 16735.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2688, pruned_loss=0.04031, over 3069203.22 frames. ], batch size: 134, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:54:42,688 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4409, 3.4667, 2.7084, 2.0791, 2.1787, 2.2930, 3.5993, 3.1028], device='cuda:3'), covar=tensor([0.2778, 0.0590, 0.1577, 0.2573, 0.2593, 0.1902, 0.0322, 0.1125], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0252, 0.0283, 0.0281, 0.0265, 0.0227, 0.0263, 0.0294], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 01:56:00,490 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-04-30 01:56:04,788 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 14, batch 9900, loss[loss=0.1631, simple_loss=0.2663, pruned_loss=0.02991, over 16837.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2689, pruned_loss=0.04021, over 3061891.88 frames. ], batch size: 102, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:03,748 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.142e+02 2.825e+02 3.311e+02 8.255e+02, threshold=5.650e+02, percent-clipped=2.0 2023-04-30 01:58:13,950 INFO [train.py:904] (3/8) Epoch 14, batch 9950, loss[loss=0.1899, simple_loss=0.2868, pruned_loss=0.04645, over 17055.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2705, pruned_loss=0.04026, over 3069368.50 frames. ], batch size: 53, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:48,333 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:21,140 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:45,323 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9307, 2.2518, 1.9074, 2.0061, 2.5605, 2.3231, 2.5636, 2.7142], device='cuda:3'), covar=tensor([0.0114, 0.0370, 0.0440, 0.0441, 0.0227, 0.0327, 0.0191, 0.0223], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0209, 0.0203, 0.0204, 0.0208, 0.0209, 0.0206, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:00:05,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0061, 5.3013, 5.0952, 5.1103, 4.8594, 4.8187, 4.6723, 5.3662], device='cuda:3'), covar=tensor([0.0980, 0.0726, 0.0793, 0.0667, 0.0612, 0.0727, 0.1051, 0.0739], device='cuda:3'), in_proj_covar=tensor([0.0560, 0.0691, 0.0566, 0.0500, 0.0440, 0.0451, 0.0579, 0.0527], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:00:16,390 INFO [train.py:904] (3/8) Epoch 14, batch 10000, loss[loss=0.1648, simple_loss=0.2638, pruned_loss=0.03292, over 16756.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2687, pruned_loss=0.03946, over 3096184.48 frames. ], batch size: 124, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:01:07,175 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:25,926 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:34,644 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.061e+02 2.551e+02 3.100e+02 8.241e+02, threshold=5.102e+02, percent-clipped=1.0 2023-04-30 02:02:01,157 INFO [train.py:904] (3/8) Epoch 14, batch 10050, loss[loss=0.1937, simple_loss=0.2896, pruned_loss=0.04885, over 15367.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2682, pruned_loss=0.03913, over 3084484.59 frames. ], batch size: 190, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:03:00,954 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:03:34,641 INFO [train.py:904] (3/8) Epoch 14, batch 10100, loss[loss=0.1524, simple_loss=0.2584, pruned_loss=0.02319, over 16826.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2685, pruned_loss=0.03961, over 3081378.98 frames. ], batch size: 102, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:04:19,943 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:04:44,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0687, 2.0261, 2.2406, 3.5301, 2.0202, 2.2996, 2.1702, 2.1470], device='cuda:3'), covar=tensor([0.1114, 0.3676, 0.2660, 0.0503, 0.4246, 0.2616, 0.3483, 0.3704], device='cuda:3'), in_proj_covar=tensor([0.0359, 0.0396, 0.0333, 0.0307, 0.0407, 0.0450, 0.0362, 0.0458], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:04:49,687 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:05:19,914 INFO [train.py:904] (3/8) Epoch 15, batch 0, loss[loss=0.18, simple_loss=0.2625, pruned_loss=0.04873, over 17280.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2625, pruned_loss=0.04873, over 17280.00 frames. ], batch size: 45, lr: 4.62e-03, grad_scale: 8.0 2023-04-30 02:05:19,914 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 02:05:27,352 INFO [train.py:938] (3/8) Epoch 15, validation: loss=0.1501, simple_loss=0.2536, pruned_loss=0.02333, over 944034.00 frames. 2023-04-30 02:05:27,353 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-30 02:05:53,971 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:06:27,843 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 50, loss[loss=0.1891, simple_loss=0.2612, pruned_loss=0.05846, over 16833.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2799, pruned_loss=0.05716, over 751833.16 frames. ], batch size: 83, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:07:24,349 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-30 02:07:44,550 INFO [optim.py:368] (3/8) 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 02:07:47,914 INFO [train.py:904] (3/8) Epoch 15, batch 100, loss[loss=0.1642, simple_loss=0.2611, pruned_loss=0.03364, over 17136.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2744, pruned_loss=0.05463, over 1317490.67 frames. ], batch size: 49, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:08:56,687 INFO [train.py:904] (3/8) Epoch 15, batch 150, loss[loss=0.1696, simple_loss=0.2653, pruned_loss=0.03696, over 17144.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.271, pruned_loss=0.0523, over 1761675.83 frames. ], batch size: 48, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:09:11,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1729, 4.2265, 2.7976, 4.7475, 3.2795, 4.7275, 2.7514, 3.4430], device='cuda:3'), covar=tensor([0.0245, 0.0332, 0.1355, 0.0254, 0.0738, 0.0388, 0.1399, 0.0631], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0164, 0.0186, 0.0137, 0.0165, 0.0201, 0.0195, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:09:25,347 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:09:37,903 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 02:09:42,883 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:10:04,594 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 200, loss[loss=0.2033, simple_loss=0.2709, pruned_loss=0.06782, over 16916.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2701, pruned_loss=0.05147, over 2115619.38 frames. ], batch size: 109, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:11:09,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6126, 3.6622, 4.1184, 2.0942, 4.2362, 4.2035, 3.2519, 3.2347], device='cuda:3'), covar=tensor([0.0780, 0.0210, 0.0146, 0.1162, 0.0062, 0.0148, 0.0389, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0100, 0.0085, 0.0136, 0.0070, 0.0109, 0.0121, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 02:11:13,368 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4316, 4.8218, 4.3177, 4.6481, 4.4072, 4.3736, 4.3755, 4.8987], device='cuda:3'), covar=tensor([0.2221, 0.1717, 0.2708, 0.1662, 0.1574, 0.1927, 0.2473, 0.1913], device='cuda:3'), in_proj_covar=tensor([0.0584, 0.0725, 0.0593, 0.0523, 0.0459, 0.0468, 0.0607, 0.0548], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:11:17,117 INFO [train.py:904] (3/8) Epoch 15, batch 250, loss[loss=0.1945, simple_loss=0.2636, pruned_loss=0.06267, over 16496.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2683, pruned_loss=0.05091, over 2389840.18 frames. ], batch size: 75, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:11:41,289 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8527, 2.8403, 2.4591, 4.2022, 3.5439, 4.1677, 1.6173, 3.0181], device='cuda:3'), covar=tensor([0.1253, 0.0587, 0.1099, 0.0140, 0.0147, 0.0343, 0.1392, 0.0700], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0164, 0.0185, 0.0162, 0.0193, 0.0209, 0.0190, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:12:22,521 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 2.346e+02 2.797e+02 3.590e+02 6.679e+02, threshold=5.594e+02, percent-clipped=6.0 2023-04-30 02:12:25,435 INFO [train.py:904] (3/8) Epoch 15, batch 300, loss[loss=0.1663, simple_loss=0.2623, pruned_loss=0.03516, over 17163.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2659, pruned_loss=0.04986, over 2605099.16 frames. ], batch size: 46, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:13:01,155 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2878, 5.1829, 5.0606, 4.5823, 4.6678, 5.0623, 5.0849, 4.7258], device='cuda:3'), covar=tensor([0.0558, 0.0521, 0.0323, 0.0322, 0.1160, 0.0497, 0.0304, 0.0777], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0349, 0.0304, 0.0284, 0.0317, 0.0332, 0.0210, 0.0358], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:13:36,236 INFO [train.py:904] (3/8) Epoch 15, batch 350, loss[loss=0.1818, simple_loss=0.2658, pruned_loss=0.04889, over 16870.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2637, pruned_loss=0.04825, over 2761961.38 frames. ], batch size: 42, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:13:53,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0361, 3.1957, 3.3643, 2.1823, 2.8121, 2.3579, 3.5004, 3.4744], device='cuda:3'), covar=tensor([0.0233, 0.0836, 0.0529, 0.1656, 0.0776, 0.0920, 0.0483, 0.0829], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0147, 0.0158, 0.0145, 0.0137, 0.0125, 0.0136, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:14:12,773 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8544, 5.1582, 5.3036, 5.1441, 5.1191, 5.7573, 5.2916, 4.9938], device='cuda:3'), covar=tensor([0.1384, 0.1931, 0.2609, 0.2417, 0.2958, 0.1245, 0.1650, 0.2625], device='cuda:3'), in_proj_covar=tensor([0.0370, 0.0523, 0.0576, 0.0445, 0.0601, 0.0601, 0.0456, 0.0595], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:14:19,143 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9996, 4.2454, 3.1481, 2.3447, 2.8106, 2.4783, 4.5768, 3.7014], device='cuda:3'), covar=tensor([0.2290, 0.0577, 0.1490, 0.2342, 0.2591, 0.1975, 0.0364, 0.1198], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0257, 0.0288, 0.0286, 0.0273, 0.0231, 0.0269, 0.0304], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:14:36,224 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7845, 3.1513, 2.7421, 4.9321, 4.1093, 4.5383, 1.6359, 3.2202], device='cuda:3'), covar=tensor([0.1309, 0.0629, 0.1117, 0.0194, 0.0262, 0.0356, 0.1520, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0164, 0.0185, 0.0163, 0.0194, 0.0210, 0.0190, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:14:42,899 INFO [optim.py:368] (3/8) 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 02:14:45,203 INFO [train.py:904] (3/8) Epoch 15, batch 400, loss[loss=0.1635, simple_loss=0.2535, pruned_loss=0.03678, over 17176.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2629, pruned_loss=0.04877, over 2884229.89 frames. ], batch size: 46, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:15:54,211 INFO [train.py:904] (3/8) Epoch 15, batch 450, loss[loss=0.1682, simple_loss=0.2396, pruned_loss=0.0484, over 16736.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.261, pruned_loss=0.04803, over 2978566.73 frames. ], batch size: 124, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:16:02,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 02:16:17,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1154, 5.0807, 5.6272, 5.5862, 5.6116, 5.2798, 5.1500, 4.9497], device='cuda:3'), covar=tensor([0.0333, 0.0438, 0.0313, 0.0383, 0.0442, 0.0305, 0.0952, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0381, 0.0380, 0.0359, 0.0424, 0.0404, 0.0490, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:16:22,467 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7873, 4.2785, 3.1391, 2.2044, 2.7856, 2.5136, 4.6138, 3.6648], device='cuda:3'), covar=tensor([0.2716, 0.0581, 0.1588, 0.2641, 0.2558, 0.1863, 0.0340, 0.1240], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0260, 0.0291, 0.0289, 0.0277, 0.0234, 0.0272, 0.0308], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:16:23,391 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:24,682 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:41,567 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:03,317 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.083e+02 2.627e+02 3.131e+02 6.454e+02, threshold=5.253e+02, percent-clipped=1.0 2023-04-30 02:17:05,223 INFO [train.py:904] (3/8) Epoch 15, batch 500, loss[loss=0.1787, simple_loss=0.2583, pruned_loss=0.04956, over 16837.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2592, pruned_loss=0.04633, over 3051192.59 frames. ], batch size: 96, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:17:19,234 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5393, 2.2413, 2.2508, 4.4595, 2.2238, 2.7395, 2.3049, 2.4418], device='cuda:3'), covar=tensor([0.1085, 0.3667, 0.2797, 0.0410, 0.3967, 0.2397, 0.3322, 0.3438], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0410, 0.0344, 0.0320, 0.0420, 0.0470, 0.0376, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:17:28,778 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:49,353 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:52,911 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7069, 2.5614, 2.2375, 2.5481, 2.9936, 2.7983, 3.5086, 3.2902], device='cuda:3'), covar=tensor([0.0103, 0.0374, 0.0457, 0.0393, 0.0259, 0.0356, 0.0187, 0.0217], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0221, 0.0212, 0.0213, 0.0219, 0.0220, 0.0224, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:18:13,784 INFO [train.py:904] (3/8) Epoch 15, batch 550, loss[loss=0.1946, simple_loss=0.2623, pruned_loss=0.06345, over 12402.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2574, pruned_loss=0.04551, over 3113538.91 frames. ], batch size: 246, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:22,092 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 600, loss[loss=0.1922, simple_loss=0.2731, pruned_loss=0.05567, over 16517.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2575, pruned_loss=0.04586, over 3159905.64 frames. ], batch size: 68, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:53,051 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 650, loss[loss=0.1823, simple_loss=0.26, pruned_loss=0.05234, over 16483.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.257, pruned_loss=0.0469, over 3192721.68 frames. ], batch size: 146, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:20:43,421 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 02:20:48,830 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7981, 1.8635, 2.2931, 2.6284, 2.6821, 2.6491, 1.8964, 2.9170], device='cuda:3'), covar=tensor([0.0151, 0.0391, 0.0294, 0.0218, 0.0241, 0.0226, 0.0425, 0.0117], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0179, 0.0163, 0.0168, 0.0177, 0.0133, 0.0180, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:21:14,213 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9509, 4.2252, 4.0610, 4.0986, 3.7949, 3.7945, 3.8837, 4.2136], device='cuda:3'), covar=tensor([0.1168, 0.0935, 0.0987, 0.0739, 0.0754, 0.1609, 0.0894, 0.1040], device='cuda:3'), in_proj_covar=tensor([0.0605, 0.0750, 0.0612, 0.0539, 0.0473, 0.0482, 0.0629, 0.0569], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:21:17,820 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:21:18,118 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 02:21:19,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2919, 3.2937, 3.4846, 2.1767, 3.0294, 2.4063, 3.7908, 3.5680], device='cuda:3'), covar=tensor([0.0206, 0.0924, 0.0631, 0.1843, 0.0800, 0.1038, 0.0445, 0.0938], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0150, 0.0161, 0.0147, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:21:27,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5829, 2.5282, 2.1560, 2.4827, 2.9462, 2.7447, 3.2894, 3.1866], device='cuda:3'), covar=tensor([0.0108, 0.0410, 0.0470, 0.0414, 0.0273, 0.0347, 0.0263, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0220, 0.0211, 0.0213, 0.0219, 0.0220, 0.0224, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:21:40,878 INFO [optim.py:368] (3/8) 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 02:21:42,124 INFO [train.py:904] (3/8) Epoch 15, batch 700, loss[loss=0.2037, simple_loss=0.2765, pruned_loss=0.06549, over 16713.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2564, pruned_loss=0.04611, over 3220856.35 frames. ], batch size: 134, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:22:31,147 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2172, 2.1689, 1.6945, 1.9477, 2.4224, 2.2477, 2.3750, 2.5732], device='cuda:3'), covar=tensor([0.0229, 0.0343, 0.0468, 0.0407, 0.0226, 0.0290, 0.0219, 0.0239], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0221, 0.0212, 0.0213, 0.0220, 0.0221, 0.0225, 0.0211], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:22:50,285 INFO [train.py:904] (3/8) Epoch 15, batch 750, loss[loss=0.148, simple_loss=0.2341, pruned_loss=0.03096, over 17264.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2567, pruned_loss=0.04583, over 3248419.17 frames. ], batch size: 43, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:23:57,714 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 800, loss[loss=0.1882, simple_loss=0.2774, pruned_loss=0.04948, over 17072.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.257, pruned_loss=0.04608, over 3260510.84 frames. ], batch size: 55, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:24:36,474 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:25:08,641 INFO [train.py:904] (3/8) Epoch 15, batch 850, loss[loss=0.1851, simple_loss=0.2714, pruned_loss=0.04941, over 16771.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2563, pruned_loss=0.0454, over 3268240.02 frames. ], batch size: 57, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:26:15,121 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 900, loss[loss=0.1584, simple_loss=0.2394, pruned_loss=0.03874, over 16870.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2558, pruned_loss=0.04479, over 3284182.59 frames. ], batch size: 90, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:27:16,611 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9303, 4.0105, 4.4249, 2.2123, 4.5052, 4.6526, 3.1510, 3.6614], device='cuda:3'), covar=tensor([0.0702, 0.0196, 0.0161, 0.1081, 0.0072, 0.0159, 0.0415, 0.0332], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0104, 0.0090, 0.0140, 0.0073, 0.0116, 0.0126, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:27:24,767 INFO [train.py:904] (3/8) Epoch 15, batch 950, loss[loss=0.1683, simple_loss=0.2469, pruned_loss=0.04481, over 16784.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2563, pruned_loss=0.04473, over 3290942.11 frames. ], batch size: 83, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:28:02,467 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:28:29,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0964, 4.1474, 4.4619, 4.4642, 4.5042, 4.1987, 4.2506, 4.0848], device='cuda:3'), covar=tensor([0.0384, 0.0751, 0.0469, 0.0472, 0.0503, 0.0517, 0.0790, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0396, 0.0393, 0.0372, 0.0441, 0.0420, 0.0506, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:28:30,229 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 1000, loss[loss=0.1972, simple_loss=0.2608, pruned_loss=0.06681, over 16932.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2553, pruned_loss=0.04494, over 3296129.76 frames. ], batch size: 109, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:29:17,361 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6909, 3.7130, 4.1658, 2.3605, 3.4141, 2.7740, 4.0059, 4.0454], device='cuda:3'), covar=tensor([0.0247, 0.0804, 0.0451, 0.1811, 0.0746, 0.0921, 0.0555, 0.0987], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0152, 0.0162, 0.0148, 0.0139, 0.0127, 0.0139, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:29:29,034 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-04-30 02:29:41,403 INFO [train.py:904] (3/8) Epoch 15, batch 1050, loss[loss=0.1668, simple_loss=0.2443, pruned_loss=0.04462, over 16496.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2545, pruned_loss=0.04467, over 3305757.71 frames. ], batch size: 146, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:30:46,963 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 1100, loss[loss=0.1688, simple_loss=0.2669, pruned_loss=0.03533, over 17043.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.254, pruned_loss=0.04435, over 3312831.14 frames. ], batch size: 50, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:31:26,419 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:31:40,427 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5961, 5.9689, 5.7247, 5.8203, 5.3859, 5.3727, 5.3837, 6.1324], device='cuda:3'), covar=tensor([0.1269, 0.0937, 0.1069, 0.0777, 0.0888, 0.0659, 0.1143, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0613, 0.0764, 0.0622, 0.0548, 0.0482, 0.0489, 0.0641, 0.0580], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:31:58,384 INFO [train.py:904] (3/8) Epoch 15, batch 1150, loss[loss=0.1527, simple_loss=0.2438, pruned_loss=0.03074, over 17301.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2533, pruned_loss=0.04347, over 3304537.11 frames. ], batch size: 52, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:32:02,111 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7520, 2.7365, 2.5479, 4.7157, 3.6800, 4.2736, 1.6485, 3.0899], device='cuda:3'), covar=tensor([0.1448, 0.0789, 0.1207, 0.0170, 0.0293, 0.0428, 0.1581, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0191, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:32:07,000 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7848, 2.6489, 2.5534, 4.7296, 3.7653, 4.2971, 1.6794, 3.0808], device='cuda:3'), covar=tensor([0.1393, 0.0822, 0.1219, 0.0186, 0.0297, 0.0435, 0.1506, 0.0800], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0191, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:32:34,544 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:33:01,268 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3904, 3.4572, 3.7477, 1.8682, 3.8686, 3.8123, 3.0644, 2.7844], device='cuda:3'), covar=tensor([0.0739, 0.0205, 0.0172, 0.1156, 0.0079, 0.0150, 0.0374, 0.0440], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0105, 0.0090, 0.0141, 0.0073, 0.0116, 0.0126, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:33:07,201 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 1200, loss[loss=0.1703, simple_loss=0.2637, pruned_loss=0.03846, over 17071.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2524, pruned_loss=0.04338, over 3298721.00 frames. ], batch size: 53, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:33:11,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8182, 3.8369, 2.3568, 4.2677, 2.8916, 4.2345, 2.4801, 3.0988], device='cuda:3'), covar=tensor([0.0220, 0.0358, 0.1455, 0.0308, 0.0764, 0.0529, 0.1479, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0171, 0.0192, 0.0147, 0.0170, 0.0213, 0.0200, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:33:41,632 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-30 02:33:43,176 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1565, 2.9819, 3.1610, 1.8254, 3.2803, 3.2453, 2.7173, 2.5806], device='cuda:3'), covar=tensor([0.0757, 0.0222, 0.0230, 0.1120, 0.0101, 0.0240, 0.0475, 0.0446], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0104, 0.0089, 0.0140, 0.0073, 0.0116, 0.0126, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 02:33:44,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6249, 4.4642, 4.6097, 4.8018, 4.9184, 4.3689, 4.7330, 4.9176], device='cuda:3'), covar=tensor([0.1533, 0.1168, 0.1351, 0.0746, 0.0551, 0.1490, 0.1959, 0.0801], device='cuda:3'), in_proj_covar=tensor([0.0594, 0.0730, 0.0872, 0.0746, 0.0559, 0.0579, 0.0591, 0.0695], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:34:16,302 INFO [train.py:904] (3/8) Epoch 15, batch 1250, loss[loss=0.183, simple_loss=0.2542, pruned_loss=0.05592, over 16711.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2532, pruned_loss=0.04363, over 3299458.74 frames. ], batch size: 134, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:34:56,624 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:35:25,424 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:35:26,162 INFO [optim.py:368] (3/8) 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 02:35:27,820 INFO [train.py:904] (3/8) Epoch 15, batch 1300, loss[loss=0.157, simple_loss=0.2446, pruned_loss=0.03464, over 16802.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2536, pruned_loss=0.04397, over 3312199.02 frames. ], batch size: 42, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:03,859 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:36:05,483 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 02:36:36,236 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 02:36:37,208 INFO [train.py:904] (3/8) Epoch 15, batch 1350, loss[loss=0.1846, simple_loss=0.2619, pruned_loss=0.05363, over 16832.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2545, pruned_loss=0.04396, over 3312435.91 frames. ], batch size: 90, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:39,914 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4225, 4.2257, 4.2695, 4.6289, 4.7289, 4.3342, 4.7130, 4.7860], device='cuda:3'), covar=tensor([0.1932, 0.1678, 0.2032, 0.1037, 0.0976, 0.1303, 0.1931, 0.1010], device='cuda:3'), in_proj_covar=tensor([0.0604, 0.0742, 0.0889, 0.0759, 0.0568, 0.0589, 0.0599, 0.0707], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:36:49,876 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 02:37:06,545 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:37:45,742 INFO [optim.py:368] (3/8) 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 02:37:47,550 INFO [train.py:904] (3/8) Epoch 15, batch 1400, loss[loss=0.1513, simple_loss=0.2395, pruned_loss=0.03155, over 17177.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2539, pruned_loss=0.04385, over 3309176.07 frames. ], batch size: 44, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:31,305 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:38:46,972 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 02:38:55,994 INFO [train.py:904] (3/8) Epoch 15, batch 1450, loss[loss=0.1555, simple_loss=0.2507, pruned_loss=0.03018, over 17202.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2538, pruned_loss=0.04384, over 3307619.32 frames. ], batch size: 46, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:39:16,051 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-30 02:40:00,364 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8596, 1.9785, 2.3405, 2.7000, 2.7522, 2.7287, 1.9641, 3.0078], device='cuda:3'), covar=tensor([0.0142, 0.0367, 0.0283, 0.0226, 0.0207, 0.0202, 0.0379, 0.0109], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0183, 0.0168, 0.0173, 0.0181, 0.0136, 0.0183, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:40:05,552 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 1500, loss[loss=0.1667, simple_loss=0.2424, pruned_loss=0.04551, over 16582.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.253, pruned_loss=0.044, over 3308477.07 frames. ], batch size: 75, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:40:10,086 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8279, 3.7326, 3.8482, 3.9923, 4.0649, 3.6413, 3.8779, 4.0929], device='cuda:3'), covar=tensor([0.1277, 0.0971, 0.1203, 0.0627, 0.0523, 0.1696, 0.1625, 0.0628], device='cuda:3'), in_proj_covar=tensor([0.0605, 0.0746, 0.0894, 0.0765, 0.0570, 0.0593, 0.0600, 0.0710], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:41:14,536 INFO [train.py:904] (3/8) Epoch 15, batch 1550, loss[loss=0.1733, simple_loss=0.2731, pruned_loss=0.03678, over 17044.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2543, pruned_loss=0.04454, over 3319327.49 frames. ], batch size: 50, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:53,849 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8594, 4.2838, 4.3233, 3.2078, 3.5575, 4.3213, 3.8512, 2.3539], device='cuda:3'), covar=tensor([0.0415, 0.0064, 0.0042, 0.0301, 0.0123, 0.0065, 0.0078, 0.0445], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0075, 0.0075, 0.0131, 0.0088, 0.0098, 0.0086, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:41:57,694 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1386, 5.6645, 5.8313, 5.5584, 5.6327, 6.1999, 5.8060, 5.5320], device='cuda:3'), covar=tensor([0.0832, 0.2054, 0.2194, 0.2147, 0.2615, 0.0953, 0.1311, 0.2161], device='cuda:3'), in_proj_covar=tensor([0.0387, 0.0554, 0.0608, 0.0470, 0.0634, 0.0637, 0.0482, 0.0624], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:41:59,509 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 02:42:22,874 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.352e+02 2.733e+02 3.255e+02 5.583e+02, threshold=5.465e+02, percent-clipped=1.0 2023-04-30 02:42:24,070 INFO [train.py:904] (3/8) Epoch 15, batch 1600, loss[loss=0.1884, simple_loss=0.2767, pruned_loss=0.05006, over 16516.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2561, pruned_loss=0.04538, over 3325272.16 frames. ], batch size: 68, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:13,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6873, 2.5948, 2.4593, 4.1655, 3.4079, 4.0732, 1.6255, 2.8542], device='cuda:3'), covar=tensor([0.1346, 0.0696, 0.1085, 0.0166, 0.0152, 0.0378, 0.1400, 0.0807], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0165, 0.0185, 0.0167, 0.0198, 0.0214, 0.0189, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:43:32,479 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-30 02:43:35,419 INFO [train.py:904] (3/8) Epoch 15, batch 1650, loss[loss=0.2048, simple_loss=0.2751, pruned_loss=0.06727, over 16759.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2575, pruned_loss=0.04621, over 3319272.07 frames. ], batch size: 124, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:40,913 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:43:48,946 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-04-30 02:44:46,123 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 1700, loss[loss=0.2054, simple_loss=0.2913, pruned_loss=0.05972, over 16717.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2597, pruned_loss=0.04693, over 3319055.34 frames. ], batch size: 62, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:44:58,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6723, 4.8113, 5.0101, 4.8479, 4.8614, 5.4799, 5.0691, 4.6957], device='cuda:3'), covar=tensor([0.1473, 0.2003, 0.2363, 0.2265, 0.2859, 0.1083, 0.1532, 0.2705], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0552, 0.0608, 0.0467, 0.0631, 0.0633, 0.0481, 0.0622], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:45:22,433 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:45:54,338 INFO [train.py:904] (3/8) Epoch 15, batch 1750, loss[loss=0.1772, simple_loss=0.2572, pruned_loss=0.0486, over 16760.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2605, pruned_loss=0.04685, over 3319722.45 frames. ], batch size: 102, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:46:04,351 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 02:47:05,601 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.250e+02 2.638e+02 3.028e+02 6.204e+02, threshold=5.275e+02, percent-clipped=1.0 2023-04-30 02:47:05,617 INFO [train.py:904] (3/8) Epoch 15, batch 1800, loss[loss=0.1726, simple_loss=0.2498, pruned_loss=0.04764, over 16783.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2613, pruned_loss=0.0465, over 3325835.18 frames. ], batch size: 39, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:47:16,670 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 02:47:17,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0830, 5.5753, 5.7924, 5.5049, 5.5466, 6.1484, 5.6502, 5.3306], device='cuda:3'), covar=tensor([0.0867, 0.1853, 0.2224, 0.2020, 0.2752, 0.1015, 0.1544, 0.2480], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0551, 0.0604, 0.0465, 0.0628, 0.0631, 0.0478, 0.0620], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:47:59,770 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 02:48:15,536 INFO [train.py:904] (3/8) Epoch 15, batch 1850, loss[loss=0.1726, simple_loss=0.2732, pruned_loss=0.036, over 17047.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2627, pruned_loss=0.0471, over 3327061.25 frames. ], batch size: 50, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:48:18,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3439, 3.3208, 3.4151, 3.5154, 3.5727, 3.2299, 3.4822, 3.6264], device='cuda:3'), covar=tensor([0.1201, 0.0971, 0.1175, 0.0611, 0.0626, 0.2604, 0.1220, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0602, 0.0741, 0.0890, 0.0763, 0.0568, 0.0591, 0.0598, 0.0707], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:48:36,073 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0252, 4.0416, 4.3271, 2.2060, 4.5392, 4.5839, 3.1476, 3.6599], device='cuda:3'), covar=tensor([0.0659, 0.0185, 0.0217, 0.1068, 0.0072, 0.0135, 0.0431, 0.0331], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0105, 0.0092, 0.0141, 0.0074, 0.0118, 0.0126, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:49:30,762 INFO [train.py:904] (3/8) Epoch 15, batch 1900, loss[loss=0.1803, simple_loss=0.2631, pruned_loss=0.04874, over 16478.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2627, pruned_loss=0.04664, over 3315121.85 frames. ], batch size: 68, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:49:31,843 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.211e+02 2.636e+02 2.995e+02 6.158e+02, threshold=5.272e+02, percent-clipped=2.0 2023-04-30 02:49:40,841 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2436, 2.1921, 1.7569, 1.9547, 2.4688, 2.2459, 2.3818, 2.5937], device='cuda:3'), covar=tensor([0.0203, 0.0312, 0.0436, 0.0372, 0.0190, 0.0276, 0.0200, 0.0222], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0223, 0.0215, 0.0215, 0.0224, 0.0223, 0.0230, 0.0216], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:50:07,582 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0269, 3.2670, 2.7609, 5.1917, 4.3914, 4.6585, 1.8065, 3.3921], device='cuda:3'), covar=tensor([0.1236, 0.0644, 0.1156, 0.0165, 0.0241, 0.0353, 0.1425, 0.0684], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0164, 0.0184, 0.0166, 0.0198, 0.0212, 0.0188, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:50:39,887 INFO [train.py:904] (3/8) Epoch 15, batch 1950, loss[loss=0.1729, simple_loss=0.27, pruned_loss=0.03785, over 17265.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2627, pruned_loss=0.04598, over 3306366.20 frames. ], batch size: 52, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:50:46,782 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 02:51:04,045 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0968, 5.0763, 5.5519, 5.5369, 5.5868, 5.2196, 5.1634, 4.9596], device='cuda:3'), covar=tensor([0.0295, 0.0494, 0.0336, 0.0459, 0.0494, 0.0380, 0.0937, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0375, 0.0398, 0.0395, 0.0375, 0.0439, 0.0419, 0.0508, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 02:51:25,928 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8142, 3.0172, 2.4723, 4.7649, 3.8504, 4.3872, 1.5482, 3.1206], device='cuda:3'), covar=tensor([0.1350, 0.0690, 0.1290, 0.0181, 0.0329, 0.0410, 0.1583, 0.0796], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0164, 0.0185, 0.0166, 0.0198, 0.0212, 0.0188, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 02:51:49,540 INFO [train.py:904] (3/8) Epoch 15, batch 2000, loss[loss=0.1641, simple_loss=0.2563, pruned_loss=0.03592, over 17202.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2615, pruned_loss=0.04548, over 3314632.35 frames. ], batch size: 46, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:51:51,361 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:23,161 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5840, 3.4623, 2.7065, 2.1578, 2.4024, 2.1743, 3.6139, 3.2356], device='cuda:3'), covar=tensor([0.2537, 0.0794, 0.1608, 0.2662, 0.2292, 0.2012, 0.0526, 0.1390], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0264, 0.0294, 0.0293, 0.0286, 0.0239, 0.0278, 0.0318], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:52:27,254 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:28,598 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2888, 5.3076, 5.1683, 4.6351, 5.1267, 2.3412, 4.9088, 5.1453], device='cuda:3'), covar=tensor([0.0083, 0.0069, 0.0145, 0.0361, 0.0094, 0.2142, 0.0130, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0138, 0.0186, 0.0170, 0.0157, 0.0198, 0.0173, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 02:52:30,125 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 02:52:58,028 INFO [train.py:904] (3/8) Epoch 15, batch 2050, loss[loss=0.2293, simple_loss=0.3051, pruned_loss=0.07673, over 11962.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2613, pruned_loss=0.04633, over 3308008.86 frames. ], batch size: 245, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:53:32,904 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:54:07,933 INFO [train.py:904] (3/8) Epoch 15, batch 2100, loss[loss=0.2121, simple_loss=0.2842, pruned_loss=0.06996, over 16190.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2634, pruned_loss=0.04772, over 3288022.85 frames. ], batch size: 165, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:54:08,982 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.513e+02 2.931e+02 3.819e+02 1.829e+03, threshold=5.862e+02, percent-clipped=10.0 2023-04-30 02:55:17,937 INFO [train.py:904] (3/8) Epoch 15, batch 2150, loss[loss=0.2008, simple_loss=0.2778, pruned_loss=0.06192, over 16691.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.264, pruned_loss=0.04775, over 3304934.89 frames. ], batch size: 134, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:55:19,620 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1256, 4.4527, 3.2387, 2.4114, 2.8907, 2.8024, 4.8449, 3.8084], device='cuda:3'), covar=tensor([0.2491, 0.0574, 0.1625, 0.2590, 0.2819, 0.1855, 0.0347, 0.1314], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0264, 0.0294, 0.0292, 0.0286, 0.0239, 0.0277, 0.0317], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:55:31,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-30 02:56:20,043 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1215, 5.2335, 4.9728, 4.6227, 4.1322, 5.2204, 5.1654, 4.6944], device='cuda:3'), covar=tensor([0.0855, 0.0550, 0.0496, 0.0440, 0.2054, 0.0430, 0.0359, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0380, 0.0330, 0.0311, 0.0343, 0.0359, 0.0225, 0.0390], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 02:56:25,351 INFO [train.py:904] (3/8) Epoch 15, batch 2200, loss[loss=0.1866, simple_loss=0.2778, pruned_loss=0.04766, over 16674.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2644, pruned_loss=0.0479, over 3305845.42 frames. ], batch size: 57, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:56:27,073 INFO [optim.py:368] (3/8) 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 02:56:52,863 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 02:57:36,226 INFO [train.py:904] (3/8) Epoch 15, batch 2250, loss[loss=0.1598, simple_loss=0.2448, pruned_loss=0.03746, over 15965.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.264, pruned_loss=0.04767, over 3314549.08 frames. ], batch size: 35, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:37,671 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 02:58:46,689 INFO [train.py:904] (3/8) Epoch 15, batch 2300, loss[loss=0.1558, simple_loss=0.2374, pruned_loss=0.0371, over 16983.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2639, pruned_loss=0.0476, over 3319397.52 frames. ], batch size: 41, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:47,877 INFO [optim.py:368] (3/8) 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 02:58:58,068 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:59:53,214 INFO [train.py:904] (3/8) Epoch 15, batch 2350, loss[loss=0.1732, simple_loss=0.2536, pruned_loss=0.04636, over 16783.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.265, pruned_loss=0.04803, over 3323444.96 frames. ], batch size: 102, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 03:00:20,655 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:01:02,764 INFO [train.py:904] (3/8) Epoch 15, batch 2400, loss[loss=0.1822, simple_loss=0.2673, pruned_loss=0.04857, over 15388.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2653, pruned_loss=0.04828, over 3322365.03 frames. ], batch size: 190, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:01:04,724 INFO [optim.py:368] (3/8) 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 03:01:22,920 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 03:02:09,964 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 03:02:10,317 INFO [train.py:904] (3/8) Epoch 15, batch 2450, loss[loss=0.1826, simple_loss=0.2564, pruned_loss=0.05439, over 16906.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2657, pruned_loss=0.04818, over 3320734.53 frames. ], batch size: 116, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:02:44,136 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2791, 2.1161, 2.2855, 4.0460, 2.2116, 2.5735, 2.2146, 2.3036], device='cuda:3'), covar=tensor([0.1175, 0.3573, 0.2570, 0.0480, 0.3508, 0.2307, 0.3538, 0.2907], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0417, 0.0349, 0.0327, 0.0423, 0.0480, 0.0382, 0.0488], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:03:17,679 INFO [train.py:904] (3/8) Epoch 15, batch 2500, loss[loss=0.1461, simple_loss=0.2355, pruned_loss=0.02841, over 16954.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2651, pruned_loss=0.04725, over 3324700.90 frames. ], batch size: 41, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:03:18,673 INFO [optim.py:368] (3/8) 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 03:03:19,137 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3033, 2.1145, 2.7314, 3.2218, 2.9845, 3.5794, 2.1049, 3.5794], device='cuda:3'), covar=tensor([0.0146, 0.0419, 0.0247, 0.0195, 0.0201, 0.0163, 0.0486, 0.0129], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0183, 0.0169, 0.0172, 0.0181, 0.0138, 0.0183, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:04:26,856 INFO [train.py:904] (3/8) Epoch 15, batch 2550, loss[loss=0.1736, simple_loss=0.2585, pruned_loss=0.04432, over 16822.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2646, pruned_loss=0.04704, over 3325662.77 frames. ], batch size: 39, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:34,894 INFO [train.py:904] (3/8) Epoch 15, batch 2600, loss[loss=0.1616, simple_loss=0.2621, pruned_loss=0.03057, over 17045.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2656, pruned_loss=0.0468, over 3311765.68 frames. ], batch size: 53, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:36,052 INFO [optim.py:368] (3/8) 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 03:05:36,416 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9193, 4.6375, 4.9371, 5.1359, 5.3302, 4.6469, 5.3215, 5.2958], device='cuda:3'), covar=tensor([0.1757, 0.1465, 0.1790, 0.0761, 0.0571, 0.0922, 0.0547, 0.0647], device='cuda:3'), in_proj_covar=tensor([0.0611, 0.0755, 0.0903, 0.0775, 0.0581, 0.0604, 0.0610, 0.0716], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:06:43,554 INFO [train.py:904] (3/8) Epoch 15, batch 2650, loss[loss=0.1815, simple_loss=0.2723, pruned_loss=0.04537, over 17134.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2659, pruned_loss=0.04624, over 3311424.99 frames. ], batch size: 49, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:07:05,959 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:07:50,252 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6705, 3.5447, 3.9106, 1.9981, 3.9496, 3.9605, 3.0648, 2.8962], device='cuda:3'), covar=tensor([0.0716, 0.0198, 0.0132, 0.1144, 0.0086, 0.0146, 0.0397, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0105, 0.0091, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 03:07:53,567 INFO [train.py:904] (3/8) Epoch 15, batch 2700, loss[loss=0.1698, simple_loss=0.2584, pruned_loss=0.04059, over 15784.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2662, pruned_loss=0.04544, over 3317744.51 frames. ], batch size: 35, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:07:54,729 INFO [optim.py:368] (3/8) 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 03:08:57,042 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6333, 2.5864, 2.5829, 4.7538, 2.4378, 2.9231, 2.6139, 2.7412], device='cuda:3'), covar=tensor([0.1112, 0.3057, 0.2449, 0.0381, 0.3642, 0.2287, 0.2899, 0.3113], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0416, 0.0347, 0.0327, 0.0422, 0.0480, 0.0380, 0.0488], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:09:02,444 INFO [train.py:904] (3/8) Epoch 15, batch 2750, loss[loss=0.178, simple_loss=0.2705, pruned_loss=0.04274, over 17259.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2666, pruned_loss=0.04515, over 3323515.16 frames. ], batch size: 52, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:09:04,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7035, 2.5706, 2.3185, 4.0355, 3.2356, 3.9906, 1.5487, 2.7168], device='cuda:3'), covar=tensor([0.1552, 0.0764, 0.1296, 0.0196, 0.0165, 0.0410, 0.1671, 0.0845], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0170, 0.0201, 0.0214, 0.0189, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 03:10:11,018 INFO [train.py:904] (3/8) Epoch 15, batch 2800, loss[loss=0.1835, simple_loss=0.2648, pruned_loss=0.05117, over 16885.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2663, pruned_loss=0.04573, over 3310832.00 frames. ], batch size: 96, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:10:12,143 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.181e+02 2.488e+02 3.014e+02 5.995e+02, threshold=4.976e+02, percent-clipped=2.0 2023-04-30 03:10:26,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6963, 2.6836, 2.2478, 2.3392, 2.9670, 2.7227, 3.4240, 3.2286], device='cuda:3'), covar=tensor([0.0113, 0.0369, 0.0427, 0.0429, 0.0258, 0.0356, 0.0214, 0.0226], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0224, 0.0213, 0.0215, 0.0225, 0.0223, 0.0230, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:11:21,054 INFO [train.py:904] (3/8) Epoch 15, batch 2850, loss[loss=0.1621, simple_loss=0.2532, pruned_loss=0.03548, over 15887.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2658, pruned_loss=0.04627, over 3309668.36 frames. ], batch size: 35, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:22,178 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 2900, loss[loss=0.1489, simple_loss=0.2354, pruned_loss=0.03123, over 17242.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2649, pruned_loss=0.0467, over 3309009.31 frames. ], batch size: 44, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:33,021 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.445e+02 2.844e+02 3.300e+02 6.709e+02, threshold=5.687e+02, percent-clipped=6.0 2023-04-30 03:13:06,957 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:13:38,116 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 03:13:40,919 INFO [train.py:904] (3/8) Epoch 15, batch 2950, loss[loss=0.1845, simple_loss=0.2581, pruned_loss=0.0555, over 16914.00 frames. ], tot_loss[loss=0.18, simple_loss=0.265, pruned_loss=0.04748, over 3297141.07 frames. ], batch size: 116, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:13:47,680 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 03:14:01,031 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:02,156 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:31,800 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 3000, loss[loss=0.1523, simple_loss=0.2503, pruned_loss=0.02712, over 17111.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2655, pruned_loss=0.04853, over 3302353.61 frames. ], batch size: 48, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:14:49,684 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 03:14:58,804 INFO [train.py:938] (3/8) Epoch 15, validation: loss=0.138, simple_loss=0.2438, pruned_loss=0.01616, over 944034.00 frames. 2023-04-30 03:14:58,805 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-30 03:15:00,799 INFO [optim.py:368] (3/8) 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 03:15:06,618 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4179, 5.8436, 5.5741, 5.6494, 5.2989, 5.1688, 5.2183, 5.9765], device='cuda:3'), covar=tensor([0.1296, 0.1029, 0.1017, 0.0765, 0.0848, 0.0688, 0.1170, 0.0854], device='cuda:3'), in_proj_covar=tensor([0.0632, 0.0785, 0.0643, 0.0562, 0.0498, 0.0500, 0.0651, 0.0596], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:15:17,418 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:34,481 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:16:07,543 INFO [train.py:904] (3/8) Epoch 15, batch 3050, loss[loss=0.2014, simple_loss=0.2776, pruned_loss=0.06258, over 16326.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2659, pruned_loss=0.04889, over 3300800.10 frames. ], batch size: 165, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:16:16,749 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0269, 5.0145, 5.4749, 5.4774, 5.4874, 5.1520, 5.0659, 4.7954], device='cuda:3'), covar=tensor([0.0282, 0.0462, 0.0342, 0.0359, 0.0428, 0.0329, 0.0920, 0.0421], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0404, 0.0399, 0.0382, 0.0448, 0.0423, 0.0518, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 03:17:18,167 INFO [train.py:904] (3/8) Epoch 15, batch 3100, loss[loss=0.1472, simple_loss=0.235, pruned_loss=0.02968, over 16959.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.265, pruned_loss=0.04867, over 3316387.73 frames. ], batch size: 41, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:17:19,336 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.445e+02 2.806e+02 3.389e+02 5.168e+02, threshold=5.611e+02, percent-clipped=0.0 2023-04-30 03:18:28,438 INFO [train.py:904] (3/8) Epoch 15, batch 3150, loss[loss=0.1723, simple_loss=0.2677, pruned_loss=0.03847, over 17088.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2641, pruned_loss=0.04758, over 3316111.98 frames. ], batch size: 47, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:18:30,128 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8069, 3.9600, 2.3738, 4.4759, 3.0615, 4.4686, 2.5152, 3.1438], device='cuda:3'), covar=tensor([0.0278, 0.0345, 0.1450, 0.0248, 0.0726, 0.0435, 0.1377, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0151, 0.0170, 0.0216, 0.0199, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 03:18:38,990 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5710, 6.0151, 5.4020, 5.9116, 5.4097, 5.1413, 5.4946, 6.0449], device='cuda:3'), covar=tensor([0.2090, 0.1346, 0.2795, 0.1117, 0.1502, 0.1329, 0.2095, 0.1735], device='cuda:3'), in_proj_covar=tensor([0.0628, 0.0779, 0.0638, 0.0558, 0.0494, 0.0496, 0.0648, 0.0590], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:19:37,116 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 03:19:37,246 INFO [train.py:904] (3/8) Epoch 15, batch 3200, loss[loss=0.1722, simple_loss=0.266, pruned_loss=0.03923, over 16645.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2627, pruned_loss=0.04735, over 3320956.12 frames. ], batch size: 62, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:19:38,468 INFO [optim.py:368] (3/8) 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 03:19:43,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2533, 3.1437, 3.3328, 1.7741, 3.4798, 3.4392, 2.7978, 2.6297], device='cuda:3'), covar=tensor([0.0744, 0.0211, 0.0229, 0.1119, 0.0096, 0.0195, 0.0419, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0106, 0.0092, 0.0140, 0.0074, 0.0119, 0.0125, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 03:20:46,516 INFO [train.py:904] (3/8) Epoch 15, batch 3250, loss[loss=0.1967, simple_loss=0.272, pruned_loss=0.06071, over 16506.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2629, pruned_loss=0.0477, over 3315529.94 frames. ], batch size: 146, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:20:46,744 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:21:13,354 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:30,554 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:36,340 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 03:21:57,391 INFO [train.py:904] (3/8) Epoch 15, batch 3300, loss[loss=0.1845, simple_loss=0.2701, pruned_loss=0.04941, over 17166.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2641, pruned_loss=0.04752, over 3308169.87 frames. ], batch size: 46, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:21:58,627 INFO [optim.py:368] (3/8) 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 03:22:22,236 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1439, 3.2489, 3.4273, 2.3002, 3.2052, 3.5717, 3.3904, 1.9276], device='cuda:3'), covar=tensor([0.0469, 0.0164, 0.0062, 0.0356, 0.0097, 0.0094, 0.0084, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0076, 0.0075, 0.0131, 0.0089, 0.0100, 0.0086, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:22:25,241 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:38,862 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:46,199 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 03:23:06,149 INFO [train.py:904] (3/8) Epoch 15, batch 3350, loss[loss=0.1783, simple_loss=0.2687, pruned_loss=0.04397, over 17110.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2642, pruned_loss=0.04708, over 3321814.43 frames. ], batch size: 47, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:17,527 INFO [train.py:904] (3/8) Epoch 15, batch 3400, loss[loss=0.1504, simple_loss=0.233, pruned_loss=0.03389, over 15944.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2636, pruned_loss=0.04641, over 3327308.12 frames. ], batch size: 35, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:18,607 INFO [optim.py:368] (3/8) 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 03:25:28,517 INFO [train.py:904] (3/8) Epoch 15, batch 3450, loss[loss=0.1959, simple_loss=0.2734, pruned_loss=0.05918, over 11570.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2616, pruned_loss=0.04541, over 3329709.08 frames. ], batch size: 248, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:25:55,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1155, 3.2115, 3.0283, 1.7699, 2.5761, 1.9970, 3.5207, 3.5887], device='cuda:3'), covar=tensor([0.0243, 0.0820, 0.0695, 0.2216, 0.1029, 0.1142, 0.0587, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0147, 0.0139, 0.0126, 0.0140, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 03:26:27,204 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 3500, loss[loss=0.1878, simple_loss=0.2638, pruned_loss=0.05591, over 16693.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2602, pruned_loss=0.04499, over 3334306.11 frames. ], batch size: 134, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:39,231 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.386e+02 2.268e+02 2.638e+02 3.199e+02 5.613e+02, threshold=5.276e+02, percent-clipped=1.0 2023-04-30 03:27:36,005 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 3550, loss[loss=0.1754, simple_loss=0.2822, pruned_loss=0.03428, over 17046.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.26, pruned_loss=0.04482, over 3330947.12 frames. ], batch size: 50, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:27:48,322 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:27:52,731 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:28:26,833 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6513, 3.7344, 1.9788, 3.8710, 2.8448, 3.8516, 1.9129, 2.8792], device='cuda:3'), covar=tensor([0.0194, 0.0269, 0.1495, 0.0242, 0.0639, 0.0514, 0.1550, 0.0578], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0152, 0.0170, 0.0216, 0.0200, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 03:28:31,821 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:57,584 INFO [train.py:904] (3/8) Epoch 15, batch 3600, loss[loss=0.1615, simple_loss=0.2459, pruned_loss=0.03857, over 16754.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2594, pruned_loss=0.04474, over 3333015.25 frames. ], batch size: 39, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:28:58,725 INFO [optim.py:368] (3/8) 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 03:29:00,392 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:26,975 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:33,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2629, 5.2256, 5.1030, 4.6119, 4.6925, 5.1580, 5.1542, 4.7990], device='cuda:3'), covar=tensor([0.0607, 0.0460, 0.0293, 0.0328, 0.1113, 0.0427, 0.0320, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0393, 0.0338, 0.0322, 0.0355, 0.0368, 0.0229, 0.0401], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:29:34,302 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:39,860 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:30:10,694 INFO [train.py:904] (3/8) Epoch 15, batch 3650, loss[loss=0.1637, simple_loss=0.2467, pruned_loss=0.04031, over 16979.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2583, pruned_loss=0.04501, over 3336863.16 frames. ], batch size: 41, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:30:38,216 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:31:24,489 INFO [train.py:904] (3/8) Epoch 15, batch 3700, loss[loss=0.2064, simple_loss=0.2793, pruned_loss=0.06679, over 11394.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2578, pruned_loss=0.04707, over 3284067.46 frames. ], batch size: 248, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:31:26,279 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.174e+02 2.701e+02 3.170e+02 5.249e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 03:31:46,964 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.04 vs. limit=5.0 2023-04-30 03:31:52,297 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2433, 3.5326, 3.6211, 2.5545, 3.3869, 3.7934, 3.5656, 2.0055], device='cuda:3'), covar=tensor([0.0484, 0.0105, 0.0057, 0.0347, 0.0088, 0.0085, 0.0079, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0075, 0.0075, 0.0130, 0.0088, 0.0099, 0.0086, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:31:53,344 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:32:37,554 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 03:32:38,909 INFO [train.py:904] (3/8) Epoch 15, batch 3750, loss[loss=0.175, simple_loss=0.2445, pruned_loss=0.05274, over 16912.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2585, pruned_loss=0.04865, over 3277817.05 frames. ], batch size: 96, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:23,457 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:33:51,791 INFO [train.py:904] (3/8) Epoch 15, batch 3800, loss[loss=0.2041, simple_loss=0.2732, pruned_loss=0.06748, over 16899.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2596, pruned_loss=0.04999, over 3266736.18 frames. ], batch size: 109, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:53,679 INFO [optim.py:368] (3/8) 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 03:34:37,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6687, 5.0144, 4.8000, 4.7870, 4.5600, 4.5262, 4.4274, 5.0644], device='cuda:3'), covar=tensor([0.1121, 0.0794, 0.0957, 0.0717, 0.0730, 0.1064, 0.0977, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0618, 0.0770, 0.0629, 0.0553, 0.0487, 0.0488, 0.0640, 0.0586], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:35:00,906 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:35:04,375 INFO [train.py:904] (3/8) Epoch 15, batch 3850, loss[loss=0.1657, simple_loss=0.2399, pruned_loss=0.04581, over 16688.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2596, pruned_loss=0.05052, over 3267633.80 frames. ], batch size: 89, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:13,420 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:36:21,017 INFO [train.py:904] (3/8) Epoch 15, batch 3900, loss[loss=0.1681, simple_loss=0.2424, pruned_loss=0.04688, over 16777.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2588, pruned_loss=0.05088, over 3273959.21 frames. ], batch size: 83, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:22,203 INFO [optim.py:368] (3/8) 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 03:36:57,941 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:37:00,467 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4799, 2.7292, 2.1862, 2.6203, 3.1220, 2.7836, 3.2243, 3.3003], device='cuda:3'), covar=tensor([0.0098, 0.0303, 0.0430, 0.0327, 0.0178, 0.0287, 0.0184, 0.0195], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0220, 0.0212, 0.0214, 0.0221, 0.0220, 0.0229, 0.0215], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:37:11,362 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4086, 3.3019, 3.5580, 1.9329, 3.6468, 3.6798, 3.0318, 2.7864], device='cuda:3'), covar=tensor([0.0713, 0.0203, 0.0156, 0.1045, 0.0096, 0.0162, 0.0353, 0.0422], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0105, 0.0090, 0.0138, 0.0073, 0.0117, 0.0122, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 03:37:32,700 INFO [train.py:904] (3/8) Epoch 15, batch 3950, loss[loss=0.1695, simple_loss=0.2517, pruned_loss=0.04366, over 17102.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2583, pruned_loss=0.0513, over 3268037.40 frames. ], batch size: 48, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:06,890 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:38:13,208 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8399, 3.9386, 4.5200, 2.2287, 4.6862, 4.8080, 3.2893, 3.3968], device='cuda:3'), covar=tensor([0.0765, 0.0239, 0.0120, 0.1135, 0.0044, 0.0057, 0.0346, 0.0438], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0106, 0.0091, 0.0139, 0.0073, 0.0118, 0.0123, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 03:38:46,133 INFO [train.py:904] (3/8) Epoch 15, batch 4000, loss[loss=0.1606, simple_loss=0.2445, pruned_loss=0.03838, over 16719.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2589, pruned_loss=0.05184, over 3264892.88 frames. ], batch size: 83, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:47,409 INFO [optim.py:368] (3/8) 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 03:39:12,675 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6432, 4.7919, 4.9152, 4.8278, 4.7821, 5.3771, 4.9104, 4.6513], device='cuda:3'), covar=tensor([0.1217, 0.1669, 0.1921, 0.1971, 0.2750, 0.0958, 0.1353, 0.2315], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0554, 0.0605, 0.0471, 0.0627, 0.0626, 0.0479, 0.0623], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:39:28,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7413, 1.8751, 2.2810, 2.6579, 2.6704, 2.9465, 1.9999, 2.8421], device='cuda:3'), covar=tensor([0.0149, 0.0403, 0.0260, 0.0243, 0.0224, 0.0153, 0.0396, 0.0094], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0182, 0.0168, 0.0173, 0.0181, 0.0139, 0.0183, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:39:56,191 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:39:59,984 INFO [train.py:904] (3/8) Epoch 15, batch 4050, loss[loss=0.1783, simple_loss=0.2668, pruned_loss=0.04494, over 17181.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2591, pruned_loss=0.05075, over 3281139.99 frames. ], batch size: 46, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:40:36,973 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:41:13,945 INFO [train.py:904] (3/8) Epoch 15, batch 4100, loss[loss=0.2112, simple_loss=0.2952, pruned_loss=0.06356, over 16816.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2603, pruned_loss=0.04998, over 3277642.95 frames. ], batch size: 116, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:41:15,745 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.975e+02 2.401e+02 2.875e+02 5.931e+02, threshold=4.803e+02, percent-clipped=1.0 2023-04-30 03:41:26,654 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:30,100 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:32,940 INFO [train.py:904] (3/8) Epoch 15, batch 4150, loss[loss=0.2266, simple_loss=0.3152, pruned_loss=0.06902, over 16640.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2669, pruned_loss=0.05217, over 3250506.39 frames. ], batch size: 134, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:42:49,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0454, 4.2401, 4.4199, 4.4117, 4.4365, 4.1989, 4.0056, 4.0425], device='cuda:3'), covar=tensor([0.0404, 0.0517, 0.0458, 0.0485, 0.0506, 0.0497, 0.1224, 0.0570], device='cuda:3'), in_proj_covar=tensor([0.0368, 0.0397, 0.0390, 0.0370, 0.0438, 0.0411, 0.0505, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 03:43:45,580 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:46,702 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:50,403 INFO [train.py:904] (3/8) Epoch 15, batch 4200, loss[loss=0.1943, simple_loss=0.2877, pruned_loss=0.05045, over 17105.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2738, pruned_loss=0.0537, over 3220959.87 frames. ], batch size: 49, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:43:53,463 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.342e+02 2.800e+02 3.448e+02 4.997e+02, threshold=5.600e+02, percent-clipped=3.0 2023-04-30 03:44:58,658 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:45:06,395 INFO [train.py:904] (3/8) Epoch 15, batch 4250, loss[loss=0.1851, simple_loss=0.2831, pruned_loss=0.04355, over 16782.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2771, pruned_loss=0.05416, over 3187054.31 frames. ], batch size: 39, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:19,356 INFO [train.py:904] (3/8) Epoch 15, batch 4300, loss[loss=0.1966, simple_loss=0.2882, pruned_loss=0.05251, over 16420.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.278, pruned_loss=0.05309, over 3186516.34 frames. ], batch size: 146, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:23,348 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.412e+02 2.971e+02 3.359e+02 7.082e+02, threshold=5.941e+02, percent-clipped=4.0 2023-04-30 03:46:34,229 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.87 vs. limit=5.0 2023-04-30 03:47:31,145 INFO [train.py:904] (3/8) Epoch 15, batch 4350, loss[loss=0.1847, simple_loss=0.2857, pruned_loss=0.04189, over 16830.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2813, pruned_loss=0.05421, over 3183108.03 frames. ], batch size: 102, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:48:08,888 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:48:45,757 INFO [train.py:904] (3/8) Epoch 15, batch 4400, loss[loss=0.2134, simple_loss=0.3035, pruned_loss=0.06167, over 16515.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2834, pruned_loss=0.05522, over 3180238.70 frames. ], batch size: 75, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:48:50,394 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.574e+02 2.972e+02 3.574e+02 6.742e+02, threshold=5.944e+02, percent-clipped=2.0 2023-04-30 03:48:51,337 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:49:21,073 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:49:58,634 INFO [train.py:904] (3/8) Epoch 15, batch 4450, loss[loss=0.2165, simple_loss=0.3016, pruned_loss=0.06567, over 16802.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2866, pruned_loss=0.05652, over 3196862.71 frames. ], batch size: 124, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:50:10,188 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 03:50:42,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5967, 5.9036, 5.6256, 5.7633, 5.4065, 5.1393, 5.3873, 6.0573], device='cuda:3'), covar=tensor([0.1072, 0.0728, 0.0963, 0.0696, 0.0671, 0.0672, 0.0971, 0.0684], device='cuda:3'), in_proj_covar=tensor([0.0597, 0.0744, 0.0614, 0.0537, 0.0468, 0.0478, 0.0618, 0.0564], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:50:48,967 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8288, 3.6134, 3.8874, 3.5686, 3.7511, 4.2288, 3.9191, 3.4989], device='cuda:3'), covar=tensor([0.2036, 0.2502, 0.2087, 0.2635, 0.2978, 0.2106, 0.1318, 0.2766], device='cuda:3'), in_proj_covar=tensor([0.0377, 0.0534, 0.0580, 0.0453, 0.0604, 0.0607, 0.0459, 0.0605], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:50:57,069 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 4500, loss[loss=0.1952, simple_loss=0.2821, pruned_loss=0.05413, over 17132.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2869, pruned_loss=0.05692, over 3196710.91 frames. ], batch size: 47, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:51:15,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7907, 3.8026, 2.2947, 4.5138, 2.9643, 4.4159, 2.5337, 3.0488], device='cuda:3'), covar=tensor([0.0222, 0.0316, 0.1558, 0.0099, 0.0708, 0.0349, 0.1366, 0.0685], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0191, 0.0145, 0.0167, 0.0211, 0.0197, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 03:51:16,066 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.915e+02 2.182e+02 2.470e+02 4.715e+02, threshold=4.363e+02, percent-clipped=0.0 2023-04-30 03:51:47,939 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:52:25,485 INFO [train.py:904] (3/8) Epoch 15, batch 4550, loss[loss=0.2132, simple_loss=0.3108, pruned_loss=0.05784, over 16794.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2882, pruned_loss=0.05799, over 3188435.11 frames. ], batch size: 89, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:52:25,904 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:53:16,027 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:53:37,390 INFO [train.py:904] (3/8) Epoch 15, batch 4600, loss[loss=0.2323, simple_loss=0.307, pruned_loss=0.07884, over 11371.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2902, pruned_loss=0.05861, over 3194587.74 frames. ], batch size: 248, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:53:41,728 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 1.940e+02 2.192e+02 2.697e+02 4.440e+02, threshold=4.384e+02, percent-clipped=1.0 2023-04-30 03:54:12,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1021, 5.7299, 6.0131, 5.4749, 5.6965, 6.3513, 5.8571, 5.4899], device='cuda:3'), covar=tensor([0.0888, 0.1672, 0.1568, 0.2094, 0.2537, 0.0820, 0.1208, 0.2129], device='cuda:3'), in_proj_covar=tensor([0.0376, 0.0531, 0.0578, 0.0453, 0.0603, 0.0606, 0.0459, 0.0603], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:54:43,989 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-04-30 03:54:49,171 INFO [train.py:904] (3/8) Epoch 15, batch 4650, loss[loss=0.1927, simple_loss=0.2685, pruned_loss=0.05843, over 16469.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2884, pruned_loss=0.05808, over 3219729.43 frames. ], batch size: 35, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:55:25,025 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0104, 2.4855, 2.5715, 1.8657, 2.7642, 2.7947, 2.4091, 2.3518], device='cuda:3'), covar=tensor([0.0718, 0.0219, 0.0201, 0.0926, 0.0102, 0.0209, 0.0441, 0.0468], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0104, 0.0091, 0.0137, 0.0072, 0.0115, 0.0122, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 03:55:48,500 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-04-30 03:56:03,159 INFO [train.py:904] (3/8) Epoch 15, batch 4700, loss[loss=0.1759, simple_loss=0.2526, pruned_loss=0.04955, over 17061.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2856, pruned_loss=0.05689, over 3214763.12 frames. ], batch size: 53, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:56:07,885 INFO [optim.py:368] (3/8) 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 03:56:08,850 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:56:35,070 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 03:56:51,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7510, 3.7427, 2.1471, 4.4734, 2.8486, 4.3878, 2.4104, 2.8706], device='cuda:3'), covar=tensor([0.0256, 0.0393, 0.1830, 0.0094, 0.0853, 0.0372, 0.1577, 0.0875], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0145, 0.0169, 0.0211, 0.0198, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 03:57:07,573 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:18,584 INFO [train.py:904] (3/8) Epoch 15, batch 4750, loss[loss=0.2043, simple_loss=0.279, pruned_loss=0.06481, over 11706.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2808, pruned_loss=0.0546, over 3222915.29 frames. ], batch size: 246, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:57:20,528 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:58:09,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2625, 2.4758, 2.0707, 2.2487, 2.8731, 2.4720, 2.9666, 3.1107], device='cuda:3'), covar=tensor([0.0104, 0.0364, 0.0461, 0.0412, 0.0212, 0.0354, 0.0172, 0.0223], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0209, 0.0216, 0.0216, 0.0222, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:58:15,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7243, 2.1731, 1.7964, 2.0192, 2.5647, 2.2161, 2.4850, 2.7814], device='cuda:3'), covar=tensor([0.0134, 0.0387, 0.0505, 0.0419, 0.0229, 0.0357, 0.0192, 0.0252], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0209, 0.0216, 0.0216, 0.0221, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:58:31,079 INFO [train.py:904] (3/8) Epoch 15, batch 4800, loss[loss=0.1815, simple_loss=0.2758, pruned_loss=0.04354, over 16833.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2778, pruned_loss=0.05276, over 3216417.57 frames. ], batch size: 102, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:58:36,178 INFO [optim.py:368] (3/8) 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 03:58:36,800 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:58:44,553 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7467, 1.3778, 1.6910, 1.7140, 1.8212, 1.9754, 1.5828, 1.7712], device='cuda:3'), covar=tensor([0.0192, 0.0355, 0.0170, 0.0232, 0.0215, 0.0168, 0.0332, 0.0110], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0183, 0.0168, 0.0173, 0.0182, 0.0138, 0.0184, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 03:59:20,512 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0570, 3.4004, 3.3729, 2.2752, 3.0485, 3.3727, 3.1639, 1.8250], device='cuda:3'), covar=tensor([0.0467, 0.0045, 0.0040, 0.0320, 0.0084, 0.0111, 0.0087, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0130, 0.0073, 0.0073, 0.0129, 0.0088, 0.0098, 0.0086, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 03:59:40,650 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 4850, loss[loss=0.2117, simple_loss=0.3007, pruned_loss=0.06138, over 15415.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2787, pruned_loss=0.05175, over 3207629.22 frames. ], batch size: 191, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:00:10,667 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0001, 3.0528, 2.5394, 4.7958, 3.6280, 4.1895, 1.8187, 3.0466], device='cuda:3'), covar=tensor([0.1133, 0.0587, 0.1100, 0.0097, 0.0201, 0.0415, 0.1329, 0.0789], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0165, 0.0185, 0.0166, 0.0199, 0.0209, 0.0188, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 04:00:18,889 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0611, 2.7532, 2.8735, 2.0294, 2.6620, 2.2034, 2.7619, 2.8698], device='cuda:3'), covar=tensor([0.0283, 0.0699, 0.0501, 0.1587, 0.0733, 0.0854, 0.0565, 0.0690], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0153, 0.0160, 0.0146, 0.0138, 0.0125, 0.0138, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 04:00:30,953 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9273, 2.7554, 2.7219, 1.9647, 2.4706, 2.7160, 2.6079, 1.7933], device='cuda:3'), covar=tensor([0.0353, 0.0061, 0.0052, 0.0290, 0.0107, 0.0101, 0.0103, 0.0380], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0073, 0.0073, 0.0130, 0.0088, 0.0098, 0.0086, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 04:00:34,143 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 04:01:03,666 INFO [train.py:904] (3/8) Epoch 15, batch 4900, loss[loss=0.1829, simple_loss=0.2823, pruned_loss=0.04175, over 16438.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2782, pruned_loss=0.05076, over 3187998.41 frames. ], batch size: 75, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:01:08,002 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 1.989e+02 2.229e+02 2.704e+02 6.823e+02, threshold=4.458e+02, percent-clipped=4.0 2023-04-30 04:02:16,312 INFO [train.py:904] (3/8) Epoch 15, batch 4950, loss[loss=0.1988, simple_loss=0.2859, pruned_loss=0.05585, over 12091.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2776, pruned_loss=0.05033, over 3181949.71 frames. ], batch size: 247, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:25,542 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4075, 3.3757, 3.4314, 3.5315, 3.5639, 3.2891, 3.5238, 3.6334], device='cuda:3'), covar=tensor([0.1138, 0.0791, 0.0961, 0.0558, 0.0557, 0.2631, 0.1000, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0566, 0.0698, 0.0842, 0.0718, 0.0535, 0.0561, 0.0570, 0.0662], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:03:28,736 INFO [train.py:904] (3/8) Epoch 15, batch 5000, loss[loss=0.1935, simple_loss=0.2818, pruned_loss=0.05259, over 12337.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2793, pruned_loss=0.05071, over 3168892.60 frames. ], batch size: 248, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:32,271 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.177e+02 2.680e+02 3.002e+02 5.827e+02, threshold=5.360e+02, percent-clipped=3.0 2023-04-30 04:04:12,241 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3097, 2.2525, 2.3503, 4.0172, 2.1229, 2.6083, 2.3192, 2.4153], device='cuda:3'), covar=tensor([0.1152, 0.3276, 0.2475, 0.0445, 0.3682, 0.2288, 0.3175, 0.3085], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0418, 0.0345, 0.0320, 0.0421, 0.0479, 0.0382, 0.0488], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:04:23,356 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 04:04:39,305 INFO [train.py:904] (3/8) Epoch 15, batch 5050, loss[loss=0.1867, simple_loss=0.2705, pruned_loss=0.05143, over 17029.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2801, pruned_loss=0.0504, over 3182869.45 frames. ], batch size: 53, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:44,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3939, 2.9346, 3.0369, 1.9236, 2.6750, 2.1440, 3.0326, 3.1270], device='cuda:3'), covar=tensor([0.0263, 0.0729, 0.0562, 0.1843, 0.0834, 0.0918, 0.0651, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0154, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 04:05:46,852 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:05:48,952 INFO [train.py:904] (3/8) Epoch 15, batch 5100, loss[loss=0.1948, simple_loss=0.2833, pruned_loss=0.05319, over 16422.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2787, pruned_loss=0.04977, over 3175565.41 frames. ], batch size: 146, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:52,960 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.025e+02 2.374e+02 2.786e+02 3.985e+02, threshold=4.748e+02, percent-clipped=0.0 2023-04-30 04:06:45,625 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:53,180 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 5150, loss[loss=0.2003, simple_loss=0.2966, pruned_loss=0.05205, over 15489.00 frames. ], tot_loss[loss=0.188, simple_loss=0.278, pruned_loss=0.04903, over 3168059.15 frames. ], batch size: 191, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:07:45,004 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:03,058 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:14,129 INFO [train.py:904] (3/8) Epoch 15, batch 5200, loss[loss=0.1875, simple_loss=0.2842, pruned_loss=0.04539, over 16748.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2762, pruned_loss=0.04845, over 3182679.22 frames. ], batch size: 83, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:08:14,606 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 04:08:25,224 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:55,066 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:09:27,026 INFO [train.py:904] (3/8) Epoch 15, batch 5250, loss[loss=0.1894, simple_loss=0.2804, pruned_loss=0.04923, over 16428.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2738, pruned_loss=0.0483, over 3196905.80 frames. ], batch size: 146, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:09:54,978 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:10:02,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2394, 2.0343, 2.7221, 3.1134, 2.9946, 3.7355, 2.3152, 3.5774], device='cuda:3'), covar=tensor([0.0164, 0.0458, 0.0261, 0.0259, 0.0256, 0.0114, 0.0453, 0.0087], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0182, 0.0168, 0.0173, 0.0181, 0.0138, 0.0185, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:10:37,510 INFO [train.py:904] (3/8) Epoch 15, batch 5300, loss[loss=0.1656, simple_loss=0.251, pruned_loss=0.04008, over 16715.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.27, pruned_loss=0.04666, over 3207136.97 frames. ], batch size: 76, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:10:40,967 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 1.977e+02 2.284e+02 2.754e+02 4.909e+02, threshold=4.569e+02, percent-clipped=0.0 2023-04-30 04:10:45,218 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:22,242 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-30 04:11:23,859 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:49,962 INFO [train.py:904] (3/8) Epoch 15, batch 5350, loss[loss=0.1977, simple_loss=0.2889, pruned_loss=0.05323, over 16186.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2685, pruned_loss=0.04606, over 3218644.93 frames. ], batch size: 165, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:12:00,612 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-04-30 04:12:14,718 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:12:53,009 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:01,200 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:03,282 INFO [train.py:904] (3/8) Epoch 15, batch 5400, loss[loss=0.1926, simple_loss=0.2831, pruned_loss=0.05104, over 16407.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2719, pruned_loss=0.04708, over 3215120.02 frames. ], batch size: 68, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:13:07,672 INFO [optim.py:368] (3/8) 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 04:14:13,936 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:14:20,744 INFO [train.py:904] (3/8) Epoch 15, batch 5450, loss[loss=0.2229, simple_loss=0.3042, pruned_loss=0.07075, over 16863.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2754, pruned_loss=0.04852, over 3222899.69 frames. ], batch size: 116, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:14:41,919 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 04:15:32,511 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:15:40,036 INFO [train.py:904] (3/8) Epoch 15, batch 5500, loss[loss=0.2064, simple_loss=0.2986, pruned_loss=0.05708, over 16410.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2819, pruned_loss=0.05269, over 3199735.47 frames. ], batch size: 68, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:15:45,683 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.456e+02 3.016e+02 4.198e+02 6.055e+02, threshold=6.032e+02, percent-clipped=7.0 2023-04-30 04:16:58,368 INFO [train.py:904] (3/8) Epoch 15, batch 5550, loss[loss=0.2053, simple_loss=0.2984, pruned_loss=0.05613, over 16910.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.29, pruned_loss=0.05875, over 3159831.98 frames. ], batch size: 96, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:17:10,688 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:22,922 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:18:21,668 INFO [train.py:904] (3/8) Epoch 15, batch 5600, loss[loss=0.2252, simple_loss=0.3073, pruned_loss=0.07153, over 16895.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2953, pruned_loss=0.06353, over 3115245.50 frames. ], batch size: 116, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:18:28,273 INFO [optim.py:368] (3/8) 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 04:18:53,049 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:19:46,181 INFO [train.py:904] (3/8) Epoch 15, batch 5650, loss[loss=0.2338, simple_loss=0.3128, pruned_loss=0.07741, over 16767.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2998, pruned_loss=0.0673, over 3104703.51 frames. ], batch size: 83, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:20:04,642 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:20:47,699 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 5700, loss[loss=0.2083, simple_loss=0.3006, pruned_loss=0.05799, over 16907.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3012, pruned_loss=0.06865, over 3098442.53 frames. ], batch size: 90, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:21:11,567 INFO [optim.py:368] (3/8) 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 04:21:26,331 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:22:23,104 INFO [train.py:904] (3/8) Epoch 15, batch 5750, loss[loss=0.2094, simple_loss=0.3055, pruned_loss=0.05665, over 16881.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3042, pruned_loss=0.07012, over 3076818.51 frames. ], batch size: 96, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:22:32,770 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:03,201 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:18,807 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 04:23:21,997 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7251, 1.7381, 1.6327, 1.5638, 1.9299, 1.5658, 1.5819, 1.9158], device='cuda:3'), covar=tensor([0.0152, 0.0245, 0.0306, 0.0278, 0.0178, 0.0221, 0.0157, 0.0185], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0214, 0.0206, 0.0207, 0.0213, 0.0213, 0.0217, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:23:36,680 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:44,674 INFO [train.py:904] (3/8) Epoch 15, batch 5800, loss[loss=0.2113, simple_loss=0.2974, pruned_loss=0.06259, over 16343.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3039, pruned_loss=0.0692, over 3067731.36 frames. ], batch size: 146, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:23:51,419 INFO [optim.py:368] (3/8) 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 04:24:00,128 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:12,288 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:55,226 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 5850, loss[loss=0.2092, simple_loss=0.299, pruned_loss=0.05972, over 16431.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3017, pruned_loss=0.06748, over 3075097.95 frames. ], batch size: 146, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:25:29,143 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:38,354 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:26:04,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7419, 1.7255, 1.4176, 1.4017, 1.8673, 1.4712, 1.7093, 1.9353], device='cuda:3'), covar=tensor([0.0205, 0.0291, 0.0444, 0.0374, 0.0199, 0.0296, 0.0174, 0.0181], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0215, 0.0208, 0.0209, 0.0214, 0.0214, 0.0218, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:26:30,239 INFO [train.py:904] (3/8) Epoch 15, batch 5900, loss[loss=0.1888, simple_loss=0.2767, pruned_loss=0.0504, over 17129.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.302, pruned_loss=0.06775, over 3072363.12 frames. ], batch size: 48, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:26:31,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-04-30 04:26:39,374 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.736e+02 3.208e+02 4.005e+02 8.372e+02, threshold=6.416e+02, percent-clipped=2.0 2023-04-30 04:26:52,639 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:27:16,065 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 04:27:45,978 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6743, 4.9519, 4.7258, 4.7188, 4.4440, 4.4657, 4.4057, 5.0252], device='cuda:3'), covar=tensor([0.1045, 0.0815, 0.0912, 0.0700, 0.0795, 0.0970, 0.1059, 0.0867], device='cuda:3'), in_proj_covar=tensor([0.0586, 0.0731, 0.0603, 0.0523, 0.0457, 0.0469, 0.0602, 0.0552], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:27:50,061 INFO [train.py:904] (3/8) Epoch 15, batch 5950, loss[loss=0.2369, simple_loss=0.3235, pruned_loss=0.07511, over 16596.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3024, pruned_loss=0.06653, over 3069732.62 frames. ], batch size: 57, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:28:08,751 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:28:48,995 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:29:08,855 INFO [train.py:904] (3/8) Epoch 15, batch 6000, loss[loss=0.1965, simple_loss=0.2792, pruned_loss=0.05689, over 16764.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.3013, pruned_loss=0.06609, over 3076761.83 frames. ], batch size: 39, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:29:08,855 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 04:29:19,437 INFO [train.py:938] (3/8) Epoch 15, validation: loss=0.1559, simple_loss=0.2691, pruned_loss=0.0214, over 944034.00 frames. 2023-04-30 04:29:19,437 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17808MB 2023-04-30 04:29:26,129 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:29:38,397 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 04:30:15,355 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:30:36,870 INFO [train.py:904] (3/8) Epoch 15, batch 6050, loss[loss=0.1947, simple_loss=0.2986, pruned_loss=0.04539, over 16581.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2998, pruned_loss=0.06479, over 3086005.30 frames. ], batch size: 68, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:31:07,972 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:31:59,320 INFO [train.py:904] (3/8) Epoch 15, batch 6100, loss[loss=0.2084, simple_loss=0.2951, pruned_loss=0.06086, over 17249.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2992, pruned_loss=0.06366, over 3104066.76 frames. ], batch size: 52, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:32:08,601 INFO [optim.py:368] (3/8) 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 04:32:17,970 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:33:18,245 INFO [train.py:904] (3/8) Epoch 15, batch 6150, loss[loss=0.2114, simple_loss=0.2905, pruned_loss=0.06614, over 16853.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2974, pruned_loss=0.06303, over 3114790.33 frames. ], batch size: 116, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:33:21,071 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 04:33:42,620 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:15,508 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:37,596 INFO [train.py:904] (3/8) Epoch 15, batch 6200, loss[loss=0.1923, simple_loss=0.2855, pruned_loss=0.04954, over 16872.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2957, pruned_loss=0.06265, over 3114175.83 frames. ], batch size: 102, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:34:46,168 INFO [optim.py:368] (3/8) 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 04:34:58,566 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:48,371 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:54,862 INFO [train.py:904] (3/8) Epoch 15, batch 6250, loss[loss=0.1909, simple_loss=0.2875, pruned_loss=0.04716, over 16818.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2948, pruned_loss=0.06215, over 3128208.63 frames. ], batch size: 102, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:36:11,761 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:36:47,399 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5484, 3.6360, 2.1485, 4.0931, 2.6819, 4.0606, 2.3718, 2.8699], device='cuda:3'), covar=tensor([0.0257, 0.0367, 0.1678, 0.0217, 0.0887, 0.0556, 0.1497, 0.0712], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0189, 0.0143, 0.0168, 0.0209, 0.0196, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 04:37:11,930 INFO [train.py:904] (3/8) Epoch 15, batch 6300, loss[loss=0.2103, simple_loss=0.2956, pruned_loss=0.06247, over 16932.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2947, pruned_loss=0.06189, over 3132200.80 frames. ], batch size: 109, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:37:21,867 INFO [optim.py:368] (3/8) 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 04:37:57,285 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:38:02,149 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7266, 5.0305, 5.1829, 5.0063, 4.9744, 5.5825, 5.0419, 4.8662], device='cuda:3'), covar=tensor([0.1086, 0.1877, 0.2244, 0.1915, 0.2448, 0.1030, 0.1541, 0.2424], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0545, 0.0591, 0.0461, 0.0614, 0.0622, 0.0472, 0.0618], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 04:38:33,525 INFO [train.py:904] (3/8) Epoch 15, batch 6350, loss[loss=0.2264, simple_loss=0.307, pruned_loss=0.07296, over 15315.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2953, pruned_loss=0.06303, over 3121051.95 frames. ], batch size: 190, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:38:39,007 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1279, 3.5134, 3.4873, 1.9711, 2.9042, 2.3456, 3.4365, 3.7352], device='cuda:3'), covar=tensor([0.0249, 0.0670, 0.0584, 0.1974, 0.0836, 0.0916, 0.0644, 0.0814], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0155, 0.0164, 0.0148, 0.0140, 0.0128, 0.0141, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 04:39:03,806 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:34,203 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:45,784 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 6400, loss[loss=0.1892, simple_loss=0.2781, pruned_loss=0.05016, over 16877.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2955, pruned_loss=0.06434, over 3102660.48 frames. ], batch size: 96, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:39:59,982 INFO [optim.py:368] (3/8) 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 04:40:08,930 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:14,077 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2431, 4.3108, 4.6649, 4.6250, 4.6290, 4.2824, 4.2989, 4.1625], device='cuda:3'), covar=tensor([0.0318, 0.0567, 0.0355, 0.0414, 0.0451, 0.0418, 0.0954, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0371, 0.0397, 0.0391, 0.0375, 0.0444, 0.0414, 0.0512, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 04:40:18,145 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:08,005 INFO [train.py:904] (3/8) Epoch 15, batch 6450, loss[loss=0.2231, simple_loss=0.2884, pruned_loss=0.07891, over 11631.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2953, pruned_loss=0.06368, over 3087229.87 frames. ], batch size: 246, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:41:19,894 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:22,169 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:31,956 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:33,563 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2665, 4.3668, 4.4882, 4.2810, 4.3104, 4.8257, 4.4296, 4.2083], device='cuda:3'), covar=tensor([0.1472, 0.1919, 0.2169, 0.2026, 0.2564, 0.1100, 0.1531, 0.2436], device='cuda:3'), in_proj_covar=tensor([0.0383, 0.0542, 0.0589, 0.0460, 0.0610, 0.0618, 0.0468, 0.0614], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 04:42:13,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4217, 3.3119, 2.6321, 2.0551, 2.2401, 2.1881, 3.3918, 3.1022], device='cuda:3'), covar=tensor([0.2858, 0.0847, 0.1695, 0.2619, 0.2367, 0.2037, 0.0587, 0.1239], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0259, 0.0288, 0.0289, 0.0282, 0.0233, 0.0274, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:42:26,998 INFO [train.py:904] (3/8) Epoch 15, batch 6500, loss[loss=0.2011, simple_loss=0.2872, pruned_loss=0.05746, over 16842.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2939, pruned_loss=0.06358, over 3070545.54 frames. ], batch size: 96, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:42:36,999 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.869e+02 3.331e+02 4.011e+02 8.248e+02, threshold=6.661e+02, percent-clipped=1.0 2023-04-30 04:42:47,762 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:33,196 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:46,605 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 6550, loss[loss=0.2813, simple_loss=0.3382, pruned_loss=0.1122, over 11940.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2969, pruned_loss=0.06511, over 3060557.07 frames. ], batch size: 247, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:44:57,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 04:45:05,098 INFO [train.py:904] (3/8) Epoch 15, batch 6600, loss[loss=0.2043, simple_loss=0.3007, pruned_loss=0.05398, over 16912.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2995, pruned_loss=0.06567, over 3049461.42 frames. ], batch size: 96, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:13,945 INFO [optim.py:368] (3/8) 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 04:45:19,954 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:26,315 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:46:22,112 INFO [train.py:904] (3/8) Epoch 15, batch 6650, loss[loss=0.1996, simple_loss=0.288, pruned_loss=0.05553, over 16728.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2999, pruned_loss=0.066, over 3075531.57 frames. ], batch size: 124, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:00,920 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:13,455 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 6700, loss[loss=0.2213, simple_loss=0.3052, pruned_loss=0.06871, over 16907.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2977, pruned_loss=0.06543, over 3098152.16 frames. ], batch size: 109, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:47,145 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.803e+02 3.506e+02 4.354e+02 9.571e+02, threshold=7.012e+02, percent-clipped=2.0 2023-04-30 04:48:54,917 INFO [train.py:904] (3/8) Epoch 15, batch 6750, loss[loss=0.2533, simple_loss=0.3209, pruned_loss=0.09287, over 11733.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2973, pruned_loss=0.06588, over 3081133.56 frames. ], batch size: 247, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:48:57,335 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9835, 3.0400, 1.8143, 3.2275, 2.3109, 3.2714, 2.0220, 2.4882], device='cuda:3'), covar=tensor([0.0305, 0.0415, 0.1672, 0.0229, 0.0807, 0.0593, 0.1599, 0.0750], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0188, 0.0142, 0.0168, 0.0209, 0.0196, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 04:48:58,422 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:50:03,067 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1016, 5.0949, 4.9204, 4.5488, 4.5852, 4.9607, 4.8796, 4.6430], device='cuda:3'), covar=tensor([0.0560, 0.0467, 0.0264, 0.0299, 0.0889, 0.0431, 0.0342, 0.0690], device='cuda:3'), in_proj_covar=tensor([0.0262, 0.0363, 0.0310, 0.0293, 0.0325, 0.0341, 0.0209, 0.0369], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:50:09,772 INFO [train.py:904] (3/8) Epoch 15, batch 6800, loss[loss=0.2353, simple_loss=0.319, pruned_loss=0.07578, over 16932.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2976, pruned_loss=0.06619, over 3067584.10 frames. ], batch size: 109, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:50:21,236 INFO [optim.py:368] (3/8) 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 04:51:15,862 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:51:28,456 INFO [train.py:904] (3/8) Epoch 15, batch 6850, loss[loss=0.2002, simple_loss=0.2994, pruned_loss=0.05054, over 16993.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2992, pruned_loss=0.06671, over 3059641.60 frames. ], batch size: 55, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:25,877 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 04:52:26,495 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:43,266 INFO [train.py:904] (3/8) Epoch 15, batch 6900, loss[loss=0.2143, simple_loss=0.3096, pruned_loss=0.05953, over 16403.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.3007, pruned_loss=0.06556, over 3073362.33 frames. ], batch size: 146, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:52,198 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149007.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:54,702 INFO [optim.py:368] (3/8) 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 04:53:58,309 INFO [train.py:904] (3/8) Epoch 15, batch 6950, loss[loss=0.1947, simple_loss=0.2866, pruned_loss=0.05136, over 16468.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.3023, pruned_loss=0.06724, over 3067210.57 frames. ], batch size: 68, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:53:59,958 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7993, 3.7343, 3.8525, 3.9985, 4.0702, 3.6823, 3.9887, 4.0692], device='cuda:3'), covar=tensor([0.1371, 0.1065, 0.1231, 0.0620, 0.0562, 0.1862, 0.0817, 0.0685], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0699, 0.0836, 0.0706, 0.0535, 0.0561, 0.0569, 0.0662], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:54:27,426 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:45,898 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:58,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3492, 3.3193, 3.3819, 3.4858, 3.5068, 3.2700, 3.4845, 3.5595], device='cuda:3'), covar=tensor([0.1092, 0.0980, 0.1047, 0.0610, 0.0613, 0.1989, 0.0954, 0.0717], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0699, 0.0837, 0.0707, 0.0536, 0.0562, 0.0570, 0.0663], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:55:00,333 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 04:55:10,868 INFO [train.py:904] (3/8) Epoch 15, batch 7000, loss[loss=0.2203, simple_loss=0.3147, pruned_loss=0.06298, over 16382.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.3014, pruned_loss=0.06567, over 3086990.10 frames. ], batch size: 146, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:55:23,348 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.893e+02 3.767e+02 4.998e+02 1.151e+03, threshold=7.533e+02, percent-clipped=10.0 2023-04-30 04:55:57,574 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:56:25,564 INFO [train.py:904] (3/8) Epoch 15, batch 7050, loss[loss=0.1989, simple_loss=0.2889, pruned_loss=0.05444, over 17015.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.3019, pruned_loss=0.06554, over 3080968.66 frames. ], batch size: 50, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:56:28,521 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:01,000 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2390, 4.2265, 4.6512, 4.6115, 4.6132, 4.2771, 4.3128, 4.1761], device='cuda:3'), covar=tensor([0.0301, 0.0642, 0.0365, 0.0436, 0.0488, 0.0418, 0.0971, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0394, 0.0385, 0.0371, 0.0438, 0.0411, 0.0508, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 04:57:40,343 INFO [train.py:904] (3/8) Epoch 15, batch 7100, loss[loss=0.1941, simple_loss=0.2844, pruned_loss=0.05193, over 17236.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.3003, pruned_loss=0.06514, over 3074641.45 frames. ], batch size: 52, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:57:40,682 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:53,981 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.990e+02 3.583e+02 4.304e+02 1.223e+03, threshold=7.166e+02, percent-clipped=1.0 2023-04-30 04:58:03,472 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3626, 3.3429, 3.4034, 3.5093, 3.5360, 3.2615, 3.4772, 3.5654], device='cuda:3'), covar=tensor([0.1211, 0.0984, 0.1050, 0.0604, 0.0655, 0.2650, 0.1053, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0699, 0.0838, 0.0706, 0.0538, 0.0563, 0.0570, 0.0663], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 04:58:55,136 INFO [train.py:904] (3/8) Epoch 15, batch 7150, loss[loss=0.2243, simple_loss=0.3068, pruned_loss=0.07088, over 16657.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2995, pruned_loss=0.06589, over 3050811.07 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 05:00:05,794 INFO [train.py:904] (3/8) Epoch 15, batch 7200, loss[loss=0.1885, simple_loss=0.2777, pruned_loss=0.04966, over 12160.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.297, pruned_loss=0.06462, over 3032553.88 frames. ], batch size: 246, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:00:09,224 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7865, 4.6223, 4.8262, 5.0099, 5.1641, 4.6305, 5.1314, 5.1439], device='cuda:3'), covar=tensor([0.1584, 0.1045, 0.1304, 0.0579, 0.0508, 0.0887, 0.0524, 0.0515], device='cuda:3'), in_proj_covar=tensor([0.0558, 0.0693, 0.0830, 0.0701, 0.0532, 0.0559, 0.0565, 0.0657], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:00:13,244 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:00:17,916 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.693e+02 3.090e+02 3.791e+02 7.265e+02, threshold=6.181e+02, percent-clipped=1.0 2023-04-30 05:01:17,027 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0654, 2.0397, 2.2298, 3.6833, 1.9727, 2.4287, 2.1933, 2.2302], device='cuda:3'), covar=tensor([0.1195, 0.3298, 0.2505, 0.0509, 0.4017, 0.2274, 0.3073, 0.3170], device='cuda:3'), in_proj_covar=tensor([0.0374, 0.0411, 0.0341, 0.0316, 0.0420, 0.0472, 0.0379, 0.0481], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:01:26,134 INFO [train.py:904] (3/8) Epoch 15, batch 7250, loss[loss=0.189, simple_loss=0.2668, pruned_loss=0.05565, over 16909.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2943, pruned_loss=0.06299, over 3048827.48 frames. ], batch size: 109, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:01:31,288 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:01:39,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6216, 4.8147, 5.0056, 4.8176, 4.8134, 5.3998, 4.8870, 4.5887], device='cuda:3'), covar=tensor([0.1172, 0.1806, 0.2100, 0.1771, 0.2374, 0.0970, 0.1565, 0.2626], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0533, 0.0582, 0.0452, 0.0599, 0.0612, 0.0462, 0.0605], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 05:01:57,295 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 7300, loss[loss=0.2039, simple_loss=0.295, pruned_loss=0.05637, over 16405.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2939, pruned_loss=0.06277, over 3055893.51 frames. ], batch size: 146, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:02:55,636 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 2.888e+02 3.399e+02 4.220e+02 6.338e+02, threshold=6.799e+02, percent-clipped=2.0 2023-04-30 05:03:10,460 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:03:38,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5381, 1.5554, 1.3493, 1.2669, 1.6470, 1.2354, 1.3639, 1.6711], device='cuda:3'), covar=tensor([0.0207, 0.0310, 0.0470, 0.0392, 0.0217, 0.0312, 0.0152, 0.0204], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0214, 0.0210, 0.0211, 0.0215, 0.0216, 0.0217, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:03:58,890 INFO [train.py:904] (3/8) Epoch 15, batch 7350, loss[loss=0.2174, simple_loss=0.2986, pruned_loss=0.06809, over 16805.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2952, pruned_loss=0.06404, over 3022584.95 frames. ], batch size: 116, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:04:16,142 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 05:05:17,915 INFO [train.py:904] (3/8) Epoch 15, batch 7400, loss[loss=0.2175, simple_loss=0.2965, pruned_loss=0.06926, over 16670.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2962, pruned_loss=0.06463, over 3033838.33 frames. ], batch size: 134, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:31,576 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0226, 3.0201, 3.2898, 1.5620, 3.3771, 3.5027, 2.8017, 2.5929], device='cuda:3'), covar=tensor([0.0882, 0.0227, 0.0201, 0.1383, 0.0087, 0.0169, 0.0381, 0.0539], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0104, 0.0090, 0.0137, 0.0073, 0.0113, 0.0122, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 05:05:32,175 INFO [optim.py:368] (3/8) 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 05:05:37,369 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:06:37,032 INFO [train.py:904] (3/8) Epoch 15, batch 7450, loss[loss=0.188, simple_loss=0.2674, pruned_loss=0.05432, over 16773.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2969, pruned_loss=0.06524, over 3049322.69 frames. ], batch size: 39, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:07:00,479 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:16,500 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:43,182 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-30 05:08:00,166 INFO [train.py:904] (3/8) Epoch 15, batch 7500, loss[loss=0.1818, simple_loss=0.2721, pruned_loss=0.04573, over 16828.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2963, pruned_loss=0.06396, over 3040842.14 frames. ], batch size: 102, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:08:16,081 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.987e+02 3.634e+02 4.459e+02 1.121e+03, threshold=7.267e+02, percent-clipped=5.0 2023-04-30 05:08:40,008 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:09:18,804 INFO [train.py:904] (3/8) Epoch 15, batch 7550, loss[loss=0.1966, simple_loss=0.2831, pruned_loss=0.05509, over 16690.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.296, pruned_loss=0.06453, over 3035748.96 frames. ], batch size: 76, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:10:35,811 INFO [train.py:904] (3/8) Epoch 15, batch 7600, loss[loss=0.1951, simple_loss=0.2826, pruned_loss=0.05387, over 16668.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2951, pruned_loss=0.0645, over 3043763.68 frames. ], batch size: 62, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:10:50,663 INFO [optim.py:368] (3/8) 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 05:11:45,903 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:11:55,062 INFO [train.py:904] (3/8) Epoch 15, batch 7650, loss[loss=0.223, simple_loss=0.3089, pruned_loss=0.06853, over 16901.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2958, pruned_loss=0.06572, over 3028852.52 frames. ], batch size: 109, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:12:59,165 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6036, 2.8063, 2.3561, 4.1308, 3.0273, 4.0185, 1.4363, 2.8680], device='cuda:3'), covar=tensor([0.1458, 0.0697, 0.1287, 0.0169, 0.0326, 0.0406, 0.1745, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0166, 0.0187, 0.0167, 0.0203, 0.0211, 0.0192, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 05:13:13,691 INFO [train.py:904] (3/8) Epoch 15, batch 7700, loss[loss=0.2182, simple_loss=0.2995, pruned_loss=0.0685, over 16555.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2955, pruned_loss=0.06557, over 3050185.61 frames. ], batch size: 62, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:13:22,691 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:13:28,381 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 05:13:29,265 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 3.235e+02 4.085e+02 5.006e+02 1.161e+03, threshold=8.169e+02, percent-clipped=3.0 2023-04-30 05:14:33,190 INFO [train.py:904] (3/8) Epoch 15, batch 7750, loss[loss=0.2277, simple_loss=0.3035, pruned_loss=0.07593, over 17058.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2951, pruned_loss=0.06508, over 3061462.60 frames. ], batch size: 53, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:15:02,331 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:15:33,348 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4075, 3.2583, 2.6718, 2.1028, 2.2312, 2.2331, 3.3346, 3.0517], device='cuda:3'), covar=tensor([0.2754, 0.0709, 0.1628, 0.2590, 0.2586, 0.1939, 0.0530, 0.1208], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0263, 0.0294, 0.0295, 0.0287, 0.0237, 0.0279, 0.0312], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 05:15:39,594 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 05:15:51,918 INFO [train.py:904] (3/8) Epoch 15, batch 7800, loss[loss=0.2032, simple_loss=0.2868, pruned_loss=0.0598, over 16516.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2961, pruned_loss=0.06532, over 3087444.80 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:16:07,341 INFO [optim.py:368] (3/8) 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 05:16:07,898 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0410, 4.0169, 3.9330, 3.2683, 3.9829, 1.8128, 3.7715, 3.5643], device='cuda:3'), covar=tensor([0.0138, 0.0115, 0.0171, 0.0307, 0.0103, 0.2588, 0.0132, 0.0241], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0131, 0.0178, 0.0165, 0.0150, 0.0191, 0.0166, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:16:20,885 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:17:06,632 INFO [train.py:904] (3/8) Epoch 15, batch 7850, loss[loss=0.2033, simple_loss=0.2912, pruned_loss=0.05764, over 16198.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2965, pruned_loss=0.06443, over 3107357.20 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:17:30,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8451, 2.7531, 2.3336, 2.6282, 3.1966, 2.8541, 3.5530, 3.5015], device='cuda:3'), covar=tensor([0.0084, 0.0342, 0.0437, 0.0374, 0.0219, 0.0339, 0.0170, 0.0203], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0216, 0.0211, 0.0212, 0.0216, 0.0217, 0.0218, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:17:36,274 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0661, 2.0829, 2.5241, 3.0088, 2.8782, 3.4744, 2.0584, 3.4172], device='cuda:3'), covar=tensor([0.0178, 0.0408, 0.0285, 0.0237, 0.0228, 0.0142, 0.0448, 0.0096], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0180, 0.0164, 0.0170, 0.0179, 0.0137, 0.0182, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:18:25,132 INFO [train.py:904] (3/8) Epoch 15, batch 7900, loss[loss=0.2088, simple_loss=0.2957, pruned_loss=0.06097, over 15415.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.296, pruned_loss=0.06367, over 3127951.44 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:18:28,215 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-30 05:18:34,870 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3080, 3.7683, 3.5683, 2.0625, 3.0821, 2.5284, 3.6032, 3.9969], device='cuda:3'), covar=tensor([0.0258, 0.0653, 0.0597, 0.2056, 0.0821, 0.0951, 0.0624, 0.0811], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0152, 0.0160, 0.0145, 0.0138, 0.0125, 0.0139, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 05:18:40,307 INFO [optim.py:368] (3/8) 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 05:19:00,803 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 05:19:36,892 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-30 05:19:43,691 INFO [train.py:904] (3/8) Epoch 15, batch 7950, loss[loss=0.1919, simple_loss=0.2763, pruned_loss=0.05373, over 16658.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2971, pruned_loss=0.06481, over 3119898.82 frames. ], batch size: 134, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:19:56,278 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4489, 1.6710, 2.0562, 2.3517, 2.4078, 2.7345, 1.8353, 2.7158], device='cuda:3'), covar=tensor([0.0188, 0.0469, 0.0279, 0.0288, 0.0290, 0.0171, 0.0433, 0.0102], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0182, 0.0166, 0.0171, 0.0180, 0.0137, 0.0183, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:20:20,121 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 05:20:25,881 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:20:58,679 INFO [train.py:904] (3/8) Epoch 15, batch 8000, loss[loss=0.2167, simple_loss=0.2999, pruned_loss=0.06674, over 16735.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2982, pruned_loss=0.06555, over 3108847.97 frames. ], batch size: 134, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:20:59,764 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150102.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:21:14,670 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.932e+02 3.545e+02 4.036e+02 7.060e+02, threshold=7.089e+02, percent-clipped=1.0 2023-04-30 05:21:55,946 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 8050, loss[loss=0.1895, simple_loss=0.2815, pruned_loss=0.04872, over 16690.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2985, pruned_loss=0.0655, over 3091404.06 frames. ], batch size: 89, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:22:41,243 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:29,298 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0775, 4.1471, 2.4085, 4.7954, 3.1317, 4.7144, 2.5033, 3.1439], device='cuda:3'), covar=tensor([0.0220, 0.0322, 0.1695, 0.0212, 0.0799, 0.0530, 0.1571, 0.0792], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0171, 0.0190, 0.0143, 0.0171, 0.0211, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 05:23:30,576 INFO [train.py:904] (3/8) Epoch 15, batch 8100, loss[loss=0.1687, simple_loss=0.2532, pruned_loss=0.04211, over 16989.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2972, pruned_loss=0.06468, over 3098793.68 frames. ], batch size: 41, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:23:45,513 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.677e+02 3.415e+02 4.132e+02 1.188e+03, threshold=6.830e+02, percent-clipped=3.0 2023-04-30 05:23:56,039 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:59,896 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:24:26,144 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8885, 2.7194, 2.5761, 2.0430, 2.5548, 2.6861, 2.5889, 1.9118], device='cuda:3'), covar=tensor([0.0401, 0.0067, 0.0063, 0.0323, 0.0110, 0.0103, 0.0099, 0.0364], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0074, 0.0073, 0.0132, 0.0088, 0.0098, 0.0085, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 05:24:45,702 INFO [train.py:904] (3/8) Epoch 15, batch 8150, loss[loss=0.175, simple_loss=0.2584, pruned_loss=0.04575, over 16588.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2946, pruned_loss=0.06374, over 3092919.69 frames. ], batch size: 76, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:25:11,110 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:19,658 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:26:00,377 INFO [train.py:904] (3/8) Epoch 15, batch 8200, loss[loss=0.2211, simple_loss=0.3029, pruned_loss=0.06965, over 16649.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2916, pruned_loss=0.06289, over 3092416.33 frames. ], batch size: 134, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:26:16,345 INFO [optim.py:368] (3/8) 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 05:26:54,561 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:27:22,317 INFO [train.py:904] (3/8) Epoch 15, batch 8250, loss[loss=0.18, simple_loss=0.2721, pruned_loss=0.04392, over 15309.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.29, pruned_loss=0.05991, over 3091942.27 frames. ], batch size: 191, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:45,004 INFO [train.py:904] (3/8) Epoch 15, batch 8300, loss[loss=0.1898, simple_loss=0.2829, pruned_loss=0.04832, over 15253.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2876, pruned_loss=0.05719, over 3097922.77 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:45,701 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 05:28:57,545 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6124, 2.1880, 2.2678, 4.4406, 2.1615, 2.7040, 2.3026, 2.4064], device='cuda:3'), covar=tensor([0.0945, 0.3673, 0.2771, 0.0331, 0.4223, 0.2416, 0.3604, 0.3369], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0409, 0.0340, 0.0313, 0.0418, 0.0469, 0.0377, 0.0478], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:29:01,272 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.394e+02 2.913e+02 3.586e+02 6.520e+02, threshold=5.826e+02, percent-clipped=0.0 2023-04-30 05:29:39,971 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:30:05,306 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:30:07,756 INFO [train.py:904] (3/8) Epoch 15, batch 8350, loss[loss=0.1934, simple_loss=0.2956, pruned_loss=0.04562, over 16882.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2868, pruned_loss=0.0556, over 3065143.73 frames. ], batch size: 116, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:30:49,923 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 05:31:29,420 INFO [train.py:904] (3/8) Epoch 15, batch 8400, loss[loss=0.176, simple_loss=0.2658, pruned_loss=0.04306, over 15247.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2842, pruned_loss=0.05352, over 3059249.80 frames. ], batch size: 191, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:46,287 INFO [optim.py:368] (3/8) 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 05:32:27,358 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:32:31,837 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5254, 3.4205, 3.7435, 1.9284, 3.8824, 3.9467, 3.0566, 3.0979], device='cuda:3'), covar=tensor([0.0689, 0.0205, 0.0152, 0.1097, 0.0044, 0.0123, 0.0329, 0.0356], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0103, 0.0088, 0.0135, 0.0071, 0.0111, 0.0120, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 05:32:49,402 INFO [train.py:904] (3/8) Epoch 15, batch 8450, loss[loss=0.1767, simple_loss=0.2733, pruned_loss=0.04008, over 16737.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2821, pruned_loss=0.0516, over 3062052.06 frames. ], batch size: 76, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:32:51,969 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 05:33:35,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7890, 3.7421, 4.1083, 4.1066, 4.0829, 3.8672, 3.8579, 3.8900], device='cuda:3'), covar=tensor([0.0373, 0.0965, 0.0453, 0.0424, 0.0502, 0.0486, 0.1043, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0363, 0.0391, 0.0379, 0.0366, 0.0434, 0.0404, 0.0499, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 05:34:02,268 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 8500, loss[loss=0.1782, simple_loss=0.2677, pruned_loss=0.04438, over 16682.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2783, pruned_loss=0.04932, over 3042483.52 frames. ], batch size: 76, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:25,005 INFO [optim.py:368] (3/8) 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 05:34:54,270 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:35:30,232 INFO [train.py:904] (3/8) Epoch 15, batch 8550, loss[loss=0.192, simple_loss=0.2943, pruned_loss=0.04484, over 15301.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2758, pruned_loss=0.04836, over 2997166.58 frames. ], batch size: 190, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:35:56,486 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5555, 4.8625, 4.6535, 4.6776, 4.3785, 4.3779, 4.3371, 4.8916], device='cuda:3'), covar=tensor([0.1117, 0.0897, 0.0953, 0.0727, 0.0837, 0.1043, 0.1121, 0.0884], device='cuda:3'), in_proj_covar=tensor([0.0584, 0.0722, 0.0598, 0.0520, 0.0452, 0.0467, 0.0601, 0.0552], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:36:22,882 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7510, 4.9703, 5.1353, 4.9490, 4.9119, 5.5233, 5.0287, 4.7716], device='cuda:3'), covar=tensor([0.0974, 0.1743, 0.1770, 0.1678, 0.2343, 0.0912, 0.1423, 0.2119], device='cuda:3'), in_proj_covar=tensor([0.0367, 0.0516, 0.0566, 0.0434, 0.0579, 0.0595, 0.0448, 0.0583], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 05:37:12,303 INFO [train.py:904] (3/8) Epoch 15, batch 8600, loss[loss=0.1869, simple_loss=0.2803, pruned_loss=0.04671, over 15432.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2763, pruned_loss=0.04736, over 3014384.17 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:32,420 INFO [optim.py:368] (3/8) 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 05:38:11,162 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 05:38:14,978 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:20,230 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:39,834 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5364, 2.5114, 2.2064, 3.7119, 2.0177, 3.7495, 1.3967, 2.7457], device='cuda:3'), covar=tensor([0.1573, 0.0797, 0.1413, 0.0174, 0.0143, 0.0405, 0.1786, 0.0791], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0162, 0.0184, 0.0162, 0.0195, 0.0205, 0.0187, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-30 05:38:51,966 INFO [train.py:904] (3/8) Epoch 15, batch 8650, loss[loss=0.1798, simple_loss=0.2694, pruned_loss=0.04513, over 12143.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2747, pruned_loss=0.04573, over 3023487.73 frames. ], batch size: 248, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:40:03,478 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:40:03,925 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 05:40:23,191 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:40:39,991 INFO [train.py:904] (3/8) Epoch 15, batch 8700, loss[loss=0.1788, simple_loss=0.2696, pruned_loss=0.04396, over 16612.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.272, pruned_loss=0.04448, over 3015282.44 frames. ], batch size: 134, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:41:01,898 INFO [optim.py:368] (3/8) 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 05:41:11,224 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1120, 2.5260, 2.7123, 1.8970, 2.7441, 2.8885, 2.4937, 2.5113], device='cuda:3'), covar=tensor([0.0644, 0.0223, 0.0178, 0.0960, 0.0080, 0.0182, 0.0374, 0.0382], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0101, 0.0087, 0.0134, 0.0070, 0.0110, 0.0119, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 05:41:55,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6301, 4.1244, 4.1423, 3.0294, 3.6861, 4.1348, 3.8560, 2.6030], device='cuda:3'), covar=tensor([0.0383, 0.0030, 0.0027, 0.0274, 0.0065, 0.0053, 0.0046, 0.0328], device='cuda:3'), in_proj_covar=tensor([0.0128, 0.0072, 0.0071, 0.0129, 0.0085, 0.0094, 0.0082, 0.0120], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 05:42:16,614 INFO [train.py:904] (3/8) Epoch 15, batch 8750, loss[loss=0.1936, simple_loss=0.2901, pruned_loss=0.04851, over 15192.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2722, pruned_loss=0.04389, over 3022347.96 frames. ], batch size: 190, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:42:46,787 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 05:43:41,307 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2089, 3.5490, 3.7349, 1.9818, 3.1423, 2.4705, 3.5608, 3.6794], device='cuda:3'), covar=tensor([0.0248, 0.0744, 0.0454, 0.1917, 0.0698, 0.0825, 0.0599, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0148, 0.0157, 0.0144, 0.0136, 0.0123, 0.0136, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 05:43:52,659 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:06,444 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 8800, loss[loss=0.1812, simple_loss=0.2786, pruned_loss=0.04187, over 16709.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2699, pruned_loss=0.04266, over 3032339.12 frames. ], batch size: 134, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:44:28,831 INFO [optim.py:368] (3/8) 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 05:45:06,321 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 8850, loss[loss=0.1529, simple_loss=0.2448, pruned_loss=0.03051, over 12134.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2725, pruned_loss=0.04212, over 3028653.06 frames. ], batch size: 247, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:46:10,224 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:46:17,332 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-30 05:46:46,307 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 8900, loss[loss=0.2069, simple_loss=0.3011, pruned_loss=0.0563, over 16256.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2727, pruned_loss=0.04166, over 3031548.78 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:47:59,489 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 15, batch 8950, loss[loss=0.1533, simple_loss=0.2503, pruned_loss=0.02813, over 15469.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2729, pruned_loss=0.04203, over 3052546.41 frames. ], batch size: 192, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:01,603 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:51:16,287 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 05:51:30,995 INFO [train.py:904] (3/8) Epoch 15, batch 9000, loss[loss=0.1722, simple_loss=0.2679, pruned_loss=0.03821, over 15398.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2695, pruned_loss=0.04067, over 3055105.92 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:30,995 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 05:51:40,828 INFO [train.py:938] (3/8) Epoch 15, validation: loss=0.15, simple_loss=0.2539, pruned_loss=0.02307, over 944034.00 frames. 2023-04-30 05:51:40,828 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 05:52:03,718 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 2.164e+02 2.577e+02 3.241e+02 6.734e+02, threshold=5.154e+02, percent-clipped=2.0 2023-04-30 05:53:23,275 INFO [train.py:904] (3/8) Epoch 15, batch 9050, loss[loss=0.1459, simple_loss=0.2367, pruned_loss=0.02752, over 16770.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2704, pruned_loss=0.0412, over 3061620.42 frames. ], batch size: 76, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:54:28,333 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:54:51,798 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 9100, loss[loss=0.1964, simple_loss=0.2948, pruned_loss=0.04896, over 15497.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2694, pruned_loss=0.04139, over 3072153.92 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:55:31,071 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.276e+02 2.684e+02 3.281e+02 5.650e+02, threshold=5.369e+02, percent-clipped=3.0 2023-04-30 05:55:55,721 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9684, 4.2350, 4.0962, 4.0803, 3.7845, 3.8295, 3.8478, 4.2546], device='cuda:3'), covar=tensor([0.1045, 0.0852, 0.0832, 0.0754, 0.0775, 0.1499, 0.0953, 0.0886], device='cuda:3'), in_proj_covar=tensor([0.0578, 0.0716, 0.0585, 0.0513, 0.0448, 0.0465, 0.0591, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 05:56:44,480 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:56:54,393 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:57:08,972 INFO [train.py:904] (3/8) Epoch 15, batch 9150, loss[loss=0.1593, simple_loss=0.2511, pruned_loss=0.03374, over 16262.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2693, pruned_loss=0.041, over 3065788.08 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:57:18,558 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:08,951 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 05:58:12,605 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:46,503 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:51,677 INFO [train.py:904] (3/8) Epoch 15, batch 9200, loss[loss=0.1726, simple_loss=0.2677, pruned_loss=0.03879, over 16166.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2657, pruned_loss=0.0403, over 3060549.92 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:59:12,237 INFO [optim.py:368] (3/8) 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 06:00:09,384 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:00:16,444 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 06:00:24,879 INFO [train.py:904] (3/8) Epoch 15, batch 9250, loss[loss=0.1601, simple_loss=0.2391, pruned_loss=0.04053, over 12105.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2657, pruned_loss=0.04053, over 3060823.41 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:00:41,069 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:01:43,243 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:02:14,982 INFO [train.py:904] (3/8) Epoch 15, batch 9300, loss[loss=0.1834, simple_loss=0.2602, pruned_loss=0.05332, over 12676.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2647, pruned_loss=0.04014, over 3059548.13 frames. ], batch size: 250, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:02:21,128 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5008, 4.4530, 4.3237, 3.8263, 4.4064, 1.7514, 4.1490, 4.1378], device='cuda:3'), covar=tensor([0.0071, 0.0080, 0.0135, 0.0211, 0.0073, 0.2359, 0.0110, 0.0172], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0127, 0.0173, 0.0156, 0.0147, 0.0187, 0.0162, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:02:37,911 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.225e+02 2.604e+02 3.286e+02 5.862e+02, threshold=5.207e+02, percent-clipped=0.0 2023-04-30 06:02:51,475 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5122, 4.5802, 4.3708, 4.1022, 4.1161, 4.5023, 4.2202, 4.1781], device='cuda:3'), covar=tensor([0.0522, 0.0539, 0.0274, 0.0252, 0.0753, 0.0480, 0.0540, 0.0610], device='cuda:3'), in_proj_covar=tensor([0.0253, 0.0348, 0.0299, 0.0282, 0.0307, 0.0326, 0.0205, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:03:29,231 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:03:59,125 INFO [train.py:904] (3/8) Epoch 15, batch 9350, loss[loss=0.1758, simple_loss=0.2596, pruned_loss=0.04603, over 12572.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2648, pruned_loss=0.04004, over 3080682.16 frames. ], batch size: 250, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:04:17,700 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-04-30 06:05:36,963 INFO [train.py:904] (3/8) Epoch 15, batch 9400, loss[loss=0.1848, simple_loss=0.2864, pruned_loss=0.04157, over 16370.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2654, pruned_loss=0.03979, over 3089786.93 frames. ], batch size: 146, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:05:59,179 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.126e+02 2.557e+02 3.057e+02 4.455e+02, threshold=5.114e+02, percent-clipped=0.0 2023-04-30 06:06:55,071 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:07:16,801 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0711, 4.0139, 3.9463, 3.3958, 3.9904, 1.7963, 3.7538, 3.6171], device='cuda:3'), covar=tensor([0.0081, 0.0086, 0.0148, 0.0215, 0.0092, 0.2489, 0.0114, 0.0202], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0127, 0.0172, 0.0156, 0.0147, 0.0187, 0.0162, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:07:17,456 INFO [train.py:904] (3/8) Epoch 15, batch 9450, loss[loss=0.1618, simple_loss=0.2598, pruned_loss=0.03195, over 15412.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2664, pruned_loss=0.03967, over 3081108.07 frames. ], batch size: 191, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:07:24,690 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:08:31,988 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7760, 2.3831, 2.3618, 3.5456, 2.1508, 3.7171, 1.3946, 2.9084], device='cuda:3'), covar=tensor([0.1393, 0.0793, 0.1158, 0.0182, 0.0095, 0.0356, 0.1720, 0.0670], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0160, 0.0181, 0.0158, 0.0187, 0.0201, 0.0185, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:3') 2023-04-30 06:08:58,269 INFO [train.py:904] (3/8) Epoch 15, batch 9500, loss[loss=0.1762, simple_loss=0.273, pruned_loss=0.03973, over 16803.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2652, pruned_loss=0.03922, over 3079090.53 frames. ], batch size: 124, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:09:04,128 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:09:22,317 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.144e+02 2.700e+02 3.388e+02 6.768e+02, threshold=5.400e+02, percent-clipped=4.0 2023-04-30 06:10:11,838 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:10:44,450 INFO [train.py:904] (3/8) Epoch 15, batch 9550, loss[loss=0.1759, simple_loss=0.2678, pruned_loss=0.04202, over 12524.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2653, pruned_loss=0.03958, over 3085005.67 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:10:49,181 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:11:58,492 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:12:24,738 INFO [train.py:904] (3/8) Epoch 15, batch 9600, loss[loss=0.1827, simple_loss=0.2663, pruned_loss=0.04951, over 12387.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.267, pruned_loss=0.04057, over 3065850.80 frames. ], batch size: 247, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:12:44,296 INFO [optim.py:368] (3/8) 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 06:14:04,281 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:14:11,686 INFO [train.py:904] (3/8) Epoch 15, batch 9650, loss[loss=0.1822, simple_loss=0.275, pruned_loss=0.04468, over 16640.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2684, pruned_loss=0.04083, over 3057544.46 frames. ], batch size: 57, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:15:20,886 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 06:15:58,143 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 9700, loss[loss=0.1774, simple_loss=0.2734, pruned_loss=0.04075, over 16645.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2673, pruned_loss=0.04055, over 3066555.79 frames. ], batch size: 134, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:16:19,921 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.225e+02 2.684e+02 3.461e+02 6.863e+02, threshold=5.368e+02, percent-clipped=3.0 2023-04-30 06:16:24,624 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 06:17:18,826 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 15, batch 9750, loss[loss=0.197, simple_loss=0.2821, pruned_loss=0.05598, over 16977.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2665, pruned_loss=0.0408, over 3062917.18 frames. ], batch size: 109, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:18:01,513 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:19:18,821 INFO [train.py:904] (3/8) Epoch 15, batch 9800, loss[loss=0.1611, simple_loss=0.2637, pruned_loss=0.02924, over 16851.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2668, pruned_loss=0.03988, over 3073841.85 frames. ], batch size: 96, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:19:40,658 INFO [optim.py:368] (3/8) 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 06:20:30,607 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:03,076 INFO [train.py:904] (3/8) Epoch 15, batch 9850, loss[loss=0.1987, simple_loss=0.2975, pruned_loss=0.04997, over 16986.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2672, pruned_loss=0.03952, over 3075516.78 frames. ], batch size: 109, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:21:08,475 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:17,653 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:57,939 INFO [train.py:904] (3/8) Epoch 15, batch 9900, loss[loss=0.1748, simple_loss=0.2763, pruned_loss=0.03661, over 16896.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2672, pruned_loss=0.03928, over 3071238.48 frames. ], batch size: 116, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:22:58,682 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 06:23:31,379 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6700, 2.6456, 1.8261, 2.8250, 2.0219, 2.8279, 2.0935, 2.3704], device='cuda:3'), covar=tensor([0.0291, 0.0406, 0.1369, 0.0211, 0.0731, 0.0533, 0.1216, 0.0641], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0161, 0.0182, 0.0135, 0.0165, 0.0197, 0.0192, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 06:24:37,869 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:24:55,583 INFO [train.py:904] (3/8) Epoch 15, batch 9950, loss[loss=0.1859, simple_loss=0.2843, pruned_loss=0.04371, over 16662.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2692, pruned_loss=0.0397, over 3070196.20 frames. ], batch size: 134, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:26:56,985 INFO [train.py:904] (3/8) Epoch 15, batch 10000, loss[loss=0.1635, simple_loss=0.2517, pruned_loss=0.03764, over 12733.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2682, pruned_loss=0.03937, over 3078296.36 frames. ], batch size: 248, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:27:17,746 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2576, 3.5300, 3.7981, 2.0814, 3.1032, 2.3377, 3.6357, 3.5959], device='cuda:3'), covar=tensor([0.0224, 0.0761, 0.0457, 0.1922, 0.0725, 0.0906, 0.0628, 0.0945], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0144, 0.0156, 0.0145, 0.0135, 0.0122, 0.0136, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 06:27:18,765 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.235e+02 2.836e+02 3.494e+02 9.282e+02, threshold=5.672e+02, percent-clipped=5.0 2023-04-30 06:27:33,913 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 06:28:35,912 INFO [train.py:904] (3/8) Epoch 15, batch 10050, loss[loss=0.1541, simple_loss=0.2546, pruned_loss=0.02681, over 16526.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2678, pruned_loss=0.03893, over 3067970.56 frames. ], batch size: 75, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:28:45,744 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:30:08,504 INFO [train.py:904] (3/8) Epoch 15, batch 10100, loss[loss=0.1637, simple_loss=0.2551, pruned_loss=0.0361, over 16890.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2685, pruned_loss=0.03939, over 3073755.41 frames. ], batch size: 96, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:30:28,184 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.142e+02 2.576e+02 3.219e+02 1.174e+03, threshold=5.152e+02, percent-clipped=3.0 2023-04-30 06:30:55,286 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:31:53,396 INFO [train.py:904] (3/8) Epoch 16, batch 0, loss[loss=0.186, simple_loss=0.2758, pruned_loss=0.04815, over 17179.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2758, pruned_loss=0.04815, over 17179.00 frames. ], batch size: 46, lr: 4.32e-03, grad_scale: 8.0 2023-04-30 06:31:53,397 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 06:32:00,892 INFO [train.py:938] (3/8) Epoch 16, validation: loss=0.1492, simple_loss=0.2525, pruned_loss=0.02288, over 944034.00 frames. 2023-04-30 06:32:00,893 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 06:32:29,714 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:32:48,559 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:33:09,961 INFO [train.py:904] (3/8) Epoch 16, batch 50, loss[loss=0.2148, simple_loss=0.2883, pruned_loss=0.07069, over 16492.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2787, pruned_loss=0.05561, over 748048.64 frames. ], batch size: 75, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:33:29,877 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 2.475e+02 3.034e+02 3.841e+02 6.460e+02, threshold=6.067e+02, percent-clipped=5.0 2023-04-30 06:33:53,581 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:07,534 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 100, loss[loss=0.2132, simple_loss=0.297, pruned_loss=0.06468, over 12381.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.275, pruned_loss=0.05285, over 1311585.24 frames. ], batch size: 247, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:34:25,694 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 06:35:14,870 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:35:26,676 INFO [train.py:904] (3/8) Epoch 16, batch 150, loss[loss=0.1669, simple_loss=0.2576, pruned_loss=0.03811, over 17224.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.272, pruned_loss=0.05059, over 1759561.31 frames. ], batch size: 45, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:48,056 INFO [optim.py:368] (3/8) 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 06:36:24,956 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:35,179 INFO [train.py:904] (3/8) Epoch 16, batch 200, loss[loss=0.2003, simple_loss=0.2744, pruned_loss=0.0631, over 16791.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2721, pruned_loss=0.05162, over 2099317.14 frames. ], batch size: 102, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:36:42,854 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:58,599 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4674, 5.4371, 5.2077, 4.7930, 5.2722, 2.2923, 5.0290, 5.1790], device='cuda:3'), covar=tensor([0.0067, 0.0068, 0.0153, 0.0282, 0.0072, 0.2171, 0.0111, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0131, 0.0176, 0.0159, 0.0151, 0.0192, 0.0165, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:37:09,273 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0918, 5.7147, 5.8649, 5.5891, 5.7037, 6.1951, 5.7263, 5.5057], device='cuda:3'), covar=tensor([0.0943, 0.1861, 0.2023, 0.1944, 0.2521, 0.1033, 0.1455, 0.2193], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0536, 0.0583, 0.0450, 0.0596, 0.0622, 0.0458, 0.0597], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 06:37:44,240 INFO [train.py:904] (3/8) Epoch 16, batch 250, loss[loss=0.1645, simple_loss=0.2557, pruned_loss=0.03658, over 17045.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2695, pruned_loss=0.05071, over 2368907.63 frames. ], batch size: 50, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:37:48,058 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:50,093 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:51,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0914, 3.4448, 2.8975, 1.8499, 2.5883, 2.1509, 3.4655, 3.6023], device='cuda:3'), covar=tensor([0.0268, 0.0730, 0.0940, 0.2397, 0.1122, 0.1222, 0.0592, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0148, 0.0159, 0.0147, 0.0138, 0.0125, 0.0138, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 06:38:05,710 INFO [optim.py:368] (3/8) 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 06:38:53,026 INFO [train.py:904] (3/8) Epoch 16, batch 300, loss[loss=0.1812, simple_loss=0.2535, pruned_loss=0.05442, over 16865.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2655, pruned_loss=0.04883, over 2585335.03 frames. ], batch size: 116, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:39:33,771 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:35,004 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:43,036 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:01,537 INFO [train.py:904] (3/8) Epoch 16, batch 350, loss[loss=0.1813, simple_loss=0.2553, pruned_loss=0.05361, over 16239.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2627, pruned_loss=0.04728, over 2752959.19 frames. ], batch size: 165, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:40:20,730 INFO [optim.py:368] (3/8) 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 06:40:38,376 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:58,721 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:06,976 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:08,822 INFO [train.py:904] (3/8) Epoch 16, batch 400, loss[loss=0.2113, simple_loss=0.2796, pruned_loss=0.07146, over 16864.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2616, pruned_loss=0.04692, over 2874106.76 frames. ], batch size: 116, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:41:12,741 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 2023-04-30 06:42:10,599 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2635, 5.2613, 5.0817, 4.5100, 5.0356, 2.2589, 4.8575, 4.9981], device='cuda:3'), covar=tensor([0.0084, 0.0078, 0.0170, 0.0387, 0.0103, 0.2286, 0.0121, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0135, 0.0181, 0.0164, 0.0155, 0.0196, 0.0170, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:42:16,091 INFO [train.py:904] (3/8) Epoch 16, batch 450, loss[loss=0.1449, simple_loss=0.2297, pruned_loss=0.03, over 17020.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2588, pruned_loss=0.04576, over 2968065.13 frames. ], batch size: 41, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:42:31,277 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0414, 5.0197, 4.8285, 4.3144, 4.8472, 2.1790, 4.6486, 4.6448], device='cuda:3'), covar=tensor([0.0086, 0.0077, 0.0172, 0.0370, 0.0105, 0.2313, 0.0120, 0.0211], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0134, 0.0180, 0.0164, 0.0155, 0.0196, 0.0169, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:42:36,232 INFO [optim.py:368] (3/8) 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 06:42:50,479 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:43:25,827 INFO [train.py:904] (3/8) Epoch 16, batch 500, loss[loss=0.1812, simple_loss=0.2575, pruned_loss=0.05248, over 16835.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2574, pruned_loss=0.04516, over 3046580.09 frames. ], batch size: 116, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:14,797 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:33,458 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:34,223 INFO [train.py:904] (3/8) Epoch 16, batch 550, loss[loss=0.1869, simple_loss=0.261, pruned_loss=0.05639, over 16845.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2563, pruned_loss=0.04486, over 3105166.57 frames. ], batch size: 96, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:40,355 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5417, 5.9545, 5.7006, 5.7363, 5.3425, 5.3686, 5.3300, 6.1033], device='cuda:3'), covar=tensor([0.1329, 0.0944, 0.1076, 0.0847, 0.1041, 0.0732, 0.1121, 0.0865], device='cuda:3'), in_proj_covar=tensor([0.0614, 0.0759, 0.0616, 0.0546, 0.0476, 0.0487, 0.0631, 0.0577], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:44:55,531 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 2.218e+02 2.743e+02 3.343e+02 5.376e+02, threshold=5.487e+02, percent-clipped=2.0 2023-04-30 06:45:17,908 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 06:45:46,473 INFO [train.py:904] (3/8) Epoch 16, batch 600, loss[loss=0.162, simple_loss=0.2449, pruned_loss=0.03955, over 12457.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2568, pruned_loss=0.04569, over 3141905.59 frames. ], batch size: 246, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:46:14,752 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 06:46:25,176 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:46:41,377 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 06:46:53,580 INFO [train.py:904] (3/8) Epoch 16, batch 650, loss[loss=0.1587, simple_loss=0.2513, pruned_loss=0.03305, over 17209.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.256, pruned_loss=0.04531, over 3187115.70 frames. ], batch size: 44, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:47:14,336 INFO [optim.py:368] (3/8) 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 06:47:30,634 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:30,807 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:43,506 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 16, batch 700, loss[loss=0.1504, simple_loss=0.2341, pruned_loss=0.03335, over 17008.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.256, pruned_loss=0.04521, over 3225619.80 frames. ], batch size: 41, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:48:37,714 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:49:10,743 INFO [train.py:904] (3/8) Epoch 16, batch 750, loss[loss=0.1733, simple_loss=0.2648, pruned_loss=0.0409, over 17081.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.256, pruned_loss=0.04537, over 3244226.08 frames. ], batch size: 55, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:49:31,101 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.303e+02 2.598e+02 3.090e+02 5.870e+02, threshold=5.196e+02, percent-clipped=1.0 2023-04-30 06:49:38,888 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 06:50:17,790 INFO [train.py:904] (3/8) Epoch 16, batch 800, loss[loss=0.1904, simple_loss=0.2639, pruned_loss=0.05847, over 16770.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2559, pruned_loss=0.04489, over 3261575.56 frames. ], batch size: 134, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:00,660 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:25,516 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 850, loss[loss=0.1691, simple_loss=0.2729, pruned_loss=0.03263, over 17265.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2561, pruned_loss=0.04499, over 3275724.22 frames. ], batch size: 52, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:46,528 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.045e+02 2.529e+02 2.957e+02 4.458e+02, threshold=5.058e+02, percent-clipped=0.0 2023-04-30 06:51:46,924 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:12,989 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4328, 1.7262, 2.0565, 2.2430, 2.3897, 2.3928, 1.6408, 2.4823], device='cuda:3'), covar=tensor([0.0164, 0.0416, 0.0265, 0.0277, 0.0254, 0.0255, 0.0479, 0.0138], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0173, 0.0182, 0.0140, 0.0185, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 06:52:32,206 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 900, loss[loss=0.1723, simple_loss=0.2512, pruned_loss=0.04664, over 16443.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2553, pruned_loss=0.04407, over 3283359.43 frames. ], batch size: 75, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:53:10,968 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:53:36,555 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5012, 3.5585, 3.8608, 2.7788, 3.5387, 3.9644, 3.7455, 2.4312], device='cuda:3'), covar=tensor([0.0435, 0.0153, 0.0046, 0.0314, 0.0089, 0.0082, 0.0074, 0.0390], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0075, 0.0074, 0.0131, 0.0088, 0.0097, 0.0086, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 06:53:43,174 INFO [train.py:904] (3/8) Epoch 16, batch 950, loss[loss=0.1987, simple_loss=0.2746, pruned_loss=0.06134, over 15580.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2556, pruned_loss=0.04419, over 3291988.14 frames. ], batch size: 191, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:54:04,602 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:42,476 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:53,047 INFO [train.py:904] (3/8) Epoch 16, batch 1000, loss[loss=0.1583, simple_loss=0.2457, pruned_loss=0.03539, over 17191.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2544, pruned_loss=0.0439, over 3301767.90 frames. ], batch size: 44, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:55:29,793 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:36,400 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 06:55:40,234 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:50,996 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 1050, loss[loss=0.1674, simple_loss=0.2664, pruned_loss=0.03421, over 17202.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2544, pruned_loss=0.04331, over 3316377.29 frames. ], batch size: 44, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:56:24,681 INFO [optim.py:368] (3/8) 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 06:56:55,449 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:57:12,821 INFO [train.py:904] (3/8) Epoch 16, batch 1100, loss[loss=0.159, simple_loss=0.253, pruned_loss=0.0325, over 17097.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2536, pruned_loss=0.0434, over 3314710.66 frames. ], batch size: 47, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:57:54,302 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:21,562 INFO [train.py:904] (3/8) Epoch 16, batch 1150, loss[loss=0.2029, simple_loss=0.2932, pruned_loss=0.0563, over 17035.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2529, pruned_loss=0.04293, over 3314636.90 frames. ], batch size: 53, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:58:42,013 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:59:27,881 INFO [train.py:904] (3/8) Epoch 16, batch 1200, loss[loss=0.1924, simple_loss=0.2822, pruned_loss=0.05128, over 16682.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2522, pruned_loss=0.0425, over 3320354.81 frames. ], batch size: 57, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 06:59:57,452 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:10,709 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:34,402 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0386, 4.6752, 4.8280, 5.2233, 5.3980, 4.8153, 5.4333, 5.4157], device='cuda:3'), covar=tensor([0.1826, 0.1589, 0.2656, 0.1103, 0.0826, 0.0782, 0.0782, 0.0787], device='cuda:3'), in_proj_covar=tensor([0.0611, 0.0759, 0.0899, 0.0770, 0.0578, 0.0603, 0.0618, 0.0712], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:00:37,161 INFO [train.py:904] (3/8) Epoch 16, batch 1250, loss[loss=0.1829, simple_loss=0.2739, pruned_loss=0.04588, over 17060.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2528, pruned_loss=0.04337, over 3321768.60 frames. ], batch size: 50, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:00:57,396 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 16, batch 1300, loss[loss=0.1831, simple_loss=0.262, pruned_loss=0.05215, over 16915.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2527, pruned_loss=0.04324, over 3319506.97 frames. ], batch size: 109, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:02:24,196 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 07:02:44,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9456, 3.8611, 4.4751, 2.1740, 4.5858, 4.7429, 3.3107, 3.5818], device='cuda:3'), covar=tensor([0.0658, 0.0223, 0.0225, 0.1089, 0.0081, 0.0126, 0.0394, 0.0357], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 07:02:52,307 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 07:02:52,642 INFO [train.py:904] (3/8) Epoch 16, batch 1350, loss[loss=0.1883, simple_loss=0.2639, pruned_loss=0.05635, over 16676.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2523, pruned_loss=0.04334, over 3309302.19 frames. ], batch size: 76, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:03:12,843 INFO [optim.py:368] (3/8) 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 07:03:37,022 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:04:02,079 INFO [train.py:904] (3/8) Epoch 16, batch 1400, loss[loss=0.1707, simple_loss=0.2588, pruned_loss=0.0413, over 16749.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2523, pruned_loss=0.04292, over 3308819.23 frames. ], batch size: 57, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:05:12,109 INFO [train.py:904] (3/8) Epoch 16, batch 1450, loss[loss=0.1675, simple_loss=0.2431, pruned_loss=0.04595, over 11822.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.251, pruned_loss=0.04226, over 3300027.54 frames. ], batch size: 246, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:05:34,100 INFO [optim.py:368] (3/8) 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 07:06:11,388 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7150, 3.8231, 2.4935, 4.3463, 2.9592, 4.3373, 2.5690, 3.1673], device='cuda:3'), covar=tensor([0.0270, 0.0339, 0.1443, 0.0257, 0.0735, 0.0395, 0.1321, 0.0596], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0193, 0.0152, 0.0174, 0.0214, 0.0202, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:06:22,460 INFO [train.py:904] (3/8) Epoch 16, batch 1500, loss[loss=0.1451, simple_loss=0.2294, pruned_loss=0.0304, over 16784.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2506, pruned_loss=0.04212, over 3306444.23 frames. ], batch size: 39, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:06:50,907 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:06:56,028 INFO [zipformer.py:625] (3/8) 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] (3/8) attn_weights_entropy = tensor([4.8242, 4.8904, 5.3818, 5.3341, 5.3519, 4.9482, 4.8750, 4.7766], device='cuda:3'), covar=tensor([0.0354, 0.0540, 0.0443, 0.0506, 0.0498, 0.0423, 0.1067, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0384, 0.0415, 0.0403, 0.0385, 0.0455, 0.0430, 0.0524, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 07:07:01,434 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9418, 4.2352, 3.3992, 2.2604, 2.8671, 2.6431, 4.6925, 3.7838], device='cuda:3'), covar=tensor([0.2672, 0.0758, 0.1483, 0.2751, 0.2783, 0.1820, 0.0349, 0.1235], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0264, 0.0293, 0.0292, 0.0283, 0.0237, 0.0279, 0.0316], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:07:09,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7702, 3.9222, 2.4594, 4.4328, 3.0548, 4.3873, 2.6045, 3.1614], device='cuda:3'), covar=tensor([0.0256, 0.0328, 0.1472, 0.0182, 0.0695, 0.0446, 0.1327, 0.0650], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0192, 0.0152, 0.0173, 0.0214, 0.0201, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:07:25,527 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 07:07:30,678 INFO [train.py:904] (3/8) Epoch 16, batch 1550, loss[loss=0.1829, simple_loss=0.2492, pruned_loss=0.0583, over 16500.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.253, pruned_loss=0.04323, over 3313478.80 frames. ], batch size: 75, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:07:53,742 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:08:40,378 INFO [train.py:904] (3/8) Epoch 16, batch 1600, loss[loss=0.2035, simple_loss=0.2832, pruned_loss=0.06189, over 16779.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2549, pruned_loss=0.04423, over 3315840.13 frames. ], batch size: 57, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:09:47,765 INFO [train.py:904] (3/8) Epoch 16, batch 1650, loss[loss=0.2006, simple_loss=0.2792, pruned_loss=0.06099, over 16581.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.257, pruned_loss=0.04517, over 3306707.80 frames. ], batch size: 75, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:10:09,059 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.549e+02 2.974e+02 3.703e+02 7.236e+02, threshold=5.949e+02, percent-clipped=5.0 2023-04-30 07:10:09,514 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2482, 5.6348, 5.3739, 5.4138, 5.0113, 5.0230, 4.9810, 5.7382], device='cuda:3'), covar=tensor([0.1324, 0.0975, 0.1150, 0.0774, 0.1062, 0.0761, 0.1138, 0.0931], device='cuda:3'), in_proj_covar=tensor([0.0632, 0.0778, 0.0634, 0.0560, 0.0491, 0.0497, 0.0650, 0.0594], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:10:16,015 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:18,362 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0281, 3.9601, 4.0009, 3.4527, 3.9871, 1.8935, 3.7992, 3.5038], device='cuda:3'), covar=tensor([0.0123, 0.0107, 0.0151, 0.0282, 0.0088, 0.2400, 0.0117, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0138, 0.0186, 0.0170, 0.0159, 0.0198, 0.0174, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:10:32,333 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:46,621 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 07:10:56,044 INFO [train.py:904] (3/8) Epoch 16, batch 1700, loss[loss=0.1679, simple_loss=0.2721, pruned_loss=0.03189, over 16737.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2587, pruned_loss=0.04532, over 3313021.37 frames. ], batch size: 62, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:11:38,410 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:40,841 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:42,775 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5299, 3.5768, 3.2611, 2.9911, 3.1651, 3.4805, 3.3238, 3.2654], device='cuda:3'), covar=tensor([0.0534, 0.0540, 0.0268, 0.0260, 0.0475, 0.0407, 0.1034, 0.0431], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0391, 0.0332, 0.0321, 0.0349, 0.0369, 0.0228, 0.0399], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:12:09,295 INFO [train.py:904] (3/8) Epoch 16, batch 1750, loss[loss=0.2113, simple_loss=0.2907, pruned_loss=0.06598, over 15568.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2599, pruned_loss=0.0453, over 3315782.35 frames. ], batch size: 190, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:12:21,055 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7448, 4.9486, 5.1497, 4.9432, 4.8360, 5.5766, 5.0645, 4.6951], device='cuda:3'), covar=tensor([0.1510, 0.2020, 0.2113, 0.2083, 0.3257, 0.1182, 0.1611, 0.2664], device='cuda:3'), in_proj_covar=tensor([0.0396, 0.0568, 0.0617, 0.0474, 0.0637, 0.0649, 0.0483, 0.0629], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:12:21,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6034, 3.4945, 3.8995, 2.7076, 3.5451, 3.9920, 3.7497, 2.3794], device='cuda:3'), covar=tensor([0.0433, 0.0300, 0.0045, 0.0347, 0.0089, 0.0075, 0.0075, 0.0392], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0076, 0.0075, 0.0132, 0.0089, 0.0098, 0.0087, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:12:33,136 INFO [optim.py:368] (3/8) 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 07:13:15,092 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:13:18,924 INFO [train.py:904] (3/8) Epoch 16, batch 1800, loss[loss=0.1549, simple_loss=0.2439, pruned_loss=0.0329, over 16836.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.261, pruned_loss=0.04542, over 3311605.90 frames. ], batch size: 42, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:13:52,581 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:13:54,189 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:03,048 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:06,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8287, 4.5564, 4.8399, 5.0295, 5.2048, 4.5733, 5.1628, 5.2032], device='cuda:3'), covar=tensor([0.1683, 0.1256, 0.1589, 0.0684, 0.0498, 0.0963, 0.0599, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0613, 0.0761, 0.0902, 0.0770, 0.0577, 0.0603, 0.0615, 0.0714], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:14:28,054 INFO [train.py:904] (3/8) Epoch 16, batch 1850, loss[loss=0.1725, simple_loss=0.2523, pruned_loss=0.04632, over 16591.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2616, pruned_loss=0.04539, over 3308361.48 frames. ], batch size: 89, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:14:38,680 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:14:39,029 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-30 07:14:50,220 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:15,204 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:15:21,231 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7582, 3.7630, 2.2087, 4.0770, 2.9013, 4.0343, 2.3947, 2.9955], device='cuda:3'), covar=tensor([0.0237, 0.0384, 0.1593, 0.0322, 0.0770, 0.0634, 0.1348, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0195, 0.0154, 0.0174, 0.0216, 0.0203, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:15:25,791 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 1900, loss[loss=0.1833, simple_loss=0.2547, pruned_loss=0.05597, over 16735.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2605, pruned_loss=0.04429, over 3321421.68 frames. ], batch size: 83, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:16:08,830 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-04-30 07:16:44,643 INFO [train.py:904] (3/8) Epoch 16, batch 1950, loss[loss=0.1598, simple_loss=0.253, pruned_loss=0.03334, over 17181.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2602, pruned_loss=0.04351, over 3325094.45 frames. ], batch size: 46, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:16:49,550 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7732, 4.4876, 4.4225, 4.9064, 5.0937, 4.5845, 5.0733, 5.1246], device='cuda:3'), covar=tensor([0.1475, 0.1215, 0.2513, 0.1069, 0.0797, 0.1149, 0.0988, 0.0835], device='cuda:3'), in_proj_covar=tensor([0.0616, 0.0763, 0.0906, 0.0772, 0.0579, 0.0603, 0.0615, 0.0714], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:16:50,702 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1606, 3.2558, 3.2986, 2.1526, 2.8857, 2.3411, 3.6548, 3.6431], device='cuda:3'), covar=tensor([0.0219, 0.0827, 0.0603, 0.1827, 0.0844, 0.1039, 0.0494, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0157, 0.0163, 0.0150, 0.0141, 0.0127, 0.0142, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:17:04,523 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.235e+02 2.551e+02 3.026e+02 6.742e+02, threshold=5.103e+02, percent-clipped=2.0 2023-04-30 07:17:10,550 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:17:51,473 INFO [train.py:904] (3/8) Epoch 16, batch 2000, loss[loss=0.1735, simple_loss=0.2608, pruned_loss=0.04307, over 16679.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2604, pruned_loss=0.04398, over 3306786.43 frames. ], batch size: 62, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:18:27,771 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:33,396 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:18:38,759 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:44,509 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:59,382 INFO [train.py:904] (3/8) Epoch 16, batch 2050, loss[loss=0.1641, simple_loss=0.2507, pruned_loss=0.03873, over 17223.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2608, pruned_loss=0.04444, over 3301152.37 frames. ], batch size: 43, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:19:19,609 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.241e+02 2.674e+02 3.116e+02 4.900e+02, threshold=5.347e+02, percent-clipped=0.0 2023-04-30 07:19:46,256 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:19:50,674 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 07:20:01,096 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:04,583 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6062, 2.5668, 2.2432, 2.3750, 2.8882, 2.7015, 3.2934, 3.1316], device='cuda:3'), covar=tensor([0.0118, 0.0390, 0.0429, 0.0431, 0.0258, 0.0366, 0.0249, 0.0244], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0225, 0.0218, 0.0218, 0.0228, 0.0229, 0.0232, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:20:06,433 INFO [train.py:904] (3/8) Epoch 16, batch 2100, loss[loss=0.1639, simple_loss=0.2442, pruned_loss=0.04185, over 15993.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2623, pruned_loss=0.04569, over 3308764.11 frames. ], batch size: 35, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:20:06,906 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:57,914 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1692, 2.0412, 2.1586, 3.8267, 2.1235, 2.4079, 2.1277, 2.2254], device='cuda:3'), covar=tensor([0.1254, 0.3458, 0.2737, 0.0552, 0.3769, 0.2377, 0.3522, 0.3108], device='cuda:3'), in_proj_covar=tensor([0.0383, 0.0420, 0.0352, 0.0323, 0.0424, 0.0483, 0.0389, 0.0493], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:21:09,569 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:21:14,695 INFO [train.py:904] (3/8) Epoch 16, batch 2150, loss[loss=0.1574, simple_loss=0.2431, pruned_loss=0.03584, over 16762.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2626, pruned_loss=0.04603, over 3316584.00 frames. ], batch size: 39, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:21:18,544 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:36,521 INFO [optim.py:368] (3/8) 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 07:21:57,842 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:22:07,247 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:22:25,078 INFO [train.py:904] (3/8) Epoch 16, batch 2200, loss[loss=0.1869, simple_loss=0.2816, pruned_loss=0.04608, over 17016.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2634, pruned_loss=0.04695, over 3313214.21 frames. ], batch size: 55, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:22:55,753 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:23:01,866 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6391, 2.6328, 2.2845, 2.4320, 2.9382, 2.7897, 3.4588, 3.2298], device='cuda:3'), covar=tensor([0.0119, 0.0325, 0.0414, 0.0376, 0.0215, 0.0279, 0.0187, 0.0215], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0225, 0.0218, 0.0218, 0.0227, 0.0228, 0.0231, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:23:16,645 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2386, 5.1894, 4.9827, 4.4625, 5.0186, 1.8763, 4.7915, 4.9994], device='cuda:3'), covar=tensor([0.0072, 0.0066, 0.0183, 0.0401, 0.0097, 0.2586, 0.0119, 0.0171], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0141, 0.0188, 0.0173, 0.0161, 0.0200, 0.0177, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:23:32,251 INFO [train.py:904] (3/8) Epoch 16, batch 2250, loss[loss=0.1628, simple_loss=0.2456, pruned_loss=0.03997, over 15865.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2642, pruned_loss=0.04789, over 3312393.79 frames. ], batch size: 35, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:23:45,842 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9023, 2.4919, 2.5128, 1.9331, 2.4622, 2.6828, 2.5375, 1.8731], device='cuda:3'), covar=tensor([0.0384, 0.0115, 0.0074, 0.0338, 0.0120, 0.0112, 0.0116, 0.0350], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0077, 0.0076, 0.0133, 0.0091, 0.0100, 0.0089, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:23:55,133 INFO [optim.py:368] (3/8) 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 07:24:18,218 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:24:19,932 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2984, 4.6031, 4.5580, 3.4215, 3.8257, 4.5742, 4.1032, 2.8356], device='cuda:3'), covar=tensor([0.0336, 0.0044, 0.0029, 0.0259, 0.0101, 0.0071, 0.0071, 0.0360], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0077, 0.0076, 0.0133, 0.0090, 0.0099, 0.0089, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:24:40,019 INFO [train.py:904] (3/8) Epoch 16, batch 2300, loss[loss=0.1798, simple_loss=0.2688, pruned_loss=0.0454, over 17043.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2643, pruned_loss=0.04759, over 3304031.95 frames. ], batch size: 55, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:25:07,234 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7968, 4.9442, 5.0793, 4.9227, 4.8983, 5.5416, 5.0637, 4.7131], device='cuda:3'), covar=tensor([0.1418, 0.2058, 0.2195, 0.2152, 0.2538, 0.1130, 0.1668, 0.2641], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0566, 0.0614, 0.0476, 0.0632, 0.0646, 0.0483, 0.0627], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:25:16,139 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:25:20,027 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:25:25,787 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9041, 4.1737, 3.9809, 4.0879, 3.6864, 3.7275, 3.8068, 4.1621], device='cuda:3'), covar=tensor([0.1271, 0.1090, 0.1230, 0.0747, 0.0971, 0.1782, 0.0976, 0.1111], device='cuda:3'), in_proj_covar=tensor([0.0637, 0.0786, 0.0641, 0.0566, 0.0495, 0.0503, 0.0655, 0.0603], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:25:33,322 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9726, 4.0896, 2.5585, 4.6764, 3.1418, 4.6328, 2.8922, 3.2657], device='cuda:3'), covar=tensor([0.0254, 0.0322, 0.1464, 0.0252, 0.0753, 0.0447, 0.1265, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0176, 0.0195, 0.0155, 0.0175, 0.0218, 0.0204, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:25:51,247 INFO [train.py:904] (3/8) Epoch 16, batch 2350, loss[loss=0.1986, simple_loss=0.2781, pruned_loss=0.05951, over 16433.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2652, pruned_loss=0.0487, over 3284440.96 frames. ], batch size: 146, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:26:11,866 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:26,639 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:45,774 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:52,318 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 2400, loss[loss=0.1973, simple_loss=0.271, pruned_loss=0.0618, over 16917.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2659, pruned_loss=0.04922, over 3295784.81 frames. ], batch size: 109, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:27:45,584 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 07:27:51,056 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:27:53,234 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1544, 5.6720, 5.8631, 5.5115, 5.5222, 6.1579, 5.6068, 5.2736], device='cuda:3'), covar=tensor([0.0878, 0.1794, 0.1907, 0.1917, 0.2630, 0.1001, 0.1351, 0.2094], device='cuda:3'), in_proj_covar=tensor([0.0396, 0.0563, 0.0611, 0.0473, 0.0631, 0.0644, 0.0480, 0.0626], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:27:55,461 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:28:06,455 INFO [train.py:904] (3/8) Epoch 16, batch 2450, loss[loss=0.1901, simple_loss=0.2642, pruned_loss=0.05794, over 16780.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2655, pruned_loss=0.0482, over 3308035.48 frames. ], batch size: 124, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:28:11,537 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:28:15,307 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0356, 2.4615, 2.0299, 2.2782, 2.8556, 2.5848, 2.9909, 3.0181], device='cuda:3'), covar=tensor([0.0165, 0.0379, 0.0477, 0.0403, 0.0208, 0.0334, 0.0245, 0.0240], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0226, 0.0219, 0.0220, 0.0229, 0.0230, 0.0234, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:28:20,257 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3220, 3.9625, 4.5651, 2.3136, 4.7221, 4.8201, 3.5673, 3.7514], device='cuda:3'), covar=tensor([0.0598, 0.0240, 0.0196, 0.1094, 0.0066, 0.0156, 0.0364, 0.0337], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0105, 0.0092, 0.0138, 0.0074, 0.0118, 0.0125, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 07:28:28,918 INFO [optim.py:368] (3/8) 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 07:28:31,858 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:50,215 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:59,848 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:16,066 INFO [train.py:904] (3/8) Epoch 16, batch 2500, loss[loss=0.1669, simple_loss=0.2572, pruned_loss=0.03832, over 17203.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2655, pruned_loss=0.04789, over 3306369.17 frames. ], batch size: 46, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:29:18,086 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:29:53,784 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:55,067 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:05,086 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:24,141 INFO [train.py:904] (3/8) Epoch 16, batch 2550, loss[loss=0.1641, simple_loss=0.2597, pruned_loss=0.03424, over 17106.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2656, pruned_loss=0.04757, over 3301491.31 frames. ], batch size: 47, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:30:47,016 INFO [optim.py:368] (3/8) 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 07:31:01,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9965, 3.8202, 4.4583, 2.3281, 4.5572, 4.6546, 3.4018, 3.4711], device='cuda:3'), covar=tensor([0.0667, 0.0224, 0.0159, 0.1052, 0.0058, 0.0160, 0.0371, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0138, 0.0074, 0.0118, 0.0126, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 07:31:02,615 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 2600, loss[loss=0.2042, simple_loss=0.2736, pruned_loss=0.06736, over 16916.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2649, pruned_loss=0.04692, over 3308084.06 frames. ], batch size: 109, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:31:46,928 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7303, 4.8202, 4.9950, 4.7844, 4.8603, 5.4821, 5.0157, 4.7187], device='cuda:3'), covar=tensor([0.1444, 0.2092, 0.2308, 0.2442, 0.2823, 0.1020, 0.1564, 0.2624], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0567, 0.0614, 0.0480, 0.0635, 0.0647, 0.0484, 0.0631], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:32:07,891 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:32:40,408 INFO [train.py:904] (3/8) Epoch 16, batch 2650, loss[loss=0.1576, simple_loss=0.2479, pruned_loss=0.03364, over 16818.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2648, pruned_loss=0.04627, over 3307924.63 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:33:01,395 INFO [optim.py:368] (3/8) 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 07:33:12,795 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:18,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8680, 3.0733, 3.0929, 2.0282, 2.7022, 2.2447, 3.3851, 3.4340], device='cuda:3'), covar=tensor([0.0266, 0.0882, 0.0654, 0.1788, 0.0863, 0.0958, 0.0573, 0.0850], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0156, 0.0162, 0.0148, 0.0139, 0.0126, 0.0140, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:33:35,793 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:43,510 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:47,342 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7128, 4.1267, 3.1056, 2.2661, 2.7826, 2.5836, 4.3731, 3.5934], device='cuda:3'), covar=tensor([0.2823, 0.0640, 0.1573, 0.2578, 0.2578, 0.1806, 0.0420, 0.1193], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0267, 0.0295, 0.0296, 0.0289, 0.0240, 0.0283, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:33:49,084 INFO [train.py:904] (3/8) Epoch 16, batch 2700, loss[loss=0.1613, simple_loss=0.2596, pruned_loss=0.03144, over 17144.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2643, pruned_loss=0.04538, over 3317449.74 frames. ], batch size: 46, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:34:21,635 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:30,143 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 07:34:32,699 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:41,401 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:45,856 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:34:47,944 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 2750, loss[loss=0.1832, simple_loss=0.2644, pruned_loss=0.05098, over 16845.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2651, pruned_loss=0.04533, over 3325371.32 frames. ], batch size: 42, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:35:18,260 INFO [optim.py:368] (3/8) 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 07:35:46,131 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:35:50,512 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:35:55,984 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3903, 1.5817, 2.0399, 2.2391, 2.3807, 2.3557, 1.6971, 2.4862], device='cuda:3'), covar=tensor([0.0174, 0.0444, 0.0288, 0.0250, 0.0258, 0.0237, 0.0450, 0.0135], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0186, 0.0172, 0.0176, 0.0186, 0.0142, 0.0185, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:36:04,754 INFO [train.py:904] (3/8) Epoch 16, batch 2800, loss[loss=0.1474, simple_loss=0.2381, pruned_loss=0.0283, over 17212.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2649, pruned_loss=0.04501, over 3323334.51 frames. ], batch size: 44, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:36:25,324 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 07:36:28,453 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:33,757 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8794, 1.9770, 2.3842, 2.8938, 2.6542, 3.2304, 2.1005, 3.1682], device='cuda:3'), covar=tensor([0.0212, 0.0412, 0.0310, 0.0245, 0.0277, 0.0157, 0.0433, 0.0131], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0187, 0.0173, 0.0176, 0.0187, 0.0143, 0.0186, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:36:38,305 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:48,125 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 07:37:08,323 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8254, 3.0576, 3.1544, 2.0744, 2.7649, 2.2510, 3.3997, 3.3739], device='cuda:3'), covar=tensor([0.0225, 0.0857, 0.0557, 0.1698, 0.0839, 0.0958, 0.0509, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0155, 0.0161, 0.0147, 0.0139, 0.0125, 0.0140, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:37:15,724 INFO [train.py:904] (3/8) Epoch 16, batch 2850, loss[loss=0.1562, simple_loss=0.2546, pruned_loss=0.02892, over 17084.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.0445, over 3325897.54 frames. ], batch size: 50, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:37:36,420 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 2.238e+02 2.628e+02 3.275e+02 9.061e+02, threshold=5.256e+02, percent-clipped=3.0 2023-04-30 07:37:54,361 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:37:54,486 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:38:23,946 INFO [train.py:904] (3/8) Epoch 16, batch 2900, loss[loss=0.1664, simple_loss=0.2468, pruned_loss=0.04301, over 16850.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2626, pruned_loss=0.04393, over 3337531.65 frames. ], batch size: 96, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:00,496 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:39:07,078 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2005, 5.0940, 5.5622, 5.4977, 5.6485, 5.2263, 5.1774, 5.0877], device='cuda:3'), covar=tensor([0.0338, 0.0615, 0.0444, 0.0533, 0.0556, 0.0439, 0.1034, 0.0516], device='cuda:3'), in_proj_covar=tensor([0.0389, 0.0420, 0.0409, 0.0386, 0.0459, 0.0435, 0.0531, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 07:39:27,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9115, 3.1223, 2.7404, 4.6629, 3.8343, 4.1954, 1.7181, 2.9492], device='cuda:3'), covar=tensor([0.1277, 0.0641, 0.1042, 0.0231, 0.0276, 0.0430, 0.1526, 0.0858], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0166, 0.0187, 0.0173, 0.0200, 0.0213, 0.0189, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:39:33,218 INFO [train.py:904] (3/8) Epoch 16, batch 2950, loss[loss=0.2055, simple_loss=0.2721, pruned_loss=0.06939, over 16879.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2618, pruned_loss=0.04483, over 3339184.14 frames. ], batch size: 116, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:54,115 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 16, batch 3000, loss[loss=0.2622, simple_loss=0.3305, pruned_loss=0.09699, over 12068.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2628, pruned_loss=0.04593, over 3326719.19 frames. ], batch size: 246, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:40:40,759 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 07:40:49,854 INFO [train.py:938] (3/8) Epoch 16, validation: loss=0.1368, simple_loss=0.2429, pruned_loss=0.01541, over 944034.00 frames. 2023-04-30 07:40:49,855 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 07:41:09,006 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0638, 5.0607, 4.8495, 4.3435, 4.9774, 1.9469, 4.7232, 4.7293], device='cuda:3'), covar=tensor([0.0083, 0.0080, 0.0187, 0.0325, 0.0078, 0.2545, 0.0115, 0.0182], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0142, 0.0191, 0.0175, 0.0162, 0.0202, 0.0179, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:41:34,709 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:41:48,436 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 07:41:59,655 INFO [train.py:904] (3/8) Epoch 16, batch 3050, loss[loss=0.1863, simple_loss=0.2634, pruned_loss=0.05462, over 15985.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2627, pruned_loss=0.0456, over 3327507.08 frames. ], batch size: 35, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:42:21,040 INFO [optim.py:368] (3/8) 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 07:42:31,181 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4673, 3.3609, 3.4806, 3.5614, 3.5944, 3.3385, 3.5485, 3.6476], device='cuda:3'), covar=tensor([0.1086, 0.0924, 0.1007, 0.0600, 0.0667, 0.2475, 0.1204, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0620, 0.0770, 0.0906, 0.0783, 0.0582, 0.0612, 0.0616, 0.0725], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:42:37,666 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 07:42:42,355 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:42,375 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:45,124 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 07:43:10,100 INFO [train.py:904] (3/8) Epoch 16, batch 3100, loss[loss=0.1677, simple_loss=0.2493, pruned_loss=0.043, over 16281.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2625, pruned_loss=0.0459, over 3309346.82 frames. ], batch size: 36, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:43:28,154 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9441, 5.2816, 5.5321, 5.2924, 5.2668, 5.9130, 5.4886, 5.1659], device='cuda:3'), covar=tensor([0.1067, 0.1938, 0.2230, 0.1841, 0.2518, 0.0972, 0.1245, 0.2203], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0572, 0.0617, 0.0484, 0.0641, 0.0647, 0.0485, 0.0638], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:43:42,802 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:15,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8265, 3.1213, 2.9968, 5.0861, 4.2132, 4.6061, 1.5570, 3.3326], device='cuda:3'), covar=tensor([0.1298, 0.0692, 0.0977, 0.0153, 0.0209, 0.0335, 0.1597, 0.0692], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0167, 0.0188, 0.0174, 0.0201, 0.0214, 0.0190, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:44:17,694 INFO [train.py:904] (3/8) Epoch 16, batch 3150, loss[loss=0.1909, simple_loss=0.2588, pruned_loss=0.06151, over 16868.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2615, pruned_loss=0.04552, over 3319850.48 frames. ], batch size: 109, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:44:39,887 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.373e+02 2.704e+02 3.278e+02 4.749e+02, threshold=5.408e+02, percent-clipped=0.0 2023-04-30 07:44:47,892 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:49,093 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:45:27,254 INFO [train.py:904] (3/8) Epoch 16, batch 3200, loss[loss=0.1718, simple_loss=0.2487, pruned_loss=0.04746, over 16479.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.26, pruned_loss=0.04484, over 3322510.90 frames. ], batch size: 68, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:45:34,239 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:45:46,290 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9040, 5.3159, 5.0484, 5.0671, 4.8325, 4.7515, 4.7098, 5.4287], device='cuda:3'), covar=tensor([0.1080, 0.0818, 0.0977, 0.0858, 0.0797, 0.1010, 0.1097, 0.0860], device='cuda:3'), in_proj_covar=tensor([0.0648, 0.0802, 0.0650, 0.0578, 0.0503, 0.0511, 0.0667, 0.0615], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:45:50,159 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 07:46:22,008 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 07:46:36,137 INFO [train.py:904] (3/8) Epoch 16, batch 3250, loss[loss=0.1661, simple_loss=0.2553, pruned_loss=0.03843, over 15952.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2602, pruned_loss=0.04481, over 3327337.21 frames. ], batch size: 35, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:46:58,461 INFO [optim.py:368] (3/8) 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 07:46:58,962 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 3300, loss[loss=0.1635, simple_loss=0.2553, pruned_loss=0.03586, over 17113.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2614, pruned_loss=0.04503, over 3325386.19 frames. ], batch size: 47, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:48:56,952 INFO [train.py:904] (3/8) Epoch 16, batch 3350, loss[loss=0.1757, simple_loss=0.2664, pruned_loss=0.04249, over 16672.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.262, pruned_loss=0.04517, over 3319524.40 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:49:01,302 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:49:39,837 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:08,446 INFO [train.py:904] (3/8) Epoch 16, batch 3400, loss[loss=0.1703, simple_loss=0.2642, pruned_loss=0.03821, over 16700.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2616, pruned_loss=0.04482, over 3329670.13 frames. ], batch size: 57, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:50:10,295 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-04-30 07:50:16,675 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:28,029 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:49,343 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:53,268 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2270, 4.1695, 4.3445, 4.1166, 4.1573, 4.7849, 4.3441, 3.9799], device='cuda:3'), covar=tensor([0.2029, 0.2456, 0.2298, 0.2655, 0.3156, 0.1389, 0.1695, 0.2955], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0572, 0.0617, 0.0481, 0.0642, 0.0647, 0.0488, 0.0638], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:51:19,269 INFO [train.py:904] (3/8) Epoch 16, batch 3450, loss[loss=0.2124, simple_loss=0.283, pruned_loss=0.0709, over 11987.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.261, pruned_loss=0.04462, over 3323790.49 frames. ], batch size: 246, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:51:38,150 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 07:51:41,373 INFO [optim.py:368] (3/8) 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 07:51:41,834 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:51:52,549 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:52:28,777 INFO [train.py:904] (3/8) Epoch 16, batch 3500, loss[loss=0.1794, simple_loss=0.2789, pruned_loss=0.04, over 17059.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2601, pruned_loss=0.04428, over 3311724.51 frames. ], batch size: 55, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:52:58,253 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:14,777 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:40,573 INFO [train.py:904] (3/8) Epoch 16, batch 3550, loss[loss=0.1933, simple_loss=0.266, pruned_loss=0.0603, over 11746.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2583, pruned_loss=0.04368, over 3307960.30 frames. ], batch size: 247, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:53:56,148 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:54:04,791 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.310e+02 2.739e+02 3.224e+02 5.856e+02, threshold=5.477e+02, percent-clipped=2.0 2023-04-30 07:54:15,886 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1871, 3.3847, 3.3660, 2.1662, 2.9311, 2.4059, 3.6837, 3.6251], device='cuda:3'), covar=tensor([0.0212, 0.0726, 0.0560, 0.1689, 0.0771, 0.0930, 0.0470, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0159, 0.0164, 0.0150, 0.0141, 0.0127, 0.0143, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:54:42,485 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:54:52,021 INFO [train.py:904] (3/8) Epoch 16, batch 3600, loss[loss=0.2072, simple_loss=0.2801, pruned_loss=0.06717, over 16267.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2583, pruned_loss=0.04383, over 3309868.53 frames. ], batch size: 165, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:03,880 INFO [train.py:904] (3/8) Epoch 16, batch 3650, loss[loss=0.18, simple_loss=0.2464, pruned_loss=0.05676, over 16736.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2566, pruned_loss=0.0443, over 3310493.70 frames. ], batch size: 83, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:28,564 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.258e+02 2.651e+02 3.350e+02 6.145e+02, threshold=5.302e+02, percent-clipped=1.0 2023-04-30 07:56:29,892 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 07:57:17,741 INFO [train.py:904] (3/8) Epoch 16, batch 3700, loss[loss=0.1862, simple_loss=0.257, pruned_loss=0.05775, over 16413.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2555, pruned_loss=0.04586, over 3272214.46 frames. ], batch size: 75, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:57:32,538 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:57:38,337 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9426, 4.8851, 4.8342, 4.5816, 4.5669, 4.9106, 4.7329, 4.6688], device='cuda:3'), covar=tensor([0.0569, 0.0645, 0.0265, 0.0253, 0.0751, 0.0480, 0.0406, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0397, 0.0337, 0.0326, 0.0353, 0.0377, 0.0230, 0.0406], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 07:58:29,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1299, 4.2222, 4.5372, 4.5400, 4.5680, 4.2196, 4.2778, 4.0776], device='cuda:3'), covar=tensor([0.0378, 0.0646, 0.0453, 0.0418, 0.0425, 0.0477, 0.0807, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0424, 0.0411, 0.0389, 0.0466, 0.0439, 0.0533, 0.0347], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 07:58:36,927 INFO [train.py:904] (3/8) Epoch 16, batch 3750, loss[loss=0.2135, simple_loss=0.28, pruned_loss=0.07347, over 16246.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2565, pruned_loss=0.04744, over 3247036.66 frames. ], batch size: 165, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:58:37,288 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3385, 4.1807, 4.3120, 4.5047, 4.6039, 4.2300, 4.4801, 4.6202], device='cuda:3'), covar=tensor([0.1625, 0.1355, 0.1723, 0.0934, 0.0815, 0.1251, 0.2244, 0.1093], device='cuda:3'), in_proj_covar=tensor([0.0620, 0.0770, 0.0907, 0.0781, 0.0582, 0.0612, 0.0615, 0.0724], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 07:58:47,505 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0014, 3.0442, 2.6503, 4.5267, 3.7148, 4.2849, 1.8598, 3.1291], device='cuda:3'), covar=tensor([0.1235, 0.0625, 0.1084, 0.0151, 0.0234, 0.0354, 0.1381, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0167, 0.0187, 0.0174, 0.0201, 0.0213, 0.0189, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 07:58:53,072 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:59:02,685 INFO [optim.py:368] (3/8) 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 07:59:51,245 INFO [train.py:904] (3/8) Epoch 16, batch 3800, loss[loss=0.177, simple_loss=0.2567, pruned_loss=0.04862, over 16340.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2573, pruned_loss=0.04884, over 3250027.45 frames. ], batch size: 165, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 08:00:17,803 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:03,105 INFO [train.py:904] (3/8) Epoch 16, batch 3850, loss[loss=0.1766, simple_loss=0.2466, pruned_loss=0.05329, over 16905.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2577, pruned_loss=0.04936, over 3257168.10 frames. ], batch size: 90, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:01:20,117 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:01:47,443 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:01:52,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.80 vs. limit=5.0 2023-04-30 08:01:59,098 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:02:16,241 INFO [train.py:904] (3/8) Epoch 16, batch 3900, loss[loss=0.171, simple_loss=0.2549, pruned_loss=0.04351, over 16225.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2572, pruned_loss=0.04972, over 3260779.27 frames. ], batch size: 35, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:02:25,034 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-30 08:02:29,442 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:02:34,219 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 08:02:36,786 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8442, 3.0621, 3.1475, 2.1544, 2.7999, 2.3012, 3.3853, 3.4190], device='cuda:3'), covar=tensor([0.0256, 0.0843, 0.0601, 0.1685, 0.0842, 0.0928, 0.0536, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0159, 0.0163, 0.0149, 0.0141, 0.0127, 0.0141, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:03:26,069 INFO [train.py:904] (3/8) Epoch 16, batch 3950, loss[loss=0.1884, simple_loss=0.2692, pruned_loss=0.05384, over 16556.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2573, pruned_loss=0.05034, over 3258561.20 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:03:47,453 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:50,339 INFO [optim.py:368] (3/8) 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 08:04:11,355 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4442, 5.7311, 5.4779, 5.5554, 5.2160, 5.0808, 5.2100, 5.8853], device='cuda:3'), covar=tensor([0.1171, 0.0788, 0.1050, 0.0772, 0.0744, 0.0638, 0.1022, 0.0811], device='cuda:3'), in_proj_covar=tensor([0.0636, 0.0788, 0.0642, 0.0571, 0.0494, 0.0505, 0.0655, 0.0603], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:04:35,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 08:04:37,101 INFO [train.py:904] (3/8) Epoch 16, batch 4000, loss[loss=0.1881, simple_loss=0.266, pruned_loss=0.0551, over 12524.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2569, pruned_loss=0.05041, over 3263571.77 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:04:38,788 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5156, 5.5031, 5.3119, 4.6498, 5.4663, 1.7320, 5.1666, 5.0384], device='cuda:3'), covar=tensor([0.0045, 0.0041, 0.0118, 0.0291, 0.0046, 0.2954, 0.0091, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0142, 0.0190, 0.0175, 0.0163, 0.0201, 0.0179, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:04:49,951 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:12,203 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:31,037 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:49,241 INFO [train.py:904] (3/8) Epoch 16, batch 4050, loss[loss=0.1688, simple_loss=0.2479, pruned_loss=0.04478, over 16496.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2574, pruned_loss=0.04987, over 3263564.40 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:05:59,641 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:04,402 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:14,225 INFO [optim.py:368] (3/8) 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 08:07:01,362 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:01,988 INFO [train.py:904] (3/8) Epoch 16, batch 4100, loss[loss=0.191, simple_loss=0.272, pruned_loss=0.05501, over 16633.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2595, pruned_loss=0.04999, over 3244756.79 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:07:13,126 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.47 vs. limit=5.0 2023-04-30 08:07:15,153 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:32,309 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8041, 2.8257, 2.2510, 2.6674, 3.1280, 2.8547, 3.4786, 3.4192], device='cuda:3'), covar=tensor([0.0060, 0.0316, 0.0420, 0.0345, 0.0196, 0.0301, 0.0163, 0.0181], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0220, 0.0212, 0.0213, 0.0222, 0.0223, 0.0228, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:07:52,024 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6556, 4.7302, 4.8895, 4.7158, 4.7701, 5.3339, 4.7869, 4.4566], device='cuda:3'), covar=tensor([0.1129, 0.1666, 0.1808, 0.1958, 0.2350, 0.0887, 0.1348, 0.2465], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0561, 0.0602, 0.0473, 0.0625, 0.0636, 0.0475, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:08:15,974 INFO [train.py:904] (3/8) Epoch 16, batch 4150, loss[loss=0.2074, simple_loss=0.2864, pruned_loss=0.06418, over 17062.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2666, pruned_loss=0.05244, over 3223002.32 frames. ], batch size: 53, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:08:40,340 INFO [optim.py:368] (3/8) 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 08:08:52,136 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:09:13,262 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 4200, loss[loss=0.2303, simple_loss=0.3078, pruned_loss=0.07639, over 11640.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.273, pruned_loss=0.0532, over 3203069.12 frames. ], batch size: 248, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:10:17,464 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8284, 2.6270, 2.6323, 1.9450, 2.5015, 2.6429, 2.5628, 1.8714], device='cuda:3'), covar=tensor([0.0412, 0.0080, 0.0076, 0.0334, 0.0109, 0.0117, 0.0109, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0076, 0.0076, 0.0130, 0.0089, 0.0099, 0.0088, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:10:24,399 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:10:43,947 INFO [train.py:904] (3/8) Epoch 16, batch 4250, loss[loss=0.1801, simple_loss=0.2738, pruned_loss=0.04321, over 16553.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2762, pruned_loss=0.05367, over 3170521.86 frames. ], batch size: 68, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:11:09,162 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.212e+02 2.674e+02 3.089e+02 6.979e+02, threshold=5.347e+02, percent-clipped=2.0 2023-04-30 08:11:56,444 INFO [train.py:904] (3/8) Epoch 16, batch 4300, loss[loss=0.2372, simple_loss=0.329, pruned_loss=0.07275, over 16833.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.278, pruned_loss=0.05296, over 3171693.77 frames. ], batch size: 116, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:12:02,041 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6141, 2.5801, 2.3906, 4.1251, 3.0212, 3.9280, 1.5423, 2.7791], device='cuda:3'), covar=tensor([0.1410, 0.0836, 0.1271, 0.0218, 0.0328, 0.0400, 0.1652, 0.0912], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0166, 0.0184, 0.0171, 0.0200, 0.0211, 0.0188, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-30 08:12:27,167 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:12:56,094 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 08:13:09,398 INFO [train.py:904] (3/8) Epoch 16, batch 4350, loss[loss=0.2077, simple_loss=0.3003, pruned_loss=0.05756, over 16328.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2816, pruned_loss=0.05389, over 3171543.50 frames. ], batch size: 165, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:13:34,596 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.270e+02 2.623e+02 3.147e+02 7.539e+02, threshold=5.245e+02, percent-clipped=2.0 2023-04-30 08:13:54,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4132, 3.2782, 2.5865, 2.0377, 2.2577, 2.1535, 3.4509, 3.0585], device='cuda:3'), covar=tensor([0.2972, 0.0775, 0.1776, 0.2512, 0.2477, 0.2051, 0.0570, 0.1183], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0265, 0.0295, 0.0295, 0.0292, 0.0240, 0.0282, 0.0320], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:14:14,560 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 4400, loss[loss=0.2164, simple_loss=0.3037, pruned_loss=0.06454, over 17037.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2834, pruned_loss=0.05463, over 3178926.51 frames. ], batch size: 55, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:14:29,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3068, 2.3596, 2.3412, 4.1702, 2.1494, 2.8263, 2.4179, 2.5419], device='cuda:3'), covar=tensor([0.1137, 0.3091, 0.2405, 0.0402, 0.3829, 0.2030, 0.2902, 0.3053], device='cuda:3'), in_proj_covar=tensor([0.0387, 0.0429, 0.0355, 0.0327, 0.0429, 0.0494, 0.0395, 0.0501], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:14:57,956 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 08:15:30,371 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.41 vs. limit=5.0 2023-04-30 08:15:32,099 INFO [train.py:904] (3/8) Epoch 16, batch 4450, loss[loss=0.2326, simple_loss=0.2995, pruned_loss=0.0828, over 11784.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2872, pruned_loss=0.05633, over 3173332.11 frames. ], batch size: 247, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:57,565 INFO [optim.py:368] (3/8) 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 08:15:58,126 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2563, 3.4606, 3.6802, 2.1116, 3.0299, 2.3144, 3.5281, 3.5955], device='cuda:3'), covar=tensor([0.0212, 0.0718, 0.0476, 0.1933, 0.0794, 0.0976, 0.0563, 0.0838], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0157, 0.0162, 0.0149, 0.0140, 0.0126, 0.0141, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:16:09,711 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:16:41,417 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:16:45,639 INFO [train.py:904] (3/8) Epoch 16, batch 4500, loss[loss=0.1828, simple_loss=0.2666, pruned_loss=0.04947, over 16540.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2878, pruned_loss=0.05725, over 3177489.57 frames. ], batch size: 68, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:17:16,184 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:17:43,388 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3964, 4.2609, 4.4482, 4.5968, 4.7208, 4.2921, 4.6711, 4.7522], device='cuda:3'), covar=tensor([0.1416, 0.1140, 0.1221, 0.0549, 0.0399, 0.0983, 0.0579, 0.0499], device='cuda:3'), in_proj_covar=tensor([0.0588, 0.0728, 0.0856, 0.0735, 0.0550, 0.0581, 0.0582, 0.0679], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:17:56,547 INFO [train.py:904] (3/8) Epoch 16, batch 4550, loss[loss=0.2126, simple_loss=0.2983, pruned_loss=0.0635, over 16666.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2882, pruned_loss=0.05781, over 3180264.85 frames. ], batch size: 134, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:18:07,965 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:18:19,417 INFO [optim.py:368] (3/8) 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 08:19:03,597 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2646, 5.8424, 6.0558, 5.6445, 5.7810, 6.3116, 5.7866, 5.5469], device='cuda:3'), covar=tensor([0.0792, 0.1609, 0.1441, 0.1926, 0.2264, 0.0819, 0.1063, 0.2076], device='cuda:3'), in_proj_covar=tensor([0.0384, 0.0549, 0.0592, 0.0463, 0.0616, 0.0628, 0.0466, 0.0618], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:19:06,328 INFO [train.py:904] (3/8) Epoch 16, batch 4600, loss[loss=0.1813, simple_loss=0.2731, pruned_loss=0.0447, over 16862.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2892, pruned_loss=0.05817, over 3188057.77 frames. ], batch size: 83, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:19:13,390 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-30 08:19:35,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2651, 3.2904, 1.7598, 3.6199, 2.4375, 3.6074, 2.0441, 2.5889], device='cuda:3'), covar=tensor([0.0267, 0.0359, 0.1872, 0.0146, 0.0871, 0.0421, 0.1536, 0.0777], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0170, 0.0189, 0.0147, 0.0170, 0.0210, 0.0198, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:19:36,295 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:19:55,722 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156885.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:19,277 INFO [train.py:904] (3/8) Epoch 16, batch 4650, loss[loss=0.1783, simple_loss=0.2658, pruned_loss=0.04544, over 17029.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2878, pruned_loss=0.05775, over 3196575.74 frames. ], batch size: 50, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:20:45,036 INFO [optim.py:368] (3/8) 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 08:20:46,966 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:25,753 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:25,839 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:33,093 INFO [train.py:904] (3/8) Epoch 16, batch 4700, loss[loss=0.1674, simple_loss=0.2575, pruned_loss=0.03869, over 16733.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.285, pruned_loss=0.05658, over 3186341.44 frames. ], batch size: 89, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:21:41,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6168, 2.3834, 2.2283, 3.4465, 2.2663, 3.5454, 1.4161, 2.5658], device='cuda:3'), covar=tensor([0.1530, 0.0831, 0.1334, 0.0185, 0.0190, 0.0384, 0.1922, 0.0895], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0168, 0.0187, 0.0171, 0.0201, 0.0212, 0.0190, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:22:36,260 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 4750, loss[loss=0.1603, simple_loss=0.2502, pruned_loss=0.03517, over 16451.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2806, pruned_loss=0.05437, over 3200268.32 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:59,301 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:22:59,659 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 08:23:11,419 INFO [optim.py:368] (3/8) 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 08:23:12,866 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 08:23:59,602 INFO [train.py:904] (3/8) Epoch 16, batch 4800, loss[loss=0.1758, simple_loss=0.2656, pruned_loss=0.04297, over 16470.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2771, pruned_loss=0.05256, over 3200233.25 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:24:19,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6742, 2.6949, 1.8323, 2.7911, 2.1264, 2.8075, 2.0394, 2.3346], device='cuda:3'), covar=tensor([0.0262, 0.0326, 0.1216, 0.0165, 0.0631, 0.0407, 0.1148, 0.0583], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0170, 0.0189, 0.0147, 0.0170, 0.0209, 0.0198, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:24:28,459 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157071.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:14,788 INFO [train.py:904] (3/8) Epoch 16, batch 4850, loss[loss=0.1782, simple_loss=0.2726, pruned_loss=0.04183, over 16478.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2778, pruned_loss=0.05146, over 3190398.31 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:25:15,197 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4674, 4.5484, 4.3462, 4.0582, 3.9842, 4.4480, 4.2133, 4.1913], device='cuda:3'), covar=tensor([0.0514, 0.0350, 0.0264, 0.0255, 0.0841, 0.0408, 0.0478, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0265, 0.0369, 0.0314, 0.0303, 0.0330, 0.0350, 0.0215, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:25:20,718 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:41,919 INFO [optim.py:368] (3/8) 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 08:26:16,250 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:26:30,905 INFO [train.py:904] (3/8) Epoch 16, batch 4900, loss[loss=0.183, simple_loss=0.2719, pruned_loss=0.04703, over 16754.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2766, pruned_loss=0.04983, over 3205348.43 frames. ], batch size: 124, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:45,016 INFO [train.py:904] (3/8) Epoch 16, batch 4950, loss[loss=0.2027, simple_loss=0.2928, pruned_loss=0.05635, over 15378.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2767, pruned_loss=0.04972, over 3205894.53 frames. ], batch size: 190, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:46,733 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:08,806 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.143e+02 2.414e+02 2.690e+02 6.197e+02, threshold=4.827e+02, percent-clipped=2.0 2023-04-30 08:28:41,102 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:57,707 INFO [train.py:904] (3/8) Epoch 16, batch 5000, loss[loss=0.1713, simple_loss=0.2665, pruned_loss=0.03805, over 16850.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2785, pruned_loss=0.04999, over 3209423.44 frames. ], batch size: 102, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:29:22,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0174, 4.0960, 3.8851, 3.6187, 3.5915, 4.0212, 3.6907, 3.7615], device='cuda:3'), covar=tensor([0.0563, 0.0508, 0.0262, 0.0250, 0.0754, 0.0404, 0.0941, 0.0478], device='cuda:3'), in_proj_covar=tensor([0.0265, 0.0370, 0.0313, 0.0303, 0.0330, 0.0351, 0.0213, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:30:07,690 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1412, 1.5157, 1.8503, 2.0316, 2.2079, 2.3582, 1.6927, 2.3078], device='cuda:3'), covar=tensor([0.0207, 0.0431, 0.0246, 0.0340, 0.0260, 0.0171, 0.0422, 0.0112], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0185, 0.0172, 0.0176, 0.0186, 0.0141, 0.0185, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:30:10,078 INFO [train.py:904] (3/8) Epoch 16, batch 5050, loss[loss=0.2064, simple_loss=0.2894, pruned_loss=0.06171, over 11894.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2792, pruned_loss=0.05008, over 3189700.02 frames. ], batch size: 247, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:25,170 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.27 vs. limit=5.0 2023-04-30 08:30:33,953 INFO [optim.py:368] (3/8) 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 08:31:03,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2357, 4.2611, 4.4545, 4.1707, 4.2649, 4.8225, 4.3997, 4.0808], device='cuda:3'), covar=tensor([0.1556, 0.1766, 0.1556, 0.2294, 0.2603, 0.1020, 0.1294, 0.2645], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0537, 0.0578, 0.0452, 0.0607, 0.0616, 0.0458, 0.0607], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:31:22,548 INFO [train.py:904] (3/8) Epoch 16, batch 5100, loss[loss=0.1921, simple_loss=0.2844, pruned_loss=0.04988, over 16694.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2778, pruned_loss=0.04928, over 3194386.09 frames. ], batch size: 124, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:31:26,093 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5171, 4.4475, 4.4411, 3.6559, 4.4275, 1.9154, 4.1312, 4.1581], device='cuda:3'), covar=tensor([0.0087, 0.0089, 0.0133, 0.0414, 0.0088, 0.2373, 0.0119, 0.0208], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0136, 0.0182, 0.0170, 0.0155, 0.0193, 0.0169, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:31:29,231 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:31:44,276 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:07,466 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:22,327 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9068, 1.9908, 2.3197, 3.1699, 2.0921, 2.1797, 2.2342, 2.1320], device='cuda:3'), covar=tensor([0.1137, 0.3339, 0.2295, 0.0635, 0.4018, 0.2509, 0.2990, 0.3258], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0423, 0.0351, 0.0321, 0.0425, 0.0487, 0.0390, 0.0493], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:32:38,193 INFO [train.py:904] (3/8) Epoch 16, batch 5150, loss[loss=0.1928, simple_loss=0.2834, pruned_loss=0.05109, over 12581.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2775, pruned_loss=0.04852, over 3191373.90 frames. ], batch size: 248, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:32:42,195 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:32:44,018 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:02,220 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:04,659 INFO [optim.py:368] (3/8) 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 08:33:40,829 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:41,058 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-04-30 08:33:53,094 INFO [train.py:904] (3/8) Epoch 16, batch 5200, loss[loss=0.1896, simple_loss=0.2695, pruned_loss=0.05484, over 16483.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2757, pruned_loss=0.04791, over 3187870.08 frames. ], batch size: 75, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:33:54,780 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:34:12,646 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:35:02,000 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:35:08,182 INFO [train.py:904] (3/8) Epoch 16, batch 5250, loss[loss=0.185, simple_loss=0.2646, pruned_loss=0.0527, over 16740.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2732, pruned_loss=0.04787, over 3191951.17 frames. ], batch size: 57, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:35:13,314 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:35:32,952 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.109e+02 2.306e+02 2.689e+02 3.834e+02, threshold=4.611e+02, percent-clipped=0.0 2023-04-30 08:36:06,461 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 5300, loss[loss=0.1746, simple_loss=0.2584, pruned_loss=0.04545, over 16888.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2703, pruned_loss=0.04703, over 3172675.38 frames. ], batch size: 109, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:36:30,158 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1934, 3.4490, 3.6152, 1.9617, 3.1585, 2.4144, 3.5584, 3.7538], device='cuda:3'), covar=tensor([0.0269, 0.0717, 0.0509, 0.1978, 0.0769, 0.0886, 0.0602, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0154, 0.0161, 0.0148, 0.0139, 0.0125, 0.0139, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:36:45,053 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:36:46,056 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:17,186 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:35,557 INFO [train.py:904] (3/8) Epoch 16, batch 5350, loss[loss=0.1802, simple_loss=0.2701, pruned_loss=0.04515, over 16483.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2687, pruned_loss=0.04642, over 3184494.08 frames. ], batch size: 75, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:00,210 INFO [optim.py:368] (3/8) 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 08:38:16,647 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:38:49,043 INFO [train.py:904] (3/8) Epoch 16, batch 5400, loss[loss=0.189, simple_loss=0.2856, pruned_loss=0.04626, over 16302.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2716, pruned_loss=0.04733, over 3206685.88 frames. ], batch size: 165, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:39:09,555 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:39:23,937 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6343, 3.7522, 1.8579, 4.4177, 2.6974, 4.2072, 2.2544, 2.8260], device='cuda:3'), covar=tensor([0.0268, 0.0354, 0.2062, 0.0124, 0.0862, 0.0470, 0.1590, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0147, 0.0172, 0.0210, 0.0200, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:40:04,938 INFO [train.py:904] (3/8) Epoch 16, batch 5450, loss[loss=0.2269, simple_loss=0.309, pruned_loss=0.07244, over 16673.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2744, pruned_loss=0.04858, over 3201924.61 frames. ], batch size: 134, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:40:20,460 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:23,493 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:41:00,215 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:41:09,152 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6374, 3.5219, 4.0966, 1.8872, 4.2462, 4.2387, 2.9407, 3.1225], device='cuda:3'), covar=tensor([0.0708, 0.0257, 0.0153, 0.1134, 0.0049, 0.0116, 0.0436, 0.0411], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0104, 0.0090, 0.0136, 0.0072, 0.0116, 0.0123, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 08:41:20,975 INFO [train.py:904] (3/8) Epoch 16, batch 5500, loss[loss=0.2454, simple_loss=0.3407, pruned_loss=0.07506, over 17130.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2821, pruned_loss=0.05356, over 3161860.17 frames. ], batch size: 47, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:41:33,656 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:41:45,346 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8206, 1.3649, 1.7692, 1.6505, 1.7860, 1.9005, 1.5583, 1.8701], device='cuda:3'), covar=tensor([0.0218, 0.0323, 0.0165, 0.0246, 0.0217, 0.0152, 0.0332, 0.0107], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0187, 0.0172, 0.0176, 0.0187, 0.0142, 0.0185, 0.0136], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:42:30,889 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:42:36,604 INFO [train.py:904] (3/8) Epoch 16, batch 5550, loss[loss=0.2074, simple_loss=0.299, pruned_loss=0.05795, over 16796.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2892, pruned_loss=0.05864, over 3133128.97 frames. ], batch size: 102, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:04,447 INFO [optim.py:368] (3/8) 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 08:43:13,984 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7698, 4.6857, 4.5438, 3.2350, 3.9574, 4.6086, 4.0171, 2.6306], device='cuda:3'), covar=tensor([0.0443, 0.0023, 0.0031, 0.0306, 0.0082, 0.0070, 0.0071, 0.0335], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0074, 0.0075, 0.0128, 0.0088, 0.0098, 0.0086, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:43:47,867 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:43:57,047 INFO [train.py:904] (3/8) Epoch 16, batch 5600, loss[loss=0.2139, simple_loss=0.3011, pruned_loss=0.06337, over 16688.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2949, pruned_loss=0.06334, over 3091655.64 frames. ], batch size: 89, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:59,550 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:44:14,256 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:45:15,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0273, 4.0124, 3.9565, 3.2507, 3.9882, 1.8280, 3.8087, 3.6147], device='cuda:3'), covar=tensor([0.0100, 0.0085, 0.0156, 0.0286, 0.0081, 0.2499, 0.0121, 0.0211], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0137, 0.0184, 0.0172, 0.0156, 0.0194, 0.0171, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 08:45:21,486 INFO [train.py:904] (3/8) Epoch 16, batch 5650, loss[loss=0.1921, simple_loss=0.2789, pruned_loss=0.05268, over 16626.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2997, pruned_loss=0.06732, over 3068642.22 frames. ], batch size: 57, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:45:40,979 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:45:50,670 INFO [optim.py:368] (3/8) 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 08:45:52,834 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-30 08:45:56,974 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:46:39,301 INFO [train.py:904] (3/8) Epoch 16, batch 5700, loss[loss=0.219, simple_loss=0.3057, pruned_loss=0.06621, over 16892.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3015, pruned_loss=0.06916, over 3054253.66 frames. ], batch size: 116, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:01,931 INFO [train.py:904] (3/8) Epoch 16, batch 5750, loss[loss=0.2809, simple_loss=0.3307, pruned_loss=0.1156, over 11157.00 frames. ], tot_loss[loss=0.2234, simple_loss=0.3047, pruned_loss=0.07104, over 3022123.99 frames. ], batch size: 247, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:17,315 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:48:30,523 INFO [optim.py:368] (3/8) 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 08:49:00,700 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:22,773 INFO [train.py:904] (3/8) Epoch 16, batch 5800, loss[loss=0.1949, simple_loss=0.2864, pruned_loss=0.05173, over 16828.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3038, pruned_loss=0.0692, over 3033582.60 frames. ], batch size: 96, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:49:35,953 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:36,082 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:50:16,597 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:50:28,912 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5910, 2.6190, 2.4118, 3.6718, 2.7157, 3.8649, 1.4519, 2.7602], device='cuda:3'), covar=tensor([0.1411, 0.0726, 0.1211, 0.0151, 0.0216, 0.0383, 0.1695, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0168, 0.0189, 0.0171, 0.0202, 0.0213, 0.0192, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 08:50:41,299 INFO [train.py:904] (3/8) Epoch 16, batch 5850, loss[loss=0.2202, simple_loss=0.3053, pruned_loss=0.06754, over 16357.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3013, pruned_loss=0.06738, over 3047541.48 frames. ], batch size: 146, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:50:51,620 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:51:08,528 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.851e+02 3.457e+02 4.318e+02 8.019e+02, threshold=6.915e+02, percent-clipped=1.0 2023-04-30 08:51:12,988 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 5900, loss[loss=0.1869, simple_loss=0.2861, pruned_loss=0.04383, over 16518.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3009, pruned_loss=0.06689, over 3065745.47 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:52:24,458 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:52:37,549 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 08:52:56,014 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:25,266 INFO [train.py:904] (3/8) Epoch 16, batch 5950, loss[loss=0.2026, simple_loss=0.2921, pruned_loss=0.05659, over 17223.00 frames. ], tot_loss[loss=0.217, simple_loss=0.3017, pruned_loss=0.06617, over 3043003.70 frames. ], batch size: 45, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:53:36,989 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:38,264 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:54:01,915 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:54:45,444 INFO [train.py:904] (3/8) Epoch 16, batch 6000, loss[loss=0.1949, simple_loss=0.2787, pruned_loss=0.05553, over 16774.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.3001, pruned_loss=0.06535, over 3061636.72 frames. ], batch size: 124, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:54:45,444 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 08:54:56,481 INFO [train.py:938] (3/8) Epoch 16, validation: loss=0.1553, simple_loss=0.2682, pruned_loss=0.0212, over 944034.00 frames. 2023-04-30 08:54:56,482 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 08:55:27,050 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:56:13,210 INFO [train.py:904] (3/8) Epoch 16, batch 6050, loss[loss=0.2036, simple_loss=0.2926, pruned_loss=0.05727, over 16534.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2985, pruned_loss=0.06439, over 3075660.12 frames. ], batch size: 62, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:56:15,988 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8176, 4.1566, 3.0600, 2.3531, 2.8836, 2.5416, 4.4668, 3.7545], device='cuda:3'), covar=tensor([0.2732, 0.0637, 0.1691, 0.2569, 0.2555, 0.1818, 0.0473, 0.1095], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0260, 0.0294, 0.0295, 0.0287, 0.0238, 0.0280, 0.0315], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 08:56:40,245 INFO [optim.py:368] (3/8) 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 08:57:31,988 INFO [train.py:904] (3/8) Epoch 16, batch 6100, loss[loss=0.2036, simple_loss=0.2852, pruned_loss=0.06101, over 15333.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.298, pruned_loss=0.06319, over 3085561.05 frames. ], batch size: 190, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:58:05,638 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 08:58:23,958 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:58:49,719 INFO [train.py:904] (3/8) Epoch 16, batch 6150, loss[loss=0.1824, simple_loss=0.2777, pruned_loss=0.04353, over 16845.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2958, pruned_loss=0.06199, over 3109508.21 frames. ], batch size: 96, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:59:18,314 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.901e+02 2.793e+02 3.279e+02 3.951e+02 7.811e+02, threshold=6.558e+02, percent-clipped=1.0 2023-04-30 08:59:57,107 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 6200, loss[loss=0.1912, simple_loss=0.2845, pruned_loss=0.04897, over 16364.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2947, pruned_loss=0.06258, over 3081309.11 frames. ], batch size: 146, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:00:48,495 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 6250, loss[loss=0.2052, simple_loss=0.2936, pruned_loss=0.05837, over 16376.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.294, pruned_loss=0.06197, over 3087803.84 frames. ], batch size: 146, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:01:34,537 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:35,113 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 09:01:50,903 INFO [optim.py:368] (3/8) 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 09:02:38,740 INFO [train.py:904] (3/8) Epoch 16, batch 6300, loss[loss=0.1987, simple_loss=0.2862, pruned_loss=0.05557, over 16639.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2934, pruned_loss=0.06143, over 3088662.34 frames. ], batch size: 57, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:02:45,560 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:03:28,140 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 09:03:55,408 INFO [train.py:904] (3/8) Epoch 16, batch 6350, loss[loss=0.1903, simple_loss=0.2726, pruned_loss=0.05406, over 16324.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.295, pruned_loss=0.06306, over 3083355.80 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:04:16,677 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1731, 1.9945, 1.6437, 1.6596, 2.2503, 1.9438, 2.0822, 2.3714], device='cuda:3'), covar=tensor([0.0179, 0.0341, 0.0491, 0.0441, 0.0227, 0.0330, 0.0197, 0.0219], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0217, 0.0211, 0.0210, 0.0218, 0.0219, 0.0221, 0.0213], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:04:24,044 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.016e+02 2.970e+02 3.577e+02 4.444e+02 9.031e+02, threshold=7.154e+02, percent-clipped=4.0 2023-04-30 09:04:46,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7090, 4.8780, 5.0750, 4.8336, 4.9058, 5.4935, 4.9982, 4.7477], device='cuda:3'), covar=tensor([0.1080, 0.1960, 0.2124, 0.2027, 0.2315, 0.0952, 0.1513, 0.2500], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0550, 0.0601, 0.0464, 0.0622, 0.0630, 0.0474, 0.0624], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 09:05:11,913 INFO [train.py:904] (3/8) Epoch 16, batch 6400, loss[loss=0.278, simple_loss=0.3439, pruned_loss=0.1061, over 11257.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2957, pruned_loss=0.06439, over 3075248.61 frames. ], batch size: 246, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:06:27,727 INFO [train.py:904] (3/8) Epoch 16, batch 6450, loss[loss=0.1855, simple_loss=0.2719, pruned_loss=0.04959, over 16151.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2955, pruned_loss=0.06395, over 3073374.44 frames. ], batch size: 165, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:06:56,386 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.821e+02 3.663e+02 4.597e+02 9.189e+02, threshold=7.326e+02, percent-clipped=6.0 2023-04-30 09:07:26,462 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:07:43,537 INFO [train.py:904] (3/8) Epoch 16, batch 6500, loss[loss=0.2065, simple_loss=0.2948, pruned_loss=0.05914, over 16292.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2936, pruned_loss=0.06303, over 3086246.96 frames. ], batch size: 165, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:08:11,321 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 09:08:21,057 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:22,806 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:28,684 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 09:08:31,620 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:01,975 INFO [train.py:904] (3/8) Epoch 16, batch 6550, loss[loss=0.2127, simple_loss=0.3113, pruned_loss=0.05701, over 16244.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.296, pruned_loss=0.06367, over 3086327.67 frames. ], batch size: 165, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:09:33,177 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:53,269 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:09:58,224 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:06,330 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 6600, loss[loss=0.2244, simple_loss=0.3097, pruned_loss=0.06957, over 17026.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2976, pruned_loss=0.06384, over 3097390.72 frames. ], batch size: 55, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:11:26,043 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 09:11:34,213 INFO [train.py:904] (3/8) Epoch 16, batch 6650, loss[loss=0.207, simple_loss=0.2903, pruned_loss=0.06183, over 16688.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.298, pruned_loss=0.06448, over 3096988.60 frames. ], batch size: 134, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:11:35,728 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 09:12:04,649 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 16, batch 6700, loss[loss=0.2353, simple_loss=0.3038, pruned_loss=0.08337, over 11518.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2967, pruned_loss=0.06436, over 3099345.89 frames. ], batch size: 248, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:13:57,219 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6229, 3.1329, 3.2097, 1.8502, 2.8699, 2.0112, 3.2624, 3.2547], device='cuda:3'), covar=tensor([0.0262, 0.0700, 0.0559, 0.2073, 0.0804, 0.1045, 0.0593, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0157, 0.0163, 0.0149, 0.0140, 0.0126, 0.0141, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 09:14:07,728 INFO [train.py:904] (3/8) Epoch 16, batch 6750, loss[loss=0.1829, simple_loss=0.2626, pruned_loss=0.05157, over 16143.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2952, pruned_loss=0.06395, over 3105437.27 frames. ], batch size: 35, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:33,099 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8394, 5.1743, 5.4048, 5.1678, 5.2606, 5.7922, 5.2709, 5.0725], device='cuda:3'), covar=tensor([0.1009, 0.1714, 0.1868, 0.1864, 0.2259, 0.0873, 0.1341, 0.2272], device='cuda:3'), in_proj_covar=tensor([0.0383, 0.0547, 0.0599, 0.0461, 0.0612, 0.0629, 0.0471, 0.0619], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 09:14:37,800 INFO [optim.py:368] (3/8) 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 09:15:05,339 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 6800, loss[loss=0.2028, simple_loss=0.2884, pruned_loss=0.05859, over 16686.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2953, pruned_loss=0.06379, over 3105160.50 frames. ], batch size: 62, lr: 4.22e-03, grad_scale: 8.0 2023-04-30 09:15:50,659 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:20,372 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 6850, loss[loss=0.1914, simple_loss=0.294, pruned_loss=0.0444, over 16880.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2957, pruned_loss=0.06278, over 3137509.80 frames. ], batch size: 96, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:17:12,960 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.851e+02 3.389e+02 4.147e+02 9.414e+02, threshold=6.778e+02, percent-clipped=4.0 2023-04-30 09:17:16,486 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 09:17:22,463 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:26,682 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:54,571 INFO [train.py:904] (3/8) Epoch 16, batch 6900, loss[loss=0.2612, simple_loss=0.3255, pruned_loss=0.09847, over 11536.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2979, pruned_loss=0.06231, over 3144977.94 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:18:07,182 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 09:19:13,870 INFO [train.py:904] (3/8) Epoch 16, batch 6950, loss[loss=0.2031, simple_loss=0.2931, pruned_loss=0.05657, over 16472.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.3009, pruned_loss=0.06562, over 3099714.37 frames. ], batch size: 146, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:19:42,743 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:19:48,718 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 3.353e+02 4.093e+02 5.038e+02 1.287e+03, threshold=8.186e+02, percent-clipped=11.0 2023-04-30 09:20:13,272 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-30 09:20:29,895 INFO [train.py:904] (3/8) Epoch 16, batch 7000, loss[loss=0.2154, simple_loss=0.3089, pruned_loss=0.0609, over 16416.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.3017, pruned_loss=0.06564, over 3092742.58 frames. ], batch size: 146, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:21:27,990 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:21:37,126 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 09:21:45,218 INFO [train.py:904] (3/8) Epoch 16, batch 7050, loss[loss=0.1951, simple_loss=0.2867, pruned_loss=0.05173, over 16788.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.3022, pruned_loss=0.06569, over 3084923.80 frames. ], batch size: 83, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:22:18,894 INFO [optim.py:368] (3/8) 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 09:23:00,641 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:23:01,356 INFO [train.py:904] (3/8) Epoch 16, batch 7100, loss[loss=0.2099, simple_loss=0.2923, pruned_loss=0.06374, over 17023.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.3004, pruned_loss=0.06529, over 3081446.22 frames. ], batch size: 41, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:23:54,849 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8138, 2.6640, 2.6429, 1.9259, 2.5329, 2.6445, 2.5781, 1.8877], device='cuda:3'), covar=tensor([0.0428, 0.0084, 0.0077, 0.0357, 0.0119, 0.0124, 0.0104, 0.0369], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0130, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 09:24:17,480 INFO [train.py:904] (3/8) Epoch 16, batch 7150, loss[loss=0.1945, simple_loss=0.2855, pruned_loss=0.0517, over 16236.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2985, pruned_loss=0.06468, over 3098102.56 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:49,381 INFO [optim.py:368] (3/8) 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 09:24:51,546 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:02,864 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:11,744 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:30,655 INFO [train.py:904] (3/8) Epoch 16, batch 7200, loss[loss=0.1816, simple_loss=0.274, pruned_loss=0.04454, over 16759.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2958, pruned_loss=0.06296, over 3090859.77 frames. ], batch size: 124, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:15,844 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:16,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4522, 4.0371, 4.0342, 2.7340, 3.6438, 4.0339, 3.6280, 2.3405], device='cuda:3'), covar=tensor([0.0451, 0.0035, 0.0039, 0.0326, 0.0083, 0.0091, 0.0079, 0.0379], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0074, 0.0075, 0.0129, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 09:26:26,897 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:33,351 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:26:50,665 INFO [train.py:904] (3/8) Epoch 16, batch 7250, loss[loss=0.2225, simple_loss=0.2891, pruned_loss=0.07799, over 11694.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2933, pruned_loss=0.06149, over 3098588.79 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:27:11,555 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:27:14,197 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5457, 3.5476, 2.1906, 4.1656, 2.7217, 4.0851, 2.2355, 2.7450], device='cuda:3'), covar=tensor([0.0291, 0.0418, 0.1650, 0.0171, 0.0819, 0.0527, 0.1512, 0.0796], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0169, 0.0192, 0.0146, 0.0172, 0.0209, 0.0200, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 09:27:23,390 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.928e+02 2.845e+02 3.560e+02 4.449e+02 7.164e+02, threshold=7.119e+02, percent-clipped=0.0 2023-04-30 09:27:47,291 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:28:06,272 INFO [train.py:904] (3/8) Epoch 16, batch 7300, loss[loss=0.2252, simple_loss=0.316, pruned_loss=0.06716, over 16404.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2932, pruned_loss=0.0617, over 3084024.62 frames. ], batch size: 146, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:01,804 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 09:29:15,882 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9195, 2.3487, 1.8881, 2.1795, 2.6747, 2.2936, 2.7462, 2.8956], device='cuda:3'), covar=tensor([0.0140, 0.0374, 0.0491, 0.0419, 0.0225, 0.0357, 0.0195, 0.0237], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0216, 0.0210, 0.0210, 0.0216, 0.0217, 0.0219, 0.0211], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:29:22,388 INFO [train.py:904] (3/8) Epoch 16, batch 7350, loss[loss=0.2108, simple_loss=0.2916, pruned_loss=0.06494, over 16623.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2937, pruned_loss=0.06243, over 3073992.64 frames. ], batch size: 62, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:56,735 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 3.169e+02 3.768e+02 4.450e+02 9.358e+02, threshold=7.536e+02, percent-clipped=3.0 2023-04-30 09:30:07,743 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:31,262 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 7400, loss[loss=0.1864, simple_loss=0.2863, pruned_loss=0.04329, over 16773.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2945, pruned_loss=0.06267, over 3088161.01 frames. ], batch size: 102, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:30:41,268 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-30 09:31:18,369 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4755, 3.4449, 3.4313, 2.7178, 3.3834, 2.0366, 3.1376, 2.7899], device='cuda:3'), covar=tensor([0.0191, 0.0146, 0.0220, 0.0276, 0.0121, 0.2426, 0.0147, 0.0275], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0134, 0.0181, 0.0166, 0.0153, 0.0192, 0.0167, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:31:41,336 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:31:57,485 INFO [train.py:904] (3/8) Epoch 16, batch 7450, loss[loss=0.2254, simple_loss=0.3176, pruned_loss=0.06659, over 15378.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2959, pruned_loss=0.06398, over 3069623.13 frames. ], batch size: 190, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:32:21,669 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 09:32:33,422 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 3.124e+02 3.664e+02 4.558e+02 8.463e+02, threshold=7.327e+02, percent-clipped=4.0 2023-04-30 09:32:36,545 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:33:17,743 INFO [train.py:904] (3/8) Epoch 16, batch 7500, loss[loss=0.2296, simple_loss=0.3123, pruned_loss=0.07349, over 16898.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2961, pruned_loss=0.06318, over 3075738.01 frames. ], batch size: 109, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:33:50,299 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:34:37,012 INFO [train.py:904] (3/8) Epoch 16, batch 7550, loss[loss=0.2261, simple_loss=0.2945, pruned_loss=0.07886, over 11746.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2946, pruned_loss=0.06319, over 3060124.85 frames. ], batch size: 250, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:34:58,338 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:35:10,141 INFO [optim.py:368] (3/8) 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 09:35:30,168 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 09:35:53,852 INFO [train.py:904] (3/8) Epoch 16, batch 7600, loss[loss=0.2364, simple_loss=0.3221, pruned_loss=0.07533, over 15259.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2939, pruned_loss=0.06325, over 3076371.76 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:35:59,038 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:09,859 INFO [train.py:904] (3/8) Epoch 16, batch 7650, loss[loss=0.2499, simple_loss=0.3222, pruned_loss=0.08876, over 16452.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2943, pruned_loss=0.06379, over 3085435.61 frames. ], batch size: 146, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:37:21,651 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:30,446 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 3.207e+02 4.068e+02 5.310e+02 2.281e+03, threshold=8.135e+02, percent-clipped=15.0 2023-04-30 09:37:50,760 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8633, 2.0820, 2.3891, 3.1777, 2.1806, 2.3090, 2.2615, 2.2161], device='cuda:3'), covar=tensor([0.1163, 0.3141, 0.2146, 0.0618, 0.3889, 0.2232, 0.3032, 0.3114], device='cuda:3'), in_proj_covar=tensor([0.0377, 0.0419, 0.0347, 0.0316, 0.0424, 0.0481, 0.0387, 0.0487], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:38:13,587 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:38:22,085 INFO [train.py:904] (3/8) Epoch 16, batch 7700, loss[loss=0.2161, simple_loss=0.2986, pruned_loss=0.06684, over 16490.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.295, pruned_loss=0.06442, over 3080273.72 frames. ], batch size: 75, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:38:31,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7742, 3.1895, 3.1882, 1.9801, 2.8545, 2.1793, 3.1829, 3.4050], device='cuda:3'), covar=tensor([0.0301, 0.0804, 0.0648, 0.2111, 0.0933, 0.1082, 0.0715, 0.0920], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0156, 0.0162, 0.0149, 0.0141, 0.0126, 0.0141, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 09:38:50,564 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:10,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7293, 4.6861, 4.5822, 3.8286, 4.6141, 1.7271, 4.3772, 4.3722], device='cuda:3'), covar=tensor([0.0093, 0.0086, 0.0165, 0.0378, 0.0090, 0.2625, 0.0137, 0.0199], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0133, 0.0180, 0.0165, 0.0152, 0.0191, 0.0166, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:39:15,436 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:41,049 INFO [train.py:904] (3/8) Epoch 16, batch 7750, loss[loss=0.1928, simple_loss=0.2828, pruned_loss=0.05141, over 16459.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.295, pruned_loss=0.06459, over 3070062.85 frames. ], batch size: 68, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:40:13,566 INFO [optim.py:368] (3/8) 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 09:40:24,179 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6328, 4.6509, 5.0373, 5.0241, 5.0199, 4.7111, 4.6690, 4.4698], device='cuda:3'), covar=tensor([0.0302, 0.0579, 0.0365, 0.0391, 0.0463, 0.0372, 0.0959, 0.0500], device='cuda:3'), in_proj_covar=tensor([0.0376, 0.0407, 0.0396, 0.0377, 0.0448, 0.0421, 0.0516, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 09:40:53,467 INFO [train.py:904] (3/8) Epoch 16, batch 7800, loss[loss=0.2228, simple_loss=0.3194, pruned_loss=0.06306, over 16796.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2963, pruned_loss=0.06535, over 3078793.32 frames. ], batch size: 102, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:41:04,243 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0006, 4.2135, 4.5871, 4.5209, 4.5094, 4.1733, 3.9641, 4.0605], device='cuda:3'), covar=tensor([0.0598, 0.0780, 0.0548, 0.0652, 0.0696, 0.0706, 0.1621, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0377, 0.0409, 0.0397, 0.0378, 0.0449, 0.0423, 0.0517, 0.0334], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 09:42:01,986 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 09:42:08,845 INFO [train.py:904] (3/8) Epoch 16, batch 7850, loss[loss=0.2297, simple_loss=0.2986, pruned_loss=0.08035, over 11443.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2978, pruned_loss=0.06594, over 3052753.90 frames. ], batch size: 248, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:43,406 INFO [optim.py:368] (3/8) 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 09:42:53,468 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-30 09:42:57,321 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2061, 4.0788, 4.2847, 4.4300, 4.5666, 4.2008, 4.5113, 4.5602], device='cuda:3'), covar=tensor([0.1841, 0.1232, 0.1470, 0.0667, 0.0575, 0.1177, 0.0731, 0.0628], device='cuda:3'), in_proj_covar=tensor([0.0578, 0.0712, 0.0837, 0.0716, 0.0540, 0.0567, 0.0579, 0.0676], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:43:25,067 INFO [train.py:904] (3/8) Epoch 16, batch 7900, loss[loss=0.19, simple_loss=0.282, pruned_loss=0.04905, over 16691.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2959, pruned_loss=0.06381, over 3078388.35 frames. ], batch size: 76, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:33,352 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6468, 3.7801, 2.8330, 2.1786, 2.5010, 2.3090, 4.0467, 3.4793], device='cuda:3'), covar=tensor([0.2817, 0.0622, 0.1695, 0.2545, 0.2832, 0.2059, 0.0444, 0.1136], device='cuda:3'), in_proj_covar=tensor([0.0317, 0.0262, 0.0296, 0.0297, 0.0289, 0.0239, 0.0281, 0.0319], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 09:44:43,662 INFO [train.py:904] (3/8) Epoch 16, batch 7950, loss[loss=0.2602, simple_loss=0.3157, pruned_loss=0.1023, over 11532.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2961, pruned_loss=0.0643, over 3080747.53 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:44,838 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6293, 4.2853, 4.1874, 2.8302, 3.7717, 4.2145, 3.8455, 2.3014], device='cuda:3'), covar=tensor([0.0489, 0.0042, 0.0046, 0.0373, 0.0077, 0.0102, 0.0075, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0073, 0.0074, 0.0128, 0.0087, 0.0097, 0.0085, 0.0121], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 09:44:56,864 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:45:07,703 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 09:45:16,536 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 16, batch 8000, loss[loss=0.2498, simple_loss=0.3138, pruned_loss=0.09285, over 11521.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2964, pruned_loss=0.06449, over 3084718.47 frames. ], batch size: 246, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:46:17,933 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:46:50,251 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 8050, loss[loss=0.2251, simple_loss=0.3, pruned_loss=0.07512, over 15162.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2959, pruned_loss=0.06418, over 3081380.69 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:47:20,300 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:47:47,809 INFO [optim.py:368] (3/8) 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 09:48:04,046 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 8100, loss[loss=0.2059, simple_loss=0.2886, pruned_loss=0.0616, over 16901.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2954, pruned_loss=0.06355, over 3096328.39 frames. ], batch size: 109, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:48:54,858 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:49:46,003 INFO [train.py:904] (3/8) Epoch 16, batch 8150, loss[loss=0.2192, simple_loss=0.2944, pruned_loss=0.07201, over 11520.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2924, pruned_loss=0.06216, over 3100581.53 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:50:21,106 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0450, 2.4798, 2.6032, 1.8474, 2.6707, 2.7827, 2.4375, 2.3446], device='cuda:3'), covar=tensor([0.0794, 0.0242, 0.0231, 0.1081, 0.0115, 0.0288, 0.0453, 0.0506], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0106, 0.0092, 0.0138, 0.0074, 0.0117, 0.0125, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 09:50:21,735 INFO [optim.py:368] (3/8) 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 09:51:05,066 INFO [train.py:904] (3/8) Epoch 16, batch 8200, loss[loss=0.1974, simple_loss=0.2879, pruned_loss=0.05341, over 16386.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2898, pruned_loss=0.06109, over 3108341.13 frames. ], batch size: 146, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:03,266 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 09:52:27,311 INFO [train.py:904] (3/8) Epoch 16, batch 8250, loss[loss=0.1969, simple_loss=0.302, pruned_loss=0.04591, over 15387.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2896, pruned_loss=0.05902, over 3098461.63 frames. ], batch size: 191, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:42,328 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:53:04,605 INFO [optim.py:368] (3/8) 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 09:53:49,472 INFO [train.py:904] (3/8) Epoch 16, batch 8300, loss[loss=0.1914, simple_loss=0.284, pruned_loss=0.04939, over 16921.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2875, pruned_loss=0.05663, over 3097148.52 frames. ], batch size: 109, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:54:00,904 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:13,042 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:56,877 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 09:55:10,565 INFO [train.py:904] (3/8) Epoch 16, batch 8350, loss[loss=0.201, simple_loss=0.2797, pruned_loss=0.06111, over 12067.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.286, pruned_loss=0.05455, over 3053685.03 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:55:30,913 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 09:56:23,283 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0857, 5.4041, 5.1857, 5.1668, 4.9054, 4.8516, 4.8133, 5.5081], device='cuda:3'), covar=tensor([0.1192, 0.0835, 0.0935, 0.0805, 0.0828, 0.0801, 0.1150, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0600, 0.0732, 0.0598, 0.0539, 0.0460, 0.0477, 0.0610, 0.0562], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:56:33,055 INFO [train.py:904] (3/8) Epoch 16, batch 8400, loss[loss=0.1708, simple_loss=0.2658, pruned_loss=0.0379, over 15331.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2834, pruned_loss=0.05248, over 3065359.74 frames. ], batch size: 191, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:56:50,697 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:57:54,561 INFO [train.py:904] (3/8) Epoch 16, batch 8450, loss[loss=0.1816, simple_loss=0.2744, pruned_loss=0.04442, over 15426.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2812, pruned_loss=0.05075, over 3059681.44 frames. ], batch size: 190, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:58:31,816 INFO [optim.py:368] (3/8) 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 09:58:39,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7783, 4.1000, 3.2863, 2.2334, 2.6731, 2.6174, 4.4576, 3.5331], device='cuda:3'), covar=tensor([0.2652, 0.0499, 0.1368, 0.2750, 0.2870, 0.1824, 0.0277, 0.1100], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0256, 0.0288, 0.0291, 0.0281, 0.0234, 0.0274, 0.0310], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 09:59:06,538 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 09:59:15,543 INFO [train.py:904] (3/8) Epoch 16, batch 8500, loss[loss=0.1857, simple_loss=0.264, pruned_loss=0.05373, over 12021.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2778, pruned_loss=0.04874, over 3055247.76 frames. ], batch size: 246, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:59:40,128 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9309, 4.3051, 3.3375, 2.3442, 2.7954, 2.6000, 4.5374, 3.7695], device='cuda:3'), covar=tensor([0.2486, 0.0458, 0.1407, 0.2845, 0.2800, 0.1909, 0.0306, 0.1072], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0255, 0.0288, 0.0290, 0.0280, 0.0234, 0.0273, 0.0309], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:00:26,854 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-30 10:00:42,447 INFO [train.py:904] (3/8) Epoch 16, batch 8550, loss[loss=0.1999, simple_loss=0.2805, pruned_loss=0.05961, over 11943.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2752, pruned_loss=0.04741, over 3042260.29 frames. ], batch size: 250, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:01:27,141 INFO [optim.py:368] (3/8) 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 10:01:49,274 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2490, 3.6586, 3.9060, 2.1057, 3.0909, 2.4513, 3.6865, 3.8228], device='cuda:3'), covar=tensor([0.0272, 0.0762, 0.0515, 0.2013, 0.0847, 0.1001, 0.0683, 0.0977], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0152, 0.0159, 0.0146, 0.0138, 0.0124, 0.0138, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 10:02:22,622 INFO [train.py:904] (3/8) Epoch 16, batch 8600, loss[loss=0.1888, simple_loss=0.2684, pruned_loss=0.05453, over 12450.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2754, pruned_loss=0.0465, over 3038910.16 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:02,589 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:04:03,215 INFO [train.py:904] (3/8) Epoch 16, batch 8650, loss[loss=0.1586, simple_loss=0.263, pruned_loss=0.02709, over 16538.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2739, pruned_loss=0.04506, over 3030486.88 frames. ], batch size: 68, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:12,830 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4102, 3.3357, 3.4615, 3.5340, 3.5862, 3.2981, 3.5513, 3.6151], device='cuda:3'), covar=tensor([0.1231, 0.0918, 0.1085, 0.0633, 0.0571, 0.2297, 0.0731, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0554, 0.0685, 0.0804, 0.0693, 0.0522, 0.0547, 0.0556, 0.0653], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:04:48,097 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 10:04:56,266 INFO [optim.py:368] (3/8) 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 10:05:52,489 INFO [train.py:904] (3/8) Epoch 16, batch 8700, loss[loss=0.1682, simple_loss=0.2606, pruned_loss=0.0379, over 12208.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2708, pruned_loss=0.04343, over 3039251.04 frames. ], batch size: 250, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:06:13,629 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:15,162 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 10:06:18,527 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:07:29,335 INFO [train.py:904] (3/8) Epoch 16, batch 8750, loss[loss=0.1948, simple_loss=0.2957, pruned_loss=0.047, over 16243.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2705, pruned_loss=0.043, over 3036856.84 frames. ], batch size: 165, lr: 4.20e-03, grad_scale: 4.0 2023-04-30 10:07:53,443 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 10:08:27,683 INFO [optim.py:368] (3/8) 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 10:08:28,271 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:08:35,210 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 10:09:20,776 INFO [train.py:904] (3/8) Epoch 16, batch 8800, loss[loss=0.1585, simple_loss=0.2572, pruned_loss=0.02988, over 16689.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2688, pruned_loss=0.04201, over 3050784.00 frames. ], batch size: 89, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:05,640 INFO [train.py:904] (3/8) Epoch 16, batch 8850, loss[loss=0.1778, simple_loss=0.2821, pruned_loss=0.03678, over 16755.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2706, pruned_loss=0.04145, over 3025509.01 frames. ], batch size: 124, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:28,062 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0524, 4.0044, 3.9108, 3.1530, 3.9787, 1.7244, 3.7830, 3.4947], device='cuda:3'), covar=tensor([0.0086, 0.0077, 0.0168, 0.0249, 0.0090, 0.2545, 0.0120, 0.0248], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0129, 0.0174, 0.0159, 0.0148, 0.0188, 0.0162, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:11:55,942 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.310e+02 2.876e+02 3.561e+02 5.593e+02, threshold=5.752e+02, percent-clipped=2.0 2023-04-30 10:12:53,419 INFO [train.py:904] (3/8) Epoch 16, batch 8900, loss[loss=0.1816, simple_loss=0.2797, pruned_loss=0.04175, over 16650.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2711, pruned_loss=0.04081, over 3031645.34 frames. ], batch size: 89, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:14:59,663 INFO [train.py:904] (3/8) Epoch 16, batch 8950, loss[loss=0.1701, simple_loss=0.2609, pruned_loss=0.03963, over 16951.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2711, pruned_loss=0.04136, over 3042775.70 frames. ], batch size: 116, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:15:29,229 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:15:47,454 INFO [optim.py:368] (3/8) 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 10:16:01,815 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:16:46,879 INFO [train.py:904] (3/8) Epoch 16, batch 9000, loss[loss=0.1616, simple_loss=0.251, pruned_loss=0.03604, over 16461.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2676, pruned_loss=0.03985, over 3045770.74 frames. ], batch size: 68, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:16:46,880 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 10:16:56,910 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 10:17:08,097 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:48,815 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:20,846 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:23,598 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 10:18:40,788 INFO [train.py:904] (3/8) Epoch 16, batch 9050, loss[loss=0.1627, simple_loss=0.2434, pruned_loss=0.04104, over 16931.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2688, pruned_loss=0.04011, over 3071313.87 frames. ], batch size: 109, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:18:53,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6753, 4.7964, 4.9897, 4.8016, 4.8619, 5.3619, 4.8569, 4.5877], device='cuda:3'), covar=tensor([0.1058, 0.1695, 0.1758, 0.1783, 0.2117, 0.0920, 0.1381, 0.2339], device='cuda:3'), in_proj_covar=tensor([0.0364, 0.0530, 0.0573, 0.0442, 0.0585, 0.0609, 0.0453, 0.0594], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:19:19,380 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:19:52,343 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1771, 2.3579, 2.0341, 2.1397, 2.7128, 2.4403, 2.8077, 2.9458], device='cuda:3'), covar=tensor([0.0125, 0.0396, 0.0471, 0.0457, 0.0229, 0.0354, 0.0201, 0.0217], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0215, 0.0209, 0.0208, 0.0213, 0.0214, 0.0215, 0.0207], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:20:22,817 INFO [train.py:904] (3/8) Epoch 16, batch 9100, loss[loss=0.1737, simple_loss=0.2688, pruned_loss=0.0393, over 16975.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2688, pruned_loss=0.04099, over 3065481.26 frames. ], batch size: 109, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:22:19,320 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:22:19,974 INFO [train.py:904] (3/8) Epoch 16, batch 9150, loss[loss=0.1797, simple_loss=0.2673, pruned_loss=0.04606, over 12168.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2688, pruned_loss=0.04044, over 3056779.54 frames. ], batch size: 250, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:23:13,971 INFO [optim.py:368] (3/8) 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 10:24:04,518 INFO [train.py:904] (3/8) Epoch 16, batch 9200, loss[loss=0.1833, simple_loss=0.2798, pruned_loss=0.04342, over 16936.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2656, pruned_loss=0.04009, over 3062928.51 frames. ], batch size: 116, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:24:20,322 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:24:24,463 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:25:08,308 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 10:25:42,686 INFO [train.py:904] (3/8) Epoch 16, batch 9250, loss[loss=0.1735, simple_loss=0.2681, pruned_loss=0.03948, over 15362.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2655, pruned_loss=0.04045, over 3038715.89 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:26:22,896 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:26:36,793 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 16, batch 9300, loss[loss=0.1747, simple_loss=0.2599, pruned_loss=0.0447, over 12356.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.264, pruned_loss=0.03979, over 3044379.41 frames. ], batch size: 248, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:27:45,412 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:23,207 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:42,122 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-30 10:28:54,333 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:29:21,683 INFO [train.py:904] (3/8) Epoch 16, batch 9350, loss[loss=0.1658, simple_loss=0.2518, pruned_loss=0.03989, over 16528.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2638, pruned_loss=0.0394, over 3071537.27 frames. ], batch size: 62, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:29:28,432 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:02,100 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.307e+02 2.790e+02 3.265e+02 7.040e+02, threshold=5.579e+02, percent-clipped=1.0 2023-04-30 10:30:13,617 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4720, 1.9881, 1.7118, 1.7512, 2.2537, 1.9519, 2.0182, 2.3653], device='cuda:3'), covar=tensor([0.0135, 0.0391, 0.0509, 0.0437, 0.0274, 0.0341, 0.0175, 0.0233], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0218, 0.0211, 0.0211, 0.0216, 0.0217, 0.0217, 0.0208], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:30:52,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9532, 3.9458, 2.7075, 4.6474, 3.0428, 4.5493, 2.6145, 3.2890], device='cuda:3'), covar=tensor([0.0228, 0.0368, 0.1381, 0.0233, 0.0752, 0.0453, 0.1397, 0.0627], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0163, 0.0184, 0.0140, 0.0166, 0.0199, 0.0194, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 10:31:03,992 INFO [train.py:904] (3/8) Epoch 16, batch 9400, loss[loss=0.1753, simple_loss=0.2734, pruned_loss=0.0386, over 16133.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2633, pruned_loss=0.03889, over 3072895.66 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:31:39,354 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:32:41,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3946, 1.6671, 2.0122, 2.3473, 2.3668, 2.6105, 1.7438, 2.5940], device='cuda:3'), covar=tensor([0.0218, 0.0495, 0.0341, 0.0309, 0.0310, 0.0217, 0.0510, 0.0170], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0179, 0.0165, 0.0167, 0.0179, 0.0135, 0.0180, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:32:44,156 INFO [train.py:904] (3/8) Epoch 16, batch 9450, loss[loss=0.1758, simple_loss=0.2633, pruned_loss=0.04412, over 12434.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2651, pruned_loss=0.03903, over 3063803.74 frames. ], batch size: 250, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:33:33,850 INFO [optim.py:368] (3/8) 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 10:34:01,025 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 10:34:23,901 INFO [train.py:904] (3/8) Epoch 16, batch 9500, loss[loss=0.158, simple_loss=0.2444, pruned_loss=0.03577, over 13014.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2639, pruned_loss=0.03859, over 3059741.85 frames. ], batch size: 248, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:34:37,541 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 9550, loss[loss=0.2, simple_loss=0.2981, pruned_loss=0.05097, over 15446.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2637, pruned_loss=0.03888, over 3050908.12 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:36:38,120 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:40,705 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:37:00,497 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.146e+02 2.559e+02 3.093e+02 5.125e+02, threshold=5.118e+02, percent-clipped=0.0 2023-04-30 10:37:51,436 INFO [train.py:904] (3/8) Epoch 16, batch 9600, loss[loss=0.1922, simple_loss=0.2675, pruned_loss=0.05849, over 12484.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2656, pruned_loss=0.03976, over 3060390.12 frames. ], batch size: 247, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:38:05,320 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4672, 3.3154, 3.6146, 1.6927, 3.7702, 3.8294, 2.9945, 2.7982], device='cuda:3'), covar=tensor([0.0784, 0.0279, 0.0179, 0.1348, 0.0082, 0.0147, 0.0412, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0103, 0.0087, 0.0136, 0.0071, 0.0113, 0.0122, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 10:38:22,292 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 10:38:29,818 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:38:38,463 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:02,372 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:37,505 INFO [train.py:904] (3/8) Epoch 16, batch 9650, loss[loss=0.1715, simple_loss=0.2636, pruned_loss=0.03969, over 16854.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2675, pruned_loss=0.03973, over 3081104.40 frames. ], batch size: 42, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:39:44,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8313, 3.7246, 3.9352, 3.7386, 3.8901, 4.3014, 3.9635, 3.7326], device='cuda:3'), covar=tensor([0.2171, 0.2421, 0.2052, 0.2331, 0.2798, 0.1608, 0.1524, 0.2408], device='cuda:3'), in_proj_covar=tensor([0.0360, 0.0523, 0.0570, 0.0437, 0.0581, 0.0605, 0.0451, 0.0586], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:40:22,071 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:36,068 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:56,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5847, 3.6296, 3.4527, 3.1477, 3.3030, 3.5314, 3.3253, 3.4153], device='cuda:3'), covar=tensor([0.0558, 0.0593, 0.0285, 0.0256, 0.0565, 0.0468, 0.1343, 0.0500], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0355, 0.0297, 0.0286, 0.0306, 0.0332, 0.0205, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-30 10:41:23,666 INFO [train.py:904] (3/8) Epoch 16, batch 9700, loss[loss=0.1694, simple_loss=0.257, pruned_loss=0.04094, over 12687.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2667, pruned_loss=0.03972, over 3080048.95 frames. ], batch size: 250, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:41:34,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7479, 4.8756, 5.0379, 4.8699, 4.9451, 5.4351, 4.9244, 4.6284], device='cuda:3'), covar=tensor([0.0985, 0.1783, 0.2138, 0.1767, 0.2193, 0.0883, 0.1338, 0.2216], device='cuda:3'), in_proj_covar=tensor([0.0357, 0.0519, 0.0567, 0.0434, 0.0576, 0.0601, 0.0447, 0.0581], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:42:47,560 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0038, 5.2480, 5.4410, 5.3021, 5.3220, 5.8221, 5.3776, 5.1184], device='cuda:3'), covar=tensor([0.0797, 0.1850, 0.2297, 0.1614, 0.2024, 0.0855, 0.1335, 0.2273], device='cuda:3'), in_proj_covar=tensor([0.0356, 0.0517, 0.0565, 0.0432, 0.0574, 0.0599, 0.0446, 0.0579], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:42:49,564 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2937, 4.3622, 4.1943, 3.8771, 3.9110, 4.2929, 4.0198, 4.0352], device='cuda:3'), covar=tensor([0.0610, 0.0612, 0.0303, 0.0291, 0.0806, 0.0465, 0.0652, 0.0635], device='cuda:3'), in_proj_covar=tensor([0.0255, 0.0353, 0.0296, 0.0285, 0.0304, 0.0331, 0.0204, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-30 10:43:08,496 INFO [train.py:904] (3/8) Epoch 16, batch 9750, loss[loss=0.1661, simple_loss=0.2483, pruned_loss=0.04196, over 12372.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2651, pruned_loss=0.03953, over 3079385.45 frames. ], batch size: 246, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:43:34,427 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 10:43:58,494 INFO [optim.py:368] (3/8) 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 10:44:37,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4494, 3.7313, 3.9049, 2.0934, 3.2610, 2.5390, 3.8164, 3.7749], device='cuda:3'), covar=tensor([0.0184, 0.0650, 0.0451, 0.1870, 0.0624, 0.0880, 0.0524, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0146, 0.0157, 0.0145, 0.0136, 0.0123, 0.0137, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 10:44:46,280 INFO [train.py:904] (3/8) Epoch 16, batch 9800, loss[loss=0.1726, simple_loss=0.275, pruned_loss=0.0351, over 16981.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2649, pruned_loss=0.03866, over 3083357.11 frames. ], batch size: 109, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:44:56,988 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:45:34,289 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:31,255 INFO [train.py:904] (3/8) Epoch 16, batch 9850, loss[loss=0.1774, simple_loss=0.258, pruned_loss=0.04838, over 12269.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2662, pruned_loss=0.03853, over 3079184.39 frames. ], batch size: 246, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:46:38,615 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:59,789 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:47:23,357 INFO [optim.py:368] (3/8) 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 10:47:51,498 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 16, batch 9900, loss[loss=0.1748, simple_loss=0.2783, pruned_loss=0.03561, over 16773.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2661, pruned_loss=0.038, over 3066772.77 frames. ], batch size: 83, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:48:52,696 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:28,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8868, 2.7198, 2.6273, 1.9785, 2.4794, 2.7199, 2.6143, 1.8670], device='cuda:3'), covar=tensor([0.0396, 0.0059, 0.0059, 0.0322, 0.0119, 0.0084, 0.0096, 0.0425], device='cuda:3'), in_proj_covar=tensor([0.0129, 0.0073, 0.0073, 0.0128, 0.0088, 0.0095, 0.0084, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:50:22,154 INFO [train.py:904] (3/8) Epoch 16, batch 9950, loss[loss=0.177, simple_loss=0.2725, pruned_loss=0.04077, over 15391.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2687, pruned_loss=0.03917, over 3059261.41 frames. ], batch size: 190, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:51:02,416 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5549, 3.6556, 2.8896, 2.2661, 2.4581, 2.4461, 3.8287, 3.4411], device='cuda:3'), covar=tensor([0.2804, 0.0569, 0.1690, 0.2737, 0.2102, 0.1745, 0.0463, 0.0972], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0251, 0.0283, 0.0285, 0.0269, 0.0231, 0.0270, 0.0303], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:51:26,523 INFO [optim.py:368] (3/8) 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 10:51:33,735 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4712, 4.6118, 4.7795, 4.6369, 4.6807, 5.1711, 4.6758, 4.4396], device='cuda:3'), covar=tensor([0.1531, 0.2016, 0.2031, 0.2012, 0.2447, 0.1042, 0.1547, 0.2658], device='cuda:3'), in_proj_covar=tensor([0.0355, 0.0514, 0.0564, 0.0433, 0.0575, 0.0598, 0.0447, 0.0578], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:51:36,405 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:52:06,251 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 10:52:23,901 INFO [train.py:904] (3/8) Epoch 16, batch 10000, loss[loss=0.1665, simple_loss=0.2523, pruned_loss=0.04035, over 12680.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2672, pruned_loss=0.03893, over 3065553.11 frames. ], batch size: 247, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:52:42,867 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6677, 4.4863, 4.7092, 4.8709, 5.0343, 4.5586, 5.0373, 5.0323], device='cuda:3'), covar=tensor([0.1528, 0.1143, 0.1450, 0.0649, 0.0470, 0.0726, 0.0444, 0.0501], device='cuda:3'), in_proj_covar=tensor([0.0555, 0.0684, 0.0802, 0.0692, 0.0522, 0.0546, 0.0558, 0.0652], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:53:46,665 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:54:04,077 INFO [train.py:904] (3/8) Epoch 16, batch 10050, loss[loss=0.1593, simple_loss=0.2571, pruned_loss=0.03076, over 16893.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.267, pruned_loss=0.03868, over 3056854.72 frames. ], batch size: 96, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:54:04,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9210, 2.2883, 2.2827, 2.9997, 1.9145, 3.2473, 1.6781, 2.7037], device='cuda:3'), covar=tensor([0.1260, 0.0671, 0.1065, 0.0137, 0.0093, 0.0365, 0.1516, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0161, 0.0184, 0.0161, 0.0188, 0.0203, 0.0189, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-30 10:54:54,659 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.225e+02 2.711e+02 3.158e+02 7.314e+02, threshold=5.421e+02, percent-clipped=6.0 2023-04-30 10:55:24,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8374, 3.8117, 4.0049, 3.8436, 3.9693, 4.3269, 3.9417, 3.6622], device='cuda:3'), covar=tensor([0.2072, 0.2354, 0.1977, 0.2109, 0.2452, 0.1458, 0.1560, 0.2526], device='cuda:3'), in_proj_covar=tensor([0.0357, 0.0515, 0.0566, 0.0434, 0.0576, 0.0599, 0.0449, 0.0581], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 10:55:38,457 INFO [train.py:904] (3/8) Epoch 16, batch 10100, loss[loss=0.1565, simple_loss=0.2412, pruned_loss=0.03586, over 12584.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2678, pruned_loss=0.0389, over 3069281.90 frames. ], batch size: 249, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:57:23,030 INFO [train.py:904] (3/8) Epoch 17, batch 0, loss[loss=0.2348, simple_loss=0.294, pruned_loss=0.08785, over 16874.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.294, pruned_loss=0.08785, over 16874.00 frames. ], batch size: 96, lr: 4.05e-03, grad_scale: 8.0 2023-04-30 10:57:23,030 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 10:57:30,750 INFO [train.py:938] (3/8) Epoch 17, validation: loss=0.1481, simple_loss=0.2518, pruned_loss=0.02217, over 944034.00 frames. 2023-04-30 10:57:30,750 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 10:57:39,282 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4604, 4.7923, 4.5642, 4.5886, 4.3349, 4.2949, 4.3416, 4.8323], device='cuda:3'), covar=tensor([0.1335, 0.0937, 0.1151, 0.0831, 0.0922, 0.1326, 0.1094, 0.1006], device='cuda:3'), in_proj_covar=tensor([0.0582, 0.0715, 0.0578, 0.0526, 0.0452, 0.0464, 0.0595, 0.0553], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 10:58:09,503 INFO [optim.py:368] (3/8) 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 10:58:14,196 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:58:32,296 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9081, 2.9808, 2.5770, 4.3604, 3.5571, 4.1735, 1.5615, 2.9959], device='cuda:3'), covar=tensor([0.1327, 0.0577, 0.1063, 0.0126, 0.0157, 0.0367, 0.1525, 0.0758], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0161, 0.0184, 0.0162, 0.0188, 0.0204, 0.0189, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:3') 2023-04-30 10:58:39,851 INFO [train.py:904] (3/8) Epoch 17, batch 50, loss[loss=0.2072, simple_loss=0.2675, pruned_loss=0.07348, over 16867.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2767, pruned_loss=0.05389, over 754785.14 frames. ], batch size: 116, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 10:59:06,442 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:59:47,951 INFO [train.py:904] (3/8) Epoch 17, batch 100, loss[loss=0.1843, simple_loss=0.2796, pruned_loss=0.04447, over 17091.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2712, pruned_loss=0.05012, over 1322760.65 frames. ], batch size: 53, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:00:00,627 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:12,328 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.355e+02 2.682e+02 3.198e+02 6.473e+02, threshold=5.364e+02, percent-clipped=1.0 2023-04-30 11:00:56,544 INFO [train.py:904] (3/8) Epoch 17, batch 150, loss[loss=0.1573, simple_loss=0.2453, pruned_loss=0.0346, over 16845.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2707, pruned_loss=0.04985, over 1752551.81 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:01:17,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2400, 5.0099, 5.2346, 5.4289, 5.6084, 4.9056, 5.5092, 5.5804], device='cuda:3'), covar=tensor([0.1822, 0.1314, 0.1790, 0.0797, 0.0575, 0.0743, 0.0649, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0565, 0.0697, 0.0821, 0.0704, 0.0534, 0.0555, 0.0570, 0.0662], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:01:23,628 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 11:01:44,470 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:02:05,937 INFO [train.py:904] (3/8) Epoch 17, batch 200, loss[loss=0.1413, simple_loss=0.2275, pruned_loss=0.02752, over 16989.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2686, pruned_loss=0.04885, over 2107985.98 frames. ], batch size: 41, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:02:34,532 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-30 11:02:43,588 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.407e+02 2.768e+02 3.167e+02 5.394e+02, threshold=5.535e+02, percent-clipped=0.0 2023-04-30 11:03:12,317 INFO [train.py:904] (3/8) Epoch 17, batch 250, loss[loss=0.1797, simple_loss=0.257, pruned_loss=0.05123, over 16456.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2676, pruned_loss=0.04886, over 2379036.95 frames. ], batch size: 68, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:20,330 INFO [train.py:904] (3/8) Epoch 17, batch 300, loss[loss=0.1723, simple_loss=0.2482, pruned_loss=0.04815, over 16757.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2655, pruned_loss=0.04717, over 2583027.63 frames. ], batch size: 134, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:55,542 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8636, 2.5815, 2.5787, 4.0237, 3.3198, 4.0385, 1.4550, 2.9143], device='cuda:3'), covar=tensor([0.1349, 0.0678, 0.1078, 0.0170, 0.0147, 0.0351, 0.1576, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0164, 0.0187, 0.0168, 0.0194, 0.0209, 0.0193, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:04:59,733 INFO [optim.py:368] (3/8) 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 11:05:03,694 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:05:29,583 INFO [train.py:904] (3/8) Epoch 17, batch 350, loss[loss=0.1863, simple_loss=0.2588, pruned_loss=0.05695, over 16900.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2622, pruned_loss=0.04621, over 2746964.41 frames. ], batch size: 96, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:05:45,429 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5172, 2.3205, 2.5059, 4.2553, 2.3426, 2.6984, 2.4487, 2.5586], device='cuda:3'), covar=tensor([0.1117, 0.3316, 0.2612, 0.0517, 0.3852, 0.2380, 0.3317, 0.3129], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0419, 0.0353, 0.0319, 0.0426, 0.0480, 0.0390, 0.0488], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:06:07,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4354, 5.8573, 5.4091, 5.8382, 5.3360, 5.0853, 5.4707, 5.9207], device='cuda:3'), covar=tensor([0.2646, 0.1837, 0.3049, 0.1320, 0.1723, 0.1396, 0.2224, 0.2092], device='cuda:3'), in_proj_covar=tensor([0.0617, 0.0762, 0.0619, 0.0559, 0.0480, 0.0489, 0.0635, 0.0586], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:06:07,975 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:06:34,570 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5808, 3.6140, 3.3049, 3.0717, 3.2302, 3.4951, 3.3166, 3.3101], device='cuda:3'), covar=tensor([0.0528, 0.0562, 0.0279, 0.0292, 0.0568, 0.0418, 0.1268, 0.0498], device='cuda:3'), in_proj_covar=tensor([0.0268, 0.0374, 0.0313, 0.0302, 0.0323, 0.0350, 0.0216, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:06:36,763 INFO [train.py:904] (3/8) Epoch 17, batch 400, loss[loss=0.2065, simple_loss=0.2778, pruned_loss=0.06756, over 16889.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2602, pruned_loss=0.04552, over 2874818.41 frames. ], batch size: 109, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:06:37,514 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 11:07:10,948 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:07:15,899 INFO [optim.py:368] (3/8) 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 11:07:34,483 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 11:07:35,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0180, 4.7853, 5.0303, 5.2597, 5.4391, 4.8078, 5.4120, 5.4237], device='cuda:3'), covar=tensor([0.1880, 0.1234, 0.1625, 0.0670, 0.0517, 0.0785, 0.0526, 0.0533], device='cuda:3'), in_proj_covar=tensor([0.0589, 0.0728, 0.0859, 0.0729, 0.0553, 0.0579, 0.0595, 0.0688], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:07:42,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2885, 3.3426, 3.5143, 2.1335, 3.0682, 2.4529, 3.6941, 3.6042], device='cuda:3'), covar=tensor([0.0244, 0.0862, 0.0579, 0.1925, 0.0773, 0.0970, 0.0516, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0149, 0.0139, 0.0126, 0.0139, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:07:46,582 INFO [train.py:904] (3/8) Epoch 17, batch 450, loss[loss=0.1668, simple_loss=0.2407, pruned_loss=0.04641, over 16465.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2578, pruned_loss=0.04491, over 2965676.05 frames. ], batch size: 146, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:08:06,778 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 11:08:34,336 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:35,593 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 500, loss[loss=0.1475, simple_loss=0.2333, pruned_loss=0.03082, over 17213.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2569, pruned_loss=0.04402, over 3049486.68 frames. ], batch size: 43, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:08:56,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5613, 3.2893, 2.7193, 2.1031, 2.2109, 2.2735, 3.4591, 3.0588], device='cuda:3'), covar=tensor([0.2698, 0.0811, 0.1725, 0.2937, 0.2656, 0.2033, 0.0569, 0.1464], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0259, 0.0292, 0.0295, 0.0282, 0.0239, 0.0279, 0.0316], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:09:32,593 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.133e+02 2.499e+02 3.161e+02 5.064e+02, threshold=4.998e+02, percent-clipped=0.0 2023-04-30 11:09:39,867 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:09:41,216 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2117, 5.1471, 5.0709, 4.5407, 4.6471, 5.1032, 5.0492, 4.7147], device='cuda:3'), covar=tensor([0.0500, 0.0447, 0.0301, 0.0342, 0.1102, 0.0413, 0.0320, 0.0736], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0379, 0.0318, 0.0306, 0.0328, 0.0356, 0.0218, 0.0381], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:09:48,067 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8767, 5.2143, 4.9649, 4.9981, 4.7298, 4.6780, 4.6597, 5.2686], device='cuda:3'), covar=tensor([0.1120, 0.0848, 0.1080, 0.0780, 0.0790, 0.1032, 0.1186, 0.0886], device='cuda:3'), in_proj_covar=tensor([0.0624, 0.0771, 0.0630, 0.0567, 0.0486, 0.0495, 0.0642, 0.0594], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:10:01,809 INFO [train.py:904] (3/8) Epoch 17, batch 550, loss[loss=0.1933, simple_loss=0.2677, pruned_loss=0.05949, over 16473.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2569, pruned_loss=0.0442, over 3116521.29 frames. ], batch size: 146, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:10:18,872 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 11:11:10,218 INFO [train.py:904] (3/8) Epoch 17, batch 600, loss[loss=0.1752, simple_loss=0.2462, pruned_loss=0.05207, over 16948.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2571, pruned_loss=0.04557, over 3164050.12 frames. ], batch size: 109, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:47,293 INFO [optim.py:368] (3/8) 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 11:12:09,265 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6717, 4.8990, 5.0697, 4.8901, 4.9368, 5.5598, 5.0572, 4.7350], device='cuda:3'), covar=tensor([0.1432, 0.2415, 0.2596, 0.2105, 0.2696, 0.1087, 0.1770, 0.2609], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0564, 0.0619, 0.0474, 0.0632, 0.0652, 0.0489, 0.0634], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:12:16,980 INFO [train.py:904] (3/8) Epoch 17, batch 650, loss[loss=0.1479, simple_loss=0.2303, pruned_loss=0.03279, over 17010.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2554, pruned_loss=0.04481, over 3188485.88 frames. ], batch size: 41, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:09,335 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:13:20,289 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 11:13:25,533 INFO [train.py:904] (3/8) Epoch 17, batch 700, loss[loss=0.1544, simple_loss=0.2435, pruned_loss=0.03271, over 16778.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2545, pruned_loss=0.04363, over 3217361.23 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:35,846 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7052, 2.3602, 2.3755, 4.5863, 2.4433, 2.7018, 2.4408, 2.5713], device='cuda:3'), covar=tensor([0.0994, 0.3521, 0.2751, 0.0403, 0.3745, 0.2472, 0.3271, 0.3529], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0420, 0.0354, 0.0322, 0.0427, 0.0483, 0.0392, 0.0492], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:13:36,824 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6603, 1.7424, 1.6305, 1.4466, 1.8832, 1.5380, 1.6369, 1.8703], device='cuda:3'), covar=tensor([0.0207, 0.0250, 0.0347, 0.0330, 0.0186, 0.0234, 0.0155, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0228, 0.0220, 0.0219, 0.0227, 0.0228, 0.0231, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:14:04,486 INFO [optim.py:368] (3/8) 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 11:14:34,353 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 750, loss[loss=0.1557, simple_loss=0.2565, pruned_loss=0.02749, over 17051.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2553, pruned_loss=0.04422, over 3228428.95 frames. ], batch size: 50, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:57,126 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 11:15:05,474 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 11:15:18,049 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:15:36,549 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 11:15:44,388 INFO [train.py:904] (3/8) Epoch 17, batch 800, loss[loss=0.2048, simple_loss=0.2828, pruned_loss=0.06346, over 16866.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2551, pruned_loss=0.04418, over 3245173.91 frames. ], batch size: 116, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:16:03,278 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 11:16:36,543 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:53,229 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7317, 4.1297, 4.1355, 2.1665, 3.3613, 2.5096, 4.1881, 4.2697], device='cuda:3'), covar=tensor([0.0245, 0.0724, 0.0530, 0.2213, 0.0873, 0.1061, 0.0594, 0.1054], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0155, 0.0162, 0.0150, 0.0141, 0.0127, 0.0141, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:16:53,792 INFO [train.py:904] (3/8) Epoch 17, batch 850, loss[loss=0.1527, simple_loss=0.2381, pruned_loss=0.03361, over 15978.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2546, pruned_loss=0.0438, over 3253447.07 frames. ], batch size: 35, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:17:17,159 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7081, 2.6804, 2.2867, 2.6729, 3.0395, 2.8800, 3.3905, 3.2924], device='cuda:3'), covar=tensor([0.0138, 0.0379, 0.0451, 0.0399, 0.0240, 0.0333, 0.0229, 0.0239], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0229, 0.0220, 0.0219, 0.0227, 0.0229, 0.0233, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:17:18,365 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0765, 4.4973, 3.2577, 2.4315, 2.8124, 2.6307, 4.8439, 3.7369], device='cuda:3'), covar=tensor([0.2431, 0.0534, 0.1600, 0.2672, 0.2882, 0.1952, 0.0299, 0.1273], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0260, 0.0293, 0.0295, 0.0283, 0.0240, 0.0280, 0.0318], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:18:01,094 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 900, loss[loss=0.1762, simple_loss=0.249, pruned_loss=0.05165, over 16931.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2541, pruned_loss=0.043, over 3275328.88 frames. ], batch size: 96, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:18:40,393 INFO [optim.py:368] (3/8) 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 11:19:09,573 INFO [train.py:904] (3/8) Epoch 17, batch 950, loss[loss=0.1686, simple_loss=0.2577, pruned_loss=0.03973, over 17201.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2543, pruned_loss=0.04327, over 3279201.60 frames. ], batch size: 44, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:20:01,574 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 11:20:17,960 INFO [train.py:904] (3/8) Epoch 17, batch 1000, loss[loss=0.1488, simple_loss=0.2339, pruned_loss=0.03179, over 17196.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2532, pruned_loss=0.04314, over 3291533.91 frames. ], batch size: 44, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:20:23,188 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 11:20:42,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5697, 3.7650, 4.1545, 2.2143, 3.3974, 2.6324, 3.9854, 3.9241], device='cuda:3'), covar=tensor([0.0246, 0.0791, 0.0420, 0.1875, 0.0704, 0.0869, 0.0568, 0.1058], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0151, 0.0141, 0.0127, 0.0142, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:20:54,826 INFO [optim.py:368] (3/8) 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 11:21:18,964 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:21:23,953 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1692, 3.2320, 3.3387, 2.1563, 2.8977, 2.4380, 3.6451, 3.5949], device='cuda:3'), covar=tensor([0.0252, 0.0905, 0.0638, 0.1894, 0.0845, 0.0971, 0.0531, 0.0931], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0156, 0.0163, 0.0151, 0.0141, 0.0127, 0.0142, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:21:26,450 INFO [train.py:904] (3/8) Epoch 17, batch 1050, loss[loss=0.161, simple_loss=0.2526, pruned_loss=0.03468, over 17252.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2524, pruned_loss=0.04292, over 3304580.73 frames. ], batch size: 52, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:22:02,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-30 11:22:10,660 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:22:10,801 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5478, 2.2899, 2.3109, 4.3832, 2.2544, 2.7312, 2.4052, 2.4471], device='cuda:3'), covar=tensor([0.1103, 0.3530, 0.2835, 0.0442, 0.4012, 0.2410, 0.3397, 0.3461], device='cuda:3'), in_proj_covar=tensor([0.0385, 0.0424, 0.0356, 0.0325, 0.0430, 0.0488, 0.0395, 0.0495], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:22:36,965 INFO [train.py:904] (3/8) Epoch 17, batch 1100, loss[loss=0.1711, simple_loss=0.2442, pruned_loss=0.049, over 16551.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.252, pruned_loss=0.04251, over 3298428.41 frames. ], batch size: 75, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:22:46,174 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4513, 4.5071, 4.8325, 4.8236, 4.8600, 4.5423, 4.5582, 4.3805], device='cuda:3'), covar=tensor([0.0352, 0.0510, 0.0410, 0.0442, 0.0482, 0.0413, 0.0836, 0.0563], device='cuda:3'), in_proj_covar=tensor([0.0389, 0.0418, 0.0410, 0.0385, 0.0457, 0.0432, 0.0527, 0.0343], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 11:23:14,823 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:23:24,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4703, 2.7919, 3.0092, 2.1099, 2.6838, 2.1330, 3.1432, 3.0446], device='cuda:3'), covar=tensor([0.0282, 0.0927, 0.0562, 0.1815, 0.0878, 0.0953, 0.0566, 0.0981], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0156, 0.0163, 0.0151, 0.0142, 0.0127, 0.0142, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:23:35,786 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1182, 5.6548, 5.7965, 5.4908, 5.6106, 6.1574, 5.7107, 5.4281], device='cuda:3'), covar=tensor([0.0934, 0.2153, 0.2267, 0.2068, 0.2680, 0.1089, 0.1477, 0.2232], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0568, 0.0626, 0.0478, 0.0639, 0.0662, 0.0495, 0.0640], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:23:43,925 INFO [train.py:904] (3/8) Epoch 17, batch 1150, loss[loss=0.1453, simple_loss=0.2192, pruned_loss=0.03573, over 16794.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2508, pruned_loss=0.04168, over 3308059.44 frames. ], batch size: 83, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:24:34,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4729, 3.3564, 3.6536, 2.6069, 3.2306, 3.6799, 3.4528, 2.1767], device='cuda:3'), covar=tensor([0.0417, 0.0149, 0.0050, 0.0334, 0.0114, 0.0087, 0.0085, 0.0424], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0132, 0.0091, 0.0101, 0.0090, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:24:43,076 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:24:43,756 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 11:24:52,262 INFO [train.py:904] (3/8) Epoch 17, batch 1200, loss[loss=0.1536, simple_loss=0.251, pruned_loss=0.0281, over 17149.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2503, pruned_loss=0.04135, over 3303296.06 frames. ], batch size: 48, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:25:29,955 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 1250, loss[loss=0.1531, simple_loss=0.2393, pruned_loss=0.03339, over 17215.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2496, pruned_loss=0.04172, over 3308245.50 frames. ], batch size: 45, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:01,066 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0295, 3.9320, 4.2396, 2.2934, 4.4358, 4.5207, 3.3271, 3.5222], device='cuda:3'), covar=tensor([0.0694, 0.0215, 0.0228, 0.1076, 0.0078, 0.0153, 0.0394, 0.0362], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0108, 0.0095, 0.0140, 0.0076, 0.0121, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 11:27:06,152 INFO [train.py:904] (3/8) Epoch 17, batch 1300, loss[loss=0.2077, simple_loss=0.2719, pruned_loss=0.07168, over 16650.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2509, pruned_loss=0.04246, over 3310762.48 frames. ], batch size: 134, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:44,996 INFO [optim.py:368] (3/8) 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 11:28:08,437 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:28:16,451 INFO [train.py:904] (3/8) Epoch 17, batch 1350, loss[loss=0.141, simple_loss=0.2225, pruned_loss=0.02969, over 16841.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2513, pruned_loss=0.04224, over 3316482.91 frames. ], batch size: 42, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:28:54,839 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 11:29:15,148 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:29:25,094 INFO [train.py:904] (3/8) Epoch 17, batch 1400, loss[loss=0.1616, simple_loss=0.2422, pruned_loss=0.04046, over 16519.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2524, pruned_loss=0.04283, over 3325707.10 frames. ], batch size: 75, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:30:05,129 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 1450, loss[loss=0.158, simple_loss=0.2313, pruned_loss=0.04236, over 11652.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2513, pruned_loss=0.0427, over 3314115.15 frames. ], batch size: 247, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:31:25,327 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4291, 5.3104, 5.2862, 4.7342, 4.8463, 5.2656, 5.2452, 4.9134], device='cuda:3'), covar=tensor([0.0488, 0.0477, 0.0270, 0.0313, 0.1102, 0.0421, 0.0269, 0.0745], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0395, 0.0331, 0.0322, 0.0344, 0.0373, 0.0228, 0.0398], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:31:38,039 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:31:45,955 INFO [train.py:904] (3/8) Epoch 17, batch 1500, loss[loss=0.1896, simple_loss=0.2703, pruned_loss=0.05443, over 16891.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2513, pruned_loss=0.04266, over 3310637.81 frames. ], batch size: 96, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:32:16,890 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7721, 1.7969, 1.5942, 1.5123, 1.9708, 1.6658, 1.7003, 1.9503], device='cuda:3'), covar=tensor([0.0195, 0.0296, 0.0372, 0.0320, 0.0199, 0.0239, 0.0198, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0230, 0.0220, 0.0219, 0.0228, 0.0229, 0.0234, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:32:24,619 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 2.188e+02 2.579e+02 3.200e+02 4.820e+02, threshold=5.157e+02, percent-clipped=0.0 2023-04-30 11:32:45,102 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:32:48,678 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7693, 1.7877, 1.5988, 1.4217, 1.9715, 1.6009, 1.6524, 1.9269], device='cuda:3'), covar=tensor([0.0156, 0.0260, 0.0337, 0.0318, 0.0172, 0.0219, 0.0163, 0.0166], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0231, 0.0221, 0.0220, 0.0229, 0.0230, 0.0235, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:32:56,203 INFO [train.py:904] (3/8) Epoch 17, batch 1550, loss[loss=0.1584, simple_loss=0.2384, pruned_loss=0.03919, over 16317.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2529, pruned_loss=0.04403, over 3302143.87 frames. ], batch size: 36, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:33:01,513 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 11:34:07,075 INFO [train.py:904] (3/8) Epoch 17, batch 1600, loss[loss=0.1855, simple_loss=0.2749, pruned_loss=0.04801, over 16715.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2547, pruned_loss=0.04437, over 3309577.99 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:34:44,777 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-30 11:34:45,103 INFO [optim.py:368] (3/8) 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 11:35:15,604 INFO [train.py:904] (3/8) Epoch 17, batch 1650, loss[loss=0.1597, simple_loss=0.2512, pruned_loss=0.03408, over 17231.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2554, pruned_loss=0.04437, over 3321765.94 frames. ], batch size: 44, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:08,233 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:36:24,910 INFO [train.py:904] (3/8) Epoch 17, batch 1700, loss[loss=0.1767, simple_loss=0.2666, pruned_loss=0.04341, over 16476.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2578, pruned_loss=0.04511, over 3316284.53 frames. ], batch size: 68, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:36,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7492, 3.9379, 2.3831, 4.5029, 2.8860, 4.4867, 2.5320, 3.1520], device='cuda:3'), covar=tensor([0.0289, 0.0382, 0.1590, 0.0271, 0.0880, 0.0458, 0.1499, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0156, 0.0174, 0.0217, 0.0203, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:37:01,949 INFO [optim.py:368] (3/8) 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 11:37:05,375 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8927, 2.5211, 1.9521, 2.2357, 2.9698, 2.6874, 3.0167, 3.0426], device='cuda:3'), covar=tensor([0.0178, 0.0359, 0.0494, 0.0422, 0.0205, 0.0280, 0.0199, 0.0221], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0231, 0.0223, 0.0222, 0.0231, 0.0232, 0.0237, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:37:07,979 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 11:37:32,247 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 1750, loss[loss=0.1774, simple_loss=0.2771, pruned_loss=0.03882, over 16634.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2583, pruned_loss=0.04453, over 3314228.10 frames. ], batch size: 57, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:37:36,815 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:37:50,540 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7729, 2.8770, 2.5970, 4.2761, 3.6091, 4.2260, 1.6902, 2.9564], device='cuda:3'), covar=tensor([0.1323, 0.0639, 0.1107, 0.0166, 0.0151, 0.0351, 0.1490, 0.0797], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0172, 0.0198, 0.0211, 0.0191, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:38:41,940 INFO [train.py:904] (3/8) Epoch 17, batch 1800, loss[loss=0.1788, simple_loss=0.2666, pruned_loss=0.04548, over 15900.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.26, pruned_loss=0.04489, over 3302865.99 frames. ], batch size: 35, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:39:01,666 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:39:09,200 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 11:39:19,765 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.179e+02 2.567e+02 3.142e+02 4.810e+02, threshold=5.135e+02, percent-clipped=0.0 2023-04-30 11:39:49,954 INFO [train.py:904] (3/8) Epoch 17, batch 1850, loss[loss=0.1922, simple_loss=0.2813, pruned_loss=0.05157, over 11865.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2607, pruned_loss=0.04529, over 3295277.29 frames. ], batch size: 246, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:40:56,917 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-30 11:41:01,367 INFO [train.py:904] (3/8) Epoch 17, batch 1900, loss[loss=0.1491, simple_loss=0.2348, pruned_loss=0.03172, over 16780.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2592, pruned_loss=0.04422, over 3293276.72 frames. ], batch size: 39, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:41,203 INFO [optim.py:368] (3/8) 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 11:42:12,314 INFO [train.py:904] (3/8) Epoch 17, batch 1950, loss[loss=0.1895, simple_loss=0.2819, pruned_loss=0.0486, over 16625.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.26, pruned_loss=0.04396, over 3307844.25 frames. ], batch size: 62, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:42:24,694 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:43:23,961 INFO [train.py:904] (3/8) Epoch 17, batch 2000, loss[loss=0.1746, simple_loss=0.2543, pruned_loss=0.04748, over 16748.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2602, pruned_loss=0.04356, over 3315454.30 frames. ], batch size: 83, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:43:30,405 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-30 11:43:51,043 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 11:44:23,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2942, 5.2684, 5.0773, 4.5594, 5.0920, 1.9824, 4.8799, 5.0424], device='cuda:3'), covar=tensor([0.0073, 0.0072, 0.0165, 0.0328, 0.0092, 0.2544, 0.0125, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0143, 0.0191, 0.0172, 0.0164, 0.0202, 0.0179, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:44:25,523 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:26,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6536, 3.9919, 4.2961, 1.9258, 4.4833, 4.6821, 3.3779, 3.2605], device='cuda:3'), covar=tensor([0.1075, 0.0176, 0.0194, 0.1377, 0.0084, 0.0158, 0.0378, 0.0541], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0094, 0.0139, 0.0075, 0.0121, 0.0127, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 11:44:32,478 INFO [train.py:904] (3/8) Epoch 17, batch 2050, loss[loss=0.1594, simple_loss=0.2564, pruned_loss=0.03115, over 17107.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2603, pruned_loss=0.0435, over 3304279.24 frames. ], batch size: 47, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:44:59,579 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1116, 5.5098, 5.7262, 5.3957, 5.4543, 6.0885, 5.5832, 5.3581], device='cuda:3'), covar=tensor([0.0917, 0.1855, 0.2021, 0.1946, 0.2650, 0.0852, 0.1551, 0.2258], device='cuda:3'), in_proj_covar=tensor([0.0396, 0.0568, 0.0627, 0.0481, 0.0647, 0.0663, 0.0497, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:45:41,559 INFO [train.py:904] (3/8) Epoch 17, batch 2100, loss[loss=0.1866, simple_loss=0.2763, pruned_loss=0.04845, over 11986.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2611, pruned_loss=0.04452, over 3304718.07 frames. ], batch size: 246, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:45:54,964 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:46:20,646 INFO [optim.py:368] (3/8) 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 11:46:26,970 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8382, 4.1209, 2.9883, 2.2596, 2.7352, 2.5225, 4.4440, 3.6011], device='cuda:3'), covar=tensor([0.2686, 0.0618, 0.1714, 0.2725, 0.2661, 0.1912, 0.0384, 0.1272], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0264, 0.0298, 0.0298, 0.0289, 0.0243, 0.0283, 0.0325], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:46:50,964 INFO [train.py:904] (3/8) Epoch 17, batch 2150, loss[loss=0.1814, simple_loss=0.2715, pruned_loss=0.04567, over 17114.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2624, pruned_loss=0.04595, over 3303177.38 frames. ], batch size: 47, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:47:13,604 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:27,034 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:58,403 INFO [train.py:904] (3/8) Epoch 17, batch 2200, loss[loss=0.2217, simple_loss=0.298, pruned_loss=0.07264, over 12474.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2624, pruned_loss=0.0462, over 3304559.41 frames. ], batch size: 248, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:48:36,622 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 11:48:44,214 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3302, 3.4970, 3.7969, 2.6581, 3.3751, 3.8223, 3.5704, 2.1450], device='cuda:3'), covar=tensor([0.0480, 0.0161, 0.0045, 0.0333, 0.0097, 0.0080, 0.0072, 0.0430], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0077, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:48:44,452 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-04-30 11:48:50,046 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 11:48:51,347 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 11:48:58,201 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0565, 3.9849, 4.4416, 2.1754, 4.6424, 4.6710, 3.2671, 3.6107], device='cuda:3'), covar=tensor([0.0670, 0.0236, 0.0181, 0.1148, 0.0060, 0.0156, 0.0447, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0095, 0.0139, 0.0075, 0.0122, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 11:49:02,799 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 11:49:06,804 INFO [train.py:904] (3/8) Epoch 17, batch 2250, loss[loss=0.1505, simple_loss=0.2404, pruned_loss=0.03029, over 17051.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2625, pruned_loss=0.0461, over 3310322.94 frames. ], batch size: 41, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:49:15,020 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 2300, loss[loss=0.1669, simple_loss=0.2524, pruned_loss=0.04068, over 17169.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.262, pruned_loss=0.04577, over 3312374.02 frames. ], batch size: 46, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:50:34,571 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:38,872 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:53,240 INFO [optim.py:368] (3/8) 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 11:51:16,722 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:23,968 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1983, 2.1410, 2.2773, 3.9678, 2.1155, 2.4980, 2.2013, 2.2964], device='cuda:3'), covar=tensor([0.1291, 0.3673, 0.2856, 0.0571, 0.3908, 0.2454, 0.3859, 0.2917], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0429, 0.0359, 0.0327, 0.0431, 0.0496, 0.0398, 0.0501], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:51:24,576 INFO [train.py:904] (3/8) Epoch 17, batch 2350, loss[loss=0.2014, simple_loss=0.2723, pruned_loss=0.06519, over 16845.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2613, pruned_loss=0.046, over 3313453.30 frames. ], batch size: 116, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:23,521 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:24,240 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 11:52:34,382 INFO [train.py:904] (3/8) Epoch 17, batch 2400, loss[loss=0.1731, simple_loss=0.2694, pruned_loss=0.03842, over 16718.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2618, pruned_loss=0.04585, over 3315198.34 frames. ], batch size: 62, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:41,599 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:46,695 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:53:12,673 INFO [optim.py:368] (3/8) 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 11:53:15,525 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7707, 2.8024, 2.6041, 4.5941, 3.6918, 4.0786, 1.6680, 3.0394], device='cuda:3'), covar=tensor([0.1325, 0.0740, 0.1142, 0.0201, 0.0266, 0.0458, 0.1524, 0.0820], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0164, 0.0185, 0.0172, 0.0198, 0.0210, 0.0190, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:53:19,544 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2990, 3.3723, 2.0209, 3.5213, 2.6996, 3.5451, 2.0795, 2.7032], device='cuda:3'), covar=tensor([0.0289, 0.0432, 0.1481, 0.0308, 0.0716, 0.0760, 0.1434, 0.0728], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0157, 0.0173, 0.0217, 0.0202, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:53:41,596 INFO [train.py:904] (3/8) Epoch 17, batch 2450, loss[loss=0.1812, simple_loss=0.255, pruned_loss=0.05369, over 16813.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2637, pruned_loss=0.04619, over 3318403.07 frames. ], batch size: 124, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:53:47,266 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5920, 4.9092, 4.6834, 4.6853, 4.4455, 4.4196, 4.3583, 4.9709], device='cuda:3'), covar=tensor([0.1106, 0.0813, 0.0953, 0.0763, 0.0790, 0.1226, 0.1058, 0.0871], device='cuda:3'), in_proj_covar=tensor([0.0647, 0.0797, 0.0649, 0.0593, 0.0506, 0.0509, 0.0662, 0.0618], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:53:51,212 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 2500, loss[loss=0.169, simple_loss=0.2625, pruned_loss=0.03773, over 17070.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2636, pruned_loss=0.04567, over 3327076.88 frames. ], batch size: 53, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:55:12,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2147, 2.1844, 2.3412, 3.9254, 2.2222, 2.5029, 2.2555, 2.3296], device='cuda:3'), covar=tensor([0.1317, 0.3478, 0.2549, 0.0562, 0.3571, 0.2431, 0.3383, 0.2980], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0427, 0.0357, 0.0326, 0.0429, 0.0494, 0.0396, 0.0500], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:55:17,566 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:55:26,645 INFO [optim.py:368] (3/8) 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 11:55:29,647 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 11:55:30,896 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 11:55:55,231 INFO [train.py:904] (3/8) Epoch 17, batch 2550, loss[loss=0.1784, simple_loss=0.2593, pruned_loss=0.04873, over 16859.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.264, pruned_loss=0.04573, over 3309281.18 frames. ], batch size: 83, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:02,087 INFO [train.py:904] (3/8) Epoch 17, batch 2600, loss[loss=0.1738, simple_loss=0.2655, pruned_loss=0.0411, over 17200.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2638, pruned_loss=0.04566, over 3301447.57 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:17,257 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:20,592 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:31,724 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8702, 2.9442, 2.6595, 4.6062, 3.7472, 4.2252, 1.7232, 3.1426], device='cuda:3'), covar=tensor([0.1262, 0.0683, 0.1143, 0.0161, 0.0266, 0.0394, 0.1490, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0174, 0.0199, 0.0212, 0.0192, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 11:57:41,409 INFO [optim.py:368] (3/8) 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] (3/8) attn_weights_entropy = tensor([4.8476, 5.2270, 4.9336, 4.9874, 4.7108, 4.6666, 4.6648, 5.3147], device='cuda:3'), covar=tensor([0.1315, 0.0865, 0.1095, 0.0831, 0.0909, 0.1169, 0.1203, 0.0948], device='cuda:3'), in_proj_covar=tensor([0.0647, 0.0800, 0.0651, 0.0594, 0.0508, 0.0509, 0.0663, 0.0618], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 11:58:08,461 INFO [train.py:904] (3/8) Epoch 17, batch 2650, loss[loss=0.1798, simple_loss=0.2655, pruned_loss=0.04704, over 16871.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2651, pruned_loss=0.04616, over 3298019.63 frames. ], batch size: 96, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:58:20,398 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9570, 3.8584, 4.3650, 1.9093, 4.6096, 4.6198, 3.1817, 3.5789], device='cuda:3'), covar=tensor([0.0720, 0.0227, 0.0197, 0.1273, 0.0052, 0.0149, 0.0451, 0.0356], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0138, 0.0075, 0.0121, 0.0127, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 11:58:26,259 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:58:27,812 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7826, 4.0896, 2.8730, 2.3248, 2.7546, 2.5618, 4.4245, 3.5490], device='cuda:3'), covar=tensor([0.2714, 0.0571, 0.1816, 0.2710, 0.2762, 0.1919, 0.0389, 0.1219], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0264, 0.0298, 0.0300, 0.0290, 0.0244, 0.0283, 0.0325], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:58:57,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2296, 4.6667, 4.5863, 3.5749, 3.9567, 4.5749, 4.1751, 2.9478], device='cuda:3'), covar=tensor([0.0407, 0.0062, 0.0041, 0.0281, 0.0091, 0.0088, 0.0069, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0076, 0.0076, 0.0129, 0.0090, 0.0099, 0.0088, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 11:59:18,027 INFO [train.py:904] (3/8) Epoch 17, batch 2700, loss[loss=0.1852, simple_loss=0.2766, pruned_loss=0.04689, over 17087.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2645, pruned_loss=0.04523, over 3311985.20 frames. ], batch size: 55, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:59:18,300 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:21,057 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 11:59:48,022 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:55,454 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7280, 3.9955, 4.3380, 1.9857, 4.5530, 4.7124, 3.3295, 3.5736], device='cuda:3'), covar=tensor([0.0973, 0.0189, 0.0202, 0.1313, 0.0080, 0.0161, 0.0401, 0.0412], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0095, 0.0139, 0.0075, 0.0122, 0.0127, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 11:59:59,036 INFO [optim.py:368] (3/8) 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 12:00:28,481 INFO [train.py:904] (3/8) Epoch 17, batch 2750, loss[loss=0.214, simple_loss=0.2863, pruned_loss=0.07079, over 12266.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2646, pruned_loss=0.04466, over 3312818.52 frames. ], batch size: 246, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:01:13,237 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 2800, loss[loss=0.1917, simple_loss=0.2803, pruned_loss=0.05153, over 16504.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2634, pruned_loss=0.04411, over 3315483.67 frames. ], batch size: 75, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:02:10,305 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:02:24,092 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:02:48,648 INFO [train.py:904] (3/8) Epoch 17, batch 2850, loss[loss=0.1584, simple_loss=0.2453, pruned_loss=0.03574, over 17211.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.263, pruned_loss=0.04406, over 3321049.67 frames. ], batch size: 44, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:03:18,270 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:18,378 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8096, 3.7560, 3.8723, 3.9835, 4.0462, 3.6329, 3.8656, 4.0852], device='cuda:3'), covar=tensor([0.1435, 0.1074, 0.1217, 0.0684, 0.0648, 0.2061, 0.1792, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0633, 0.0779, 0.0930, 0.0789, 0.0593, 0.0629, 0.0633, 0.0738], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:03:31,431 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:33,870 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7782, 4.9002, 5.0384, 4.8493, 4.8596, 5.4907, 4.9396, 4.6508], device='cuda:3'), covar=tensor([0.1275, 0.2112, 0.2312, 0.1985, 0.2683, 0.1057, 0.1707, 0.2670], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0572, 0.0631, 0.0485, 0.0650, 0.0666, 0.0500, 0.0650], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 12:03:47,563 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2533, 5.8226, 5.9497, 5.6967, 5.7993, 6.2806, 5.7531, 5.5510], device='cuda:3'), covar=tensor([0.0782, 0.1907, 0.2062, 0.1806, 0.2330, 0.0983, 0.1486, 0.2242], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0571, 0.0631, 0.0484, 0.0649, 0.0665, 0.0499, 0.0649], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 12:03:58,711 INFO [train.py:904] (3/8) Epoch 17, batch 2900, loss[loss=0.1652, simple_loss=0.2526, pruned_loss=0.03892, over 16656.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2609, pruned_loss=0.04408, over 3326188.51 frames. ], batch size: 62, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:04:16,974 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:04:39,534 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.311e+02 2.666e+02 3.214e+02 5.469e+02, threshold=5.331e+02, percent-clipped=1.0 2023-04-30 12:05:09,243 INFO [train.py:904] (3/8) Epoch 17, batch 2950, loss[loss=0.213, simple_loss=0.2878, pruned_loss=0.0691, over 16388.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.26, pruned_loss=0.04464, over 3324509.45 frames. ], batch size: 146, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:05:19,678 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0414, 5.4559, 5.6203, 5.3251, 5.3808, 6.0084, 5.4833, 5.2031], device='cuda:3'), covar=tensor([0.1030, 0.1954, 0.2014, 0.2240, 0.2839, 0.0970, 0.1387, 0.2491], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0575, 0.0636, 0.0489, 0.0654, 0.0671, 0.0502, 0.0654], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 12:05:23,902 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:20,614 INFO [train.py:904] (3/8) Epoch 17, batch 3000, loss[loss=0.1707, simple_loss=0.2647, pruned_loss=0.03837, over 17032.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2612, pruned_loss=0.04555, over 3320241.69 frames. ], batch size: 55, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:06:20,614 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 12:06:29,131 INFO [train.py:938] (3/8) Epoch 17, validation: loss=0.1364, simple_loss=0.2422, pruned_loss=0.0153, over 944034.00 frames. 2023-04-30 12:06:29,131 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 12:06:29,480 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:07:09,773 INFO [optim.py:368] (3/8) 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 12:07:23,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9671, 3.1446, 3.1430, 2.1688, 2.7933, 2.2072, 3.4608, 3.4212], device='cuda:3'), covar=tensor([0.0243, 0.0860, 0.0628, 0.1769, 0.0839, 0.0990, 0.0579, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0158, 0.0163, 0.0149, 0.0140, 0.0126, 0.0140, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 12:07:36,652 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3050, loss[loss=0.1899, simple_loss=0.2666, pruned_loss=0.05654, over 16728.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.262, pruned_loss=0.04595, over 3314137.02 frames. ], batch size: 134, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:08:12,779 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.56 vs. limit=2.0 2023-04-30 12:08:15,873 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3100, loss[loss=0.1759, simple_loss=0.2832, pruned_loss=0.03434, over 17269.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2616, pruned_loss=0.04594, over 3301751.55 frames. ], batch size: 52, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:09:27,660 INFO [optim.py:368] (3/8) 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 12:09:54,959 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3150, loss[loss=0.1853, simple_loss=0.2765, pruned_loss=0.04704, over 17026.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2602, pruned_loss=0.04534, over 3305078.71 frames. ], batch size: 55, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:10:17,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 12:11:06,274 INFO [train.py:904] (3/8) Epoch 17, batch 3200, loss[loss=0.1783, simple_loss=0.27, pruned_loss=0.04337, over 16827.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.259, pruned_loss=0.04477, over 3310750.97 frames. ], batch size: 102, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:11:21,048 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:11:23,482 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2369, 2.0362, 1.7178, 1.8375, 2.3089, 2.1123, 2.2815, 2.4456], device='cuda:3'), covar=tensor([0.0208, 0.0302, 0.0388, 0.0349, 0.0186, 0.0268, 0.0173, 0.0203], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0231, 0.0221, 0.0222, 0.0233, 0.0232, 0.0238, 0.0228], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:11:49,643 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.164e+02 2.494e+02 3.096e+02 4.677e+02, threshold=4.988e+02, percent-clipped=0.0 2023-04-30 12:12:15,438 INFO [train.py:904] (3/8) Epoch 17, batch 3250, loss[loss=0.1843, simple_loss=0.2728, pruned_loss=0.04783, over 17099.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.259, pruned_loss=0.04497, over 3309786.59 frames. ], batch size: 53, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:12:49,775 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 12:13:23,327 INFO [train.py:904] (3/8) Epoch 17, batch 3300, loss[loss=0.1471, simple_loss=0.2361, pruned_loss=0.02905, over 17209.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2593, pruned_loss=0.04462, over 3322881.13 frames. ], batch size: 44, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:14:06,777 INFO [optim.py:368] (3/8) 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 12:14:30,178 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3350, loss[loss=0.186, simple_loss=0.2687, pruned_loss=0.05167, over 16406.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2603, pruned_loss=0.04439, over 3328710.04 frames. ], batch size: 146, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:08,288 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-30 12:15:11,013 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3400, loss[loss=0.1698, simple_loss=0.2517, pruned_loss=0.04398, over 16807.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2605, pruned_loss=0.04449, over 3320942.67 frames. ], batch size: 102, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:44,583 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:56,243 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:27,736 INFO [optim.py:368] (3/8) 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 12:16:54,368 INFO [train.py:904] (3/8) Epoch 17, batch 3450, loss[loss=0.1853, simple_loss=0.2711, pruned_loss=0.04969, over 15470.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2583, pruned_loss=0.04393, over 3327515.99 frames. ], batch size: 191, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:17:11,515 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3500, loss[loss=0.171, simple_loss=0.2652, pruned_loss=0.03836, over 16528.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2579, pruned_loss=0.04377, over 3318324.52 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:18:13,077 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:18:45,787 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-30 12:18:49,909 INFO [optim.py:368] (3/8) 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 12:19:16,034 INFO [train.py:904] (3/8) Epoch 17, batch 3550, loss[loss=0.1867, simple_loss=0.262, pruned_loss=0.05571, over 16748.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2571, pruned_loss=0.04345, over 3322731.64 frames. ], batch size: 124, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:19:59,851 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1289, 2.1057, 2.2566, 3.6531, 2.1566, 2.4009, 2.2315, 2.2352], device='cuda:3'), covar=tensor([0.1313, 0.3465, 0.2679, 0.0662, 0.3693, 0.2547, 0.3324, 0.3124], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0429, 0.0357, 0.0327, 0.0430, 0.0498, 0.0399, 0.0503], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:20:27,926 INFO [train.py:904] (3/8) Epoch 17, batch 3600, loss[loss=0.1557, simple_loss=0.2455, pruned_loss=0.03293, over 17264.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2574, pruned_loss=0.04346, over 3331291.98 frames. ], batch size: 45, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:20:58,438 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8400, 3.7702, 4.0792, 2.1089, 4.3000, 4.2408, 3.2714, 3.2835], device='cuda:3'), covar=tensor([0.0718, 0.0219, 0.0180, 0.1124, 0.0083, 0.0199, 0.0383, 0.0385], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0138, 0.0076, 0.0123, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 12:21:12,106 INFO [optim.py:368] (3/8) 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 12:21:26,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1772, 3.4895, 3.5719, 3.5276, 3.5273, 3.4096, 3.4064, 3.4577], device='cuda:3'), covar=tensor([0.0443, 0.0582, 0.0434, 0.0475, 0.0586, 0.0497, 0.0736, 0.0531], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0428, 0.0417, 0.0394, 0.0467, 0.0441, 0.0539, 0.0349], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 12:21:40,299 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3714, 4.3440, 4.3613, 3.3126, 4.3835, 1.6818, 4.0257, 3.8355], device='cuda:3'), covar=tensor([0.0194, 0.0177, 0.0210, 0.0678, 0.0132, 0.3466, 0.0229, 0.0328], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0143, 0.0190, 0.0174, 0.0164, 0.0198, 0.0179, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:21:40,964 INFO [train.py:904] (3/8) Epoch 17, batch 3650, loss[loss=0.1962, simple_loss=0.268, pruned_loss=0.06225, over 16794.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2563, pruned_loss=0.04412, over 3316800.80 frames. ], batch size: 124, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:21:44,593 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:21:50,988 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-30 12:22:05,591 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1291, 5.2071, 5.5943, 5.5566, 5.5925, 5.2632, 5.1939, 4.9678], device='cuda:3'), covar=tensor([0.0335, 0.0557, 0.0410, 0.0451, 0.0562, 0.0361, 0.0879, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0426, 0.0415, 0.0393, 0.0465, 0.0439, 0.0535, 0.0347], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 12:22:55,839 INFO [train.py:904] (3/8) Epoch 17, batch 3700, loss[loss=0.1871, simple_loss=0.2581, pruned_loss=0.058, over 16877.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2551, pruned_loss=0.04531, over 3281082.71 frames. ], batch size: 109, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:23:02,183 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:23:15,998 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 12:23:21,122 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3984, 3.5528, 3.7992, 2.7067, 3.4140, 3.8726, 3.6680, 2.2190], device='cuda:3'), covar=tensor([0.0442, 0.0084, 0.0044, 0.0335, 0.0093, 0.0080, 0.0063, 0.0444], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0077, 0.0077, 0.0131, 0.0091, 0.0101, 0.0088, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 12:23:42,388 INFO [optim.py:368] (3/8) 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 12:24:03,683 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:07,683 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9307, 3.1462, 3.1758, 2.1748, 2.9462, 3.2774, 3.0749, 1.9314], device='cuda:3'), covar=tensor([0.0476, 0.0100, 0.0063, 0.0365, 0.0114, 0.0095, 0.0086, 0.0410], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0076, 0.0076, 0.0130, 0.0090, 0.0100, 0.0088, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 12:24:10,330 INFO [train.py:904] (3/8) Epoch 17, batch 3750, loss[loss=0.1652, simple_loss=0.2556, pruned_loss=0.03743, over 16538.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2553, pruned_loss=0.04653, over 3259540.27 frames. ], batch size: 62, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:24:21,420 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:47,316 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 12:25:14,890 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 12:25:24,422 INFO [train.py:904] (3/8) Epoch 17, batch 3800, loss[loss=0.2054, simple_loss=0.29, pruned_loss=0.06033, over 12077.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2569, pruned_loss=0.04773, over 3251208.27 frames. ], batch size: 246, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:25:32,388 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:33,726 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:26:10,618 INFO [optim.py:368] (3/8) 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 12:26:38,505 INFO [train.py:904] (3/8) Epoch 17, batch 3850, loss[loss=0.1855, simple_loss=0.259, pruned_loss=0.05603, over 16900.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2575, pruned_loss=0.04851, over 3244603.11 frames. ], batch size: 109, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:26:44,003 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 3900, loss[loss=0.1792, simple_loss=0.2598, pruned_loss=0.04936, over 16506.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2574, pruned_loss=0.04929, over 3247555.57 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:28:01,059 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:28:37,861 INFO [optim.py:368] (3/8) 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 12:29:07,065 INFO [train.py:904] (3/8) Epoch 17, batch 3950, loss[loss=0.1672, simple_loss=0.24, pruned_loss=0.04723, over 16858.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2572, pruned_loss=0.04982, over 3252171.90 frames. ], batch size: 116, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:29:30,577 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:18,532 INFO [train.py:904] (3/8) Epoch 17, batch 4000, loss[loss=0.163, simple_loss=0.2517, pruned_loss=0.03715, over 16643.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2566, pruned_loss=0.05, over 3268936.61 frames. ], batch size: 89, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:30:25,169 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:31,137 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:31:03,720 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 4050, loss[loss=0.1673, simple_loss=0.2497, pruned_loss=0.04247, over 16311.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2569, pruned_loss=0.04907, over 3274197.73 frames. ], batch size: 35, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:31:34,056 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:42,521 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:46,059 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2555, 3.2506, 3.5501, 1.6337, 3.7206, 3.7375, 2.9055, 2.7083], device='cuda:3'), covar=tensor([0.0886, 0.0291, 0.0182, 0.1381, 0.0083, 0.0127, 0.0415, 0.0505], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0108, 0.0095, 0.0139, 0.0077, 0.0123, 0.0128, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 12:32:14,542 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:44,112 INFO [train.py:904] (3/8) Epoch 17, batch 4100, loss[loss=0.1982, simple_loss=0.2951, pruned_loss=0.0507, over 16896.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.259, pruned_loss=0.04856, over 3272556.45 frames. ], batch size: 96, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:32:46,281 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:54,571 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:29,237 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:33:45,767 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:59,690 INFO [train.py:904] (3/8) Epoch 17, batch 4150, loss[loss=0.2284, simple_loss=0.3118, pruned_loss=0.07248, over 15351.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2663, pruned_loss=0.05129, over 3236847.46 frames. ], batch size: 191, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:34:28,444 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:01,099 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:14,004 INFO [train.py:904] (3/8) Epoch 17, batch 4200, loss[loss=0.2138, simple_loss=0.3032, pruned_loss=0.06225, over 17112.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2735, pruned_loss=0.05361, over 3179737.71 frames. ], batch size: 48, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:00,016 INFO [optim.py:368] (3/8) 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 12:36:13,737 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 12:36:27,722 INFO [train.py:904] (3/8) Epoch 17, batch 4250, loss[loss=0.1996, simple_loss=0.2876, pruned_loss=0.05577, over 16698.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2774, pruned_loss=0.05368, over 3168086.21 frames. ], batch size: 57, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:43,547 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:36:49,760 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3698, 4.4915, 4.7314, 4.7007, 4.7333, 4.4353, 4.4147, 4.2402], device='cuda:3'), covar=tensor([0.0302, 0.0479, 0.0441, 0.0500, 0.0400, 0.0364, 0.0924, 0.0504], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0415, 0.0404, 0.0381, 0.0450, 0.0426, 0.0524, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 12:37:33,331 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0859, 2.0255, 1.6504, 1.7157, 2.2463, 1.8731, 2.0417, 2.3679], device='cuda:3'), covar=tensor([0.0185, 0.0364, 0.0482, 0.0414, 0.0218, 0.0326, 0.0199, 0.0227], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0225, 0.0216, 0.0218, 0.0227, 0.0226, 0.0230, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:37:39,001 INFO [train.py:904] (3/8) Epoch 17, batch 4300, loss[loss=0.2041, simple_loss=0.2899, pruned_loss=0.05918, over 11358.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2781, pruned_loss=0.05276, over 3149827.17 frames. ], batch size: 246, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:37:51,464 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 12:38:24,692 INFO [optim.py:368] (3/8) 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 12:38:28,894 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:38:52,930 INFO [train.py:904] (3/8) Epoch 17, batch 4350, loss[loss=0.1971, simple_loss=0.2839, pruned_loss=0.05514, over 17024.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2808, pruned_loss=0.05328, over 3148745.24 frames. ], batch size: 55, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:39:01,860 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:39:04,624 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5706, 2.1851, 1.6947, 2.0295, 2.5104, 2.1930, 2.4597, 2.7043], device='cuda:3'), covar=tensor([0.0157, 0.0368, 0.0526, 0.0435, 0.0223, 0.0338, 0.0215, 0.0215], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0222, 0.0215, 0.0215, 0.0224, 0.0224, 0.0227, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:39:10,383 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:39:57,551 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:02,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 12:40:05,343 INFO [train.py:904] (3/8) Epoch 17, batch 4400, loss[loss=0.1984, simple_loss=0.2876, pruned_loss=0.05464, over 17201.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2825, pruned_loss=0.05448, over 3143249.84 frames. ], batch size: 44, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:40:06,833 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:37,964 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:49,145 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:41:16,588 INFO [train.py:904] (3/8) Epoch 17, batch 4450, loss[loss=0.2338, simple_loss=0.3183, pruned_loss=0.07465, over 16697.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2864, pruned_loss=0.05584, over 3173940.43 frames. ], batch size: 62, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:41:36,424 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:07,491 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:28,857 INFO [train.py:904] (3/8) Epoch 17, batch 4500, loss[loss=0.1814, simple_loss=0.2717, pruned_loss=0.04553, over 16778.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2865, pruned_loss=0.05642, over 3176379.62 frames. ], batch size: 76, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:03,521 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-30 12:43:07,227 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:43:31,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0699, 1.7924, 2.5802, 2.9513, 2.7811, 3.2977, 2.0462, 3.3421], device='cuda:3'), covar=tensor([0.0173, 0.0463, 0.0267, 0.0226, 0.0250, 0.0186, 0.0459, 0.0121], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0176, 0.0186, 0.0143, 0.0187, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:43:40,951 INFO [train.py:904] (3/8) Epoch 17, batch 4550, loss[loss=0.1804, simple_loss=0.2741, pruned_loss=0.04336, over 16836.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2873, pruned_loss=0.05725, over 3206113.03 frames. ], batch size: 102, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:57,135 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:20,127 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4940, 3.5063, 3.4399, 2.7557, 3.2337, 2.1568, 3.0499, 2.6579], device='cuda:3'), covar=tensor([0.0110, 0.0092, 0.0159, 0.0196, 0.0084, 0.2147, 0.0111, 0.0205], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0141, 0.0187, 0.0172, 0.0162, 0.0197, 0.0177, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:44:35,601 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:53,171 INFO [train.py:904] (3/8) Epoch 17, batch 4600, loss[loss=0.2297, simple_loss=0.3002, pruned_loss=0.07961, over 11628.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.288, pruned_loss=0.05735, over 3205639.09 frames. ], batch size: 246, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:45:07,100 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:45:38,256 INFO [optim.py:368] (3/8) 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 12:46:05,394 INFO [train.py:904] (3/8) Epoch 17, batch 4650, loss[loss=0.1929, simple_loss=0.2778, pruned_loss=0.05397, over 16879.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2868, pruned_loss=0.05718, over 3204289.67 frames. ], batch size: 96, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:46:43,200 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8881, 3.2494, 3.2016, 2.0637, 3.0046, 3.2292, 3.0791, 1.8856], device='cuda:3'), covar=tensor([0.0484, 0.0048, 0.0047, 0.0399, 0.0084, 0.0110, 0.0083, 0.0405], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0129, 0.0090, 0.0099, 0.0088, 0.0122], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 12:46:58,517 INFO [scaling.py:679] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:13,887 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 4700, loss[loss=0.201, simple_loss=0.2835, pruned_loss=0.05928, over 16429.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2837, pruned_loss=0.05568, over 3227970.43 frames. ], batch size: 68, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:47:18,980 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 12:48:06,768 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 4750, loss[loss=0.1853, simple_loss=0.2764, pruned_loss=0.04714, over 15397.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2797, pruned_loss=0.05374, over 3221848.67 frames. ], batch size: 191, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:48:42,088 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:44,928 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5938, 2.5555, 2.4471, 3.8906, 2.5877, 3.8417, 1.4730, 2.7978], device='cuda:3'), covar=tensor([0.1400, 0.0796, 0.1285, 0.0146, 0.0205, 0.0401, 0.1700, 0.0863], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0175, 0.0203, 0.0211, 0.0192, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 12:48:48,027 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,114 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:16,407 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:18,952 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 4800, loss[loss=0.193, simple_loss=0.2831, pruned_loss=0.05139, over 15368.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2761, pruned_loss=0.05147, over 3219807.04 frames. ], batch size: 190, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:49:58,978 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:27,971 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.008e+02 2.170e+02 2.594e+02 4.937e+02, threshold=4.340e+02, percent-clipped=1.0 2023-04-30 12:50:32,465 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 4850, loss[loss=0.1832, simple_loss=0.279, pruned_loss=0.04369, over 15440.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2771, pruned_loss=0.05118, over 3186314.11 frames. ], batch size: 190, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:51:46,660 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:51:56,979 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 12:52:12,188 INFO [train.py:904] (3/8) Epoch 17, batch 4900, loss[loss=0.1797, simple_loss=0.2764, pruned_loss=0.04156, over 16676.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2765, pruned_loss=0.04992, over 3168062.03 frames. ], batch size: 89, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:52:31,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8679, 2.2179, 2.1700, 2.7414, 1.7624, 3.2493, 1.6947, 2.7136], device='cuda:3'), covar=tensor([0.1225, 0.0750, 0.1198, 0.0143, 0.0107, 0.0328, 0.1465, 0.0697], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0168, 0.0189, 0.0175, 0.0203, 0.0211, 0.0193, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 12:52:55,966 INFO [optim.py:368] (3/8) 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 12:53:24,979 INFO [train.py:904] (3/8) Epoch 17, batch 4950, loss[loss=0.2055, simple_loss=0.3016, pruned_loss=0.05468, over 16629.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2763, pruned_loss=0.0494, over 3174261.18 frames. ], batch size: 68, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:53:26,101 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1941, 1.5215, 1.9048, 2.1009, 2.2863, 2.3325, 1.7175, 2.2150], device='cuda:3'), covar=tensor([0.0194, 0.0436, 0.0251, 0.0290, 0.0259, 0.0171, 0.0445, 0.0130], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0177, 0.0187, 0.0143, 0.0188, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:54:16,304 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 12:54:22,610 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:54:37,755 INFO [train.py:904] (3/8) Epoch 17, batch 5000, loss[loss=0.1867, simple_loss=0.2763, pruned_loss=0.04855, over 16278.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2778, pruned_loss=0.04936, over 3185671.43 frames. ], batch size: 35, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:54:45,387 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5252, 2.5625, 2.0665, 2.4230, 2.9659, 2.6475, 3.2010, 3.1780], device='cuda:3'), covar=tensor([0.0076, 0.0384, 0.0529, 0.0384, 0.0246, 0.0342, 0.0169, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0223, 0.0216, 0.0216, 0.0225, 0.0224, 0.0226, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:55:02,780 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:19,942 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.024e+02 2.500e+02 2.980e+02 7.048e+02, threshold=5.000e+02, percent-clipped=1.0 2023-04-30 12:55:29,658 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:42,129 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-04-30 12:55:46,692 INFO [train.py:904] (3/8) Epoch 17, batch 5050, loss[loss=0.1773, simple_loss=0.2682, pruned_loss=0.04319, over 16525.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2782, pruned_loss=0.04924, over 3180589.51 frames. ], batch size: 75, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:51,603 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:57,231 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:08,483 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:43,566 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:56,777 INFO [train.py:904] (3/8) Epoch 17, batch 5100, loss[loss=0.1914, simple_loss=0.2782, pruned_loss=0.05227, over 16862.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2762, pruned_loss=0.04858, over 3179561.54 frames. ], batch size: 116, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:57:32,052 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2399, 2.9564, 3.2008, 1.8411, 3.2967, 3.3667, 2.7578, 2.5817], device='cuda:3'), covar=tensor([0.0754, 0.0236, 0.0127, 0.1075, 0.0063, 0.0143, 0.0370, 0.0471], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0106, 0.0092, 0.0137, 0.0075, 0.0119, 0.0126, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 12:57:39,789 INFO [optim.py:368] (3/8) 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 12:57:50,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4305, 1.5672, 2.0448, 2.3110, 2.3913, 2.6241, 1.7051, 2.5733], device='cuda:3'), covar=tensor([0.0164, 0.0535, 0.0282, 0.0305, 0.0278, 0.0156, 0.0501, 0.0113], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0188, 0.0174, 0.0178, 0.0188, 0.0144, 0.0189, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 12:58:08,923 INFO [train.py:904] (3/8) Epoch 17, batch 5150, loss[loss=0.2116, simple_loss=0.3019, pruned_loss=0.06068, over 16451.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2772, pruned_loss=0.04842, over 3182642.49 frames. ], batch size: 146, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:58:45,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8499, 3.9194, 2.1943, 4.6558, 3.0397, 4.6048, 2.5147, 3.1369], device='cuda:3'), covar=tensor([0.0265, 0.0359, 0.1898, 0.0199, 0.0842, 0.0412, 0.1578, 0.0803], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0169, 0.0188, 0.0147, 0.0168, 0.0207, 0.0195, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 12:58:56,749 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 5200, loss[loss=0.1603, simple_loss=0.2401, pruned_loss=0.04022, over 17044.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2749, pruned_loss=0.04754, over 3180975.53 frames. ], batch size: 55, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:59:21,932 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 12:59:40,943 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 13:00:07,259 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 17, batch 5250, loss[loss=0.1657, simple_loss=0.2577, pruned_loss=0.03687, over 16728.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2726, pruned_loss=0.04721, over 3192506.75 frames. ], batch size: 89, lr: 3.99e-03, grad_scale: 16.0 2023-04-30 13:01:48,348 INFO [train.py:904] (3/8) Epoch 17, batch 5300, loss[loss=0.1674, simple_loss=0.2529, pruned_loss=0.04094, over 16746.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2685, pruned_loss=0.0457, over 3202772.31 frames. ], batch size: 124, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:02:12,777 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:02:14,615 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1787, 4.1132, 4.0256, 3.2177, 4.0648, 1.6431, 3.8447, 3.6264], device='cuda:3'), covar=tensor([0.0101, 0.0113, 0.0188, 0.0436, 0.0109, 0.2869, 0.0162, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0139, 0.0184, 0.0170, 0.0160, 0.0195, 0.0174, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:02:35,320 INFO [optim.py:368] (3/8) 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 13:03:01,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8021, 3.9176, 4.1771, 4.1425, 4.1325, 3.9022, 3.7413, 3.8836], device='cuda:3'), covar=tensor([0.0500, 0.0903, 0.0534, 0.0567, 0.0733, 0.0545, 0.1381, 0.0654], device='cuda:3'), in_proj_covar=tensor([0.0376, 0.0411, 0.0402, 0.0377, 0.0448, 0.0421, 0.0521, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 13:03:01,788 INFO [train.py:904] (3/8) Epoch 17, batch 5350, loss[loss=0.1914, simple_loss=0.2849, pruned_loss=0.04891, over 16176.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2673, pruned_loss=0.04532, over 3192755.72 frames. ], batch size: 165, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:03:08,174 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:13,807 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:23,273 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:41,677 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:53,663 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:03,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3867, 4.4566, 4.2489, 3.9782, 3.9373, 4.3530, 4.0611, 4.0922], device='cuda:3'), covar=tensor([0.0619, 0.0579, 0.0299, 0.0267, 0.0842, 0.0521, 0.0654, 0.0577], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0386, 0.0327, 0.0315, 0.0336, 0.0367, 0.0223, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:04:14,426 INFO [train.py:904] (3/8) Epoch 17, batch 5400, loss[loss=0.1907, simple_loss=0.2839, pruned_loss=0.04876, over 16793.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2694, pruned_loss=0.04575, over 3211940.47 frames. ], batch size: 124, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:04:17,678 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:54,015 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:57,150 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8680, 2.7696, 2.7344, 2.0184, 2.5542, 2.7041, 2.5942, 1.9416], device='cuda:3'), covar=tensor([0.0393, 0.0061, 0.0062, 0.0315, 0.0108, 0.0092, 0.0103, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0075, 0.0076, 0.0129, 0.0090, 0.0099, 0.0088, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 13:05:02,680 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.037e+02 2.360e+02 2.749e+02 5.264e+02, threshold=4.720e+02, percent-clipped=1.0 2023-04-30 13:05:31,682 INFO [train.py:904] (3/8) Epoch 17, batch 5450, loss[loss=0.2118, simple_loss=0.2924, pruned_loss=0.06556, over 16924.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2732, pruned_loss=0.04784, over 3187494.45 frames. ], batch size: 116, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:05:32,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5184, 3.5124, 3.4975, 2.8055, 3.3901, 2.0849, 3.2083, 2.8711], device='cuda:3'), covar=tensor([0.0156, 0.0144, 0.0162, 0.0272, 0.0106, 0.2264, 0.0152, 0.0212], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0139, 0.0185, 0.0170, 0.0160, 0.0196, 0.0174, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:05:59,792 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4469, 1.6931, 2.1401, 2.3407, 2.4232, 2.7032, 1.8251, 2.6169], device='cuda:3'), covar=tensor([0.0186, 0.0452, 0.0268, 0.0318, 0.0286, 0.0167, 0.0429, 0.0127], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0187, 0.0174, 0.0178, 0.0187, 0.0143, 0.0188, 0.0138], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:06:48,591 INFO [train.py:904] (3/8) Epoch 17, batch 5500, loss[loss=0.278, simple_loss=0.3458, pruned_loss=0.1051, over 15366.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2807, pruned_loss=0.05262, over 3154183.89 frames. ], batch size: 190, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:06:51,282 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2327, 2.2250, 2.2011, 3.9108, 2.0607, 2.5945, 2.3298, 2.3707], device='cuda:3'), covar=tensor([0.1115, 0.3192, 0.2521, 0.0434, 0.3652, 0.2153, 0.3057, 0.2926], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0421, 0.0349, 0.0318, 0.0422, 0.0487, 0.0392, 0.0493], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:07:01,529 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 13:07:39,670 INFO [optim.py:368] (3/8) 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 13:08:02,150 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 13:08:05,266 INFO [train.py:904] (3/8) Epoch 17, batch 5550, loss[loss=0.2134, simple_loss=0.2976, pruned_loss=0.06464, over 16512.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2877, pruned_loss=0.05747, over 3135615.33 frames. ], batch size: 68, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:09:31,455 INFO [train.py:904] (3/8) Epoch 17, batch 5600, loss[loss=0.2209, simple_loss=0.3061, pruned_loss=0.06783, over 16727.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2926, pruned_loss=0.06127, over 3123616.75 frames. ], batch size: 76, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:10:27,345 INFO [optim.py:368] (3/8) 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 13:10:27,922 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:10:54,415 INFO [train.py:904] (3/8) Epoch 17, batch 5650, loss[loss=0.2004, simple_loss=0.2921, pruned_loss=0.05429, over 16894.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2978, pruned_loss=0.06558, over 3089065.68 frames. ], batch size: 96, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:11:24,155 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1276, 1.9671, 2.6018, 2.9995, 2.8803, 3.5337, 2.1055, 3.3606], device='cuda:3'), covar=tensor([0.0157, 0.0448, 0.0259, 0.0254, 0.0237, 0.0122, 0.0463, 0.0121], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0186, 0.0172, 0.0176, 0.0186, 0.0142, 0.0188, 0.0137], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:11:30,930 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:11:50,846 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:04,692 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 5700, loss[loss=0.1905, simple_loss=0.2771, pruned_loss=0.0519, over 16623.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2991, pruned_loss=0.06697, over 3076667.23 frames. ], batch size: 57, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:12:45,722 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:58,682 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4654, 4.4402, 4.8488, 4.8025, 4.8633, 4.5164, 4.5455, 4.3183], device='cuda:3'), covar=tensor([0.0336, 0.0594, 0.0369, 0.0441, 0.0383, 0.0340, 0.0889, 0.0532], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0412, 0.0402, 0.0378, 0.0448, 0.0422, 0.0521, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 13:13:04,293 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.185e+02 3.066e+02 3.948e+02 4.929e+02 1.585e+03, threshold=7.895e+02, percent-clipped=5.0 2023-04-30 13:13:06,079 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:31,681 INFO [train.py:904] (3/8) Epoch 17, batch 5750, loss[loss=0.1955, simple_loss=0.285, pruned_loss=0.05296, over 16794.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3013, pruned_loss=0.06821, over 3054129.63 frames. ], batch size: 83, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:14:50,309 INFO [train.py:904] (3/8) Epoch 17, batch 5800, loss[loss=0.1844, simple_loss=0.2839, pruned_loss=0.04252, over 16838.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.302, pruned_loss=0.06811, over 3045149.39 frames. ], batch size: 96, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:15:09,336 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 13:15:18,389 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9295, 3.8789, 4.0335, 4.1599, 4.2460, 3.8416, 4.2040, 4.2730], device='cuda:3'), covar=tensor([0.1642, 0.1132, 0.1297, 0.0665, 0.0591, 0.1740, 0.0711, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0588, 0.0725, 0.0865, 0.0739, 0.0553, 0.0587, 0.0591, 0.0691], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:15:41,161 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:15:43,237 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 5850, loss[loss=0.2002, simple_loss=0.292, pruned_loss=0.05416, over 16915.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2994, pruned_loss=0.06607, over 3054266.63 frames. ], batch size: 109, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:17:16,869 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9971, 3.9871, 4.1237, 4.2646, 4.3626, 3.9994, 4.2886, 4.3785], device='cuda:3'), covar=tensor([0.1770, 0.1059, 0.1388, 0.0656, 0.0521, 0.1416, 0.0706, 0.0603], device='cuda:3'), in_proj_covar=tensor([0.0595, 0.0732, 0.0872, 0.0746, 0.0558, 0.0592, 0.0596, 0.0697], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:17:19,305 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:17:22,306 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5614, 4.4012, 4.3202, 3.0097, 3.9631, 4.3991, 3.9838, 2.5498], device='cuda:3'), covar=tensor([0.0494, 0.0032, 0.0046, 0.0338, 0.0077, 0.0083, 0.0076, 0.0392], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0077, 0.0078, 0.0132, 0.0091, 0.0102, 0.0090, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 13:17:32,944 INFO [train.py:904] (3/8) Epoch 17, batch 5900, loss[loss=0.1929, simple_loss=0.2774, pruned_loss=0.05419, over 15430.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.299, pruned_loss=0.0661, over 3059761.60 frames. ], batch size: 191, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:17:34,490 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7215, 3.1379, 3.3640, 1.9935, 2.9425, 2.2048, 3.3752, 3.3135], device='cuda:3'), covar=tensor([0.0206, 0.0710, 0.0531, 0.1858, 0.0748, 0.0946, 0.0551, 0.0814], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0156, 0.0161, 0.0148, 0.0141, 0.0126, 0.0139, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 13:17:42,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5927, 3.5823, 3.9957, 1.8318, 4.1746, 4.1743, 3.1713, 3.1040], device='cuda:3'), covar=tensor([0.0774, 0.0208, 0.0163, 0.1211, 0.0055, 0.0115, 0.0332, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0104, 0.0092, 0.0136, 0.0074, 0.0118, 0.0124, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 13:18:28,194 INFO [optim.py:368] (3/8) 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 13:18:51,807 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7850, 3.7333, 3.8846, 3.9933, 4.0549, 3.6776, 4.0139, 4.0907], device='cuda:3'), covar=tensor([0.1513, 0.1083, 0.1183, 0.0616, 0.0617, 0.1754, 0.0779, 0.0679], device='cuda:3'), in_proj_covar=tensor([0.0590, 0.0727, 0.0864, 0.0742, 0.0555, 0.0589, 0.0593, 0.0692], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:18:52,642 INFO [train.py:904] (3/8) Epoch 17, batch 5950, loss[loss=0.2249, simple_loss=0.3211, pruned_loss=0.06437, over 15142.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2991, pruned_loss=0.06392, over 3086150.83 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:19:29,619 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:20:12,026 INFO [train.py:904] (3/8) Epoch 17, batch 6000, loss[loss=0.2387, simple_loss=0.3124, pruned_loss=0.08247, over 15432.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2985, pruned_loss=0.06375, over 3090406.80 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:20:12,026 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 13:20:21,981 INFO [train.py:938] (3/8) Epoch 17, validation: loss=0.1535, simple_loss=0.2667, pruned_loss=0.0202, over 944034.00 frames. 2023-04-30 13:20:21,981 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 13:20:22,745 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 13:20:53,933 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:20:54,037 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:21:03,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0858, 5.1007, 4.9322, 4.5661, 4.5535, 4.9900, 4.9069, 4.6965], device='cuda:3'), covar=tensor([0.0680, 0.0666, 0.0290, 0.0328, 0.1016, 0.0542, 0.0381, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0380, 0.0321, 0.0309, 0.0329, 0.0361, 0.0220, 0.0383], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:21:15,098 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-30 13:21:15,434 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 2.768e+02 3.388e+02 4.185e+02 7.743e+02, threshold=6.777e+02, percent-clipped=1.0 2023-04-30 13:21:24,612 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-30 13:21:39,917 INFO [train.py:904] (3/8) Epoch 17, batch 6050, loss[loss=0.2082, simple_loss=0.3034, pruned_loss=0.05646, over 16497.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2975, pruned_loss=0.06311, over 3093041.92 frames. ], batch size: 75, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:22:09,419 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:22:15,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3765, 3.3536, 3.4106, 3.5132, 3.5385, 3.2827, 3.5119, 3.5788], device='cuda:3'), covar=tensor([0.1210, 0.0878, 0.0978, 0.0573, 0.0596, 0.2633, 0.0980, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0582, 0.0719, 0.0855, 0.0734, 0.0550, 0.0583, 0.0588, 0.0685], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:22:59,843 INFO [train.py:904] (3/8) Epoch 17, batch 6100, loss[loss=0.2185, simple_loss=0.3048, pruned_loss=0.06613, over 16225.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2964, pruned_loss=0.06179, over 3111649.45 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:23:55,310 INFO [optim.py:368] (3/8) 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 13:24:03,968 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4256, 4.5282, 4.8908, 4.8542, 4.8511, 4.5376, 4.5183, 4.3969], device='cuda:3'), covar=tensor([0.0326, 0.0534, 0.0315, 0.0376, 0.0450, 0.0380, 0.0867, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0384, 0.0418, 0.0407, 0.0386, 0.0456, 0.0430, 0.0529, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 13:24:18,357 INFO [train.py:904] (3/8) Epoch 17, batch 6150, loss[loss=0.2045, simple_loss=0.2927, pruned_loss=0.05813, over 16248.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2945, pruned_loss=0.06158, over 3102217.30 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:25:17,263 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 6200, loss[loss=0.2144, simple_loss=0.301, pruned_loss=0.06392, over 15275.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2928, pruned_loss=0.06143, over 3107821.47 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:26:15,015 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-30 13:26:31,028 INFO [optim.py:368] (3/8) 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 13:26:53,707 INFO [train.py:904] (3/8) Epoch 17, batch 6250, loss[loss=0.2165, simple_loss=0.317, pruned_loss=0.05798, over 16872.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2926, pruned_loss=0.06099, over 3112418.98 frames. ], batch size: 116, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:27:53,796 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 6300, loss[loss=0.1897, simple_loss=0.2814, pruned_loss=0.04904, over 16562.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2921, pruned_loss=0.05988, over 3117047.74 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:28:25,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5151, 4.4876, 4.3784, 3.6895, 4.4428, 1.7467, 4.1971, 4.1176], device='cuda:3'), covar=tensor([0.0086, 0.0073, 0.0159, 0.0302, 0.0077, 0.2677, 0.0126, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0139, 0.0186, 0.0171, 0.0159, 0.0196, 0.0173, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:29:06,430 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:29:29,581 INFO [train.py:904] (3/8) Epoch 17, batch 6350, loss[loss=0.2092, simple_loss=0.2894, pruned_loss=0.06452, over 16547.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2926, pruned_loss=0.06102, over 3109734.70 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:29:31,988 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9304, 3.2018, 3.2117, 2.0438, 2.9381, 3.1430, 3.0305, 1.9941], device='cuda:3'), covar=tensor([0.0532, 0.0059, 0.0062, 0.0425, 0.0107, 0.0126, 0.0098, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0076, 0.0078, 0.0132, 0.0090, 0.0102, 0.0089, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 13:29:42,681 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1194, 5.1175, 4.9938, 4.5997, 4.5840, 5.0324, 4.9861, 4.7252], device='cuda:3'), covar=tensor([0.0575, 0.0387, 0.0270, 0.0298, 0.1046, 0.0390, 0.0274, 0.0682], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0385, 0.0324, 0.0312, 0.0332, 0.0364, 0.0221, 0.0388], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:30:46,692 INFO [train.py:904] (3/8) Epoch 17, batch 6400, loss[loss=0.2662, simple_loss=0.3322, pruned_loss=0.1001, over 11286.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2938, pruned_loss=0.06332, over 3073001.14 frames. ], batch size: 247, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:31:29,475 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0257, 5.6836, 5.9034, 5.5167, 5.5886, 6.2186, 5.6783, 5.4626], device='cuda:3'), covar=tensor([0.0904, 0.1771, 0.2252, 0.1992, 0.2613, 0.0844, 0.1511, 0.2216], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0471, 0.0633, 0.0648, 0.0488, 0.0633], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 13:31:42,103 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 6450, loss[loss=0.1769, simple_loss=0.2736, pruned_loss=0.04013, over 17033.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2944, pruned_loss=0.06293, over 3064653.67 frames. ], batch size: 50, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:32:04,259 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-04-30 13:33:02,238 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 6500, loss[loss=0.21, simple_loss=0.2958, pruned_loss=0.0621, over 16211.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2921, pruned_loss=0.06195, over 3073717.97 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:34:15,420 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:34:16,252 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.744e+02 3.257e+02 4.082e+02 7.063e+02, threshold=6.513e+02, percent-clipped=0.0 2023-04-30 13:34:40,482 INFO [train.py:904] (3/8) Epoch 17, batch 6550, loss[loss=0.2124, simple_loss=0.3063, pruned_loss=0.05926, over 16892.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2949, pruned_loss=0.06223, over 3097020.40 frames. ], batch size: 116, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:34:44,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4494, 3.3885, 2.6554, 2.0843, 2.2180, 2.2090, 3.5036, 3.1073], device='cuda:3'), covar=tensor([0.2890, 0.0634, 0.1703, 0.2889, 0.2768, 0.2147, 0.0491, 0.1297], device='cuda:3'), in_proj_covar=tensor([0.0317, 0.0259, 0.0295, 0.0299, 0.0289, 0.0242, 0.0283, 0.0319], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 13:35:03,818 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:35:06,287 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:35:58,171 INFO [train.py:904] (3/8) Epoch 17, batch 6600, loss[loss=0.2185, simple_loss=0.3079, pruned_loss=0.06458, over 16742.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2967, pruned_loss=0.06269, over 3091484.10 frames. ], batch size: 124, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:36:37,546 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:36:39,969 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:36:50,873 INFO [optim.py:368] (3/8) 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 13:36:54,447 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4856, 3.4678, 3.4511, 2.7881, 3.3288, 2.1229, 3.1387, 2.8636], device='cuda:3'), covar=tensor([0.0142, 0.0113, 0.0168, 0.0218, 0.0094, 0.1973, 0.0124, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0139, 0.0185, 0.0169, 0.0159, 0.0196, 0.0172, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:36:57,740 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5586, 4.8252, 4.5600, 4.5809, 4.4032, 4.3456, 4.3576, 4.8926], device='cuda:3'), covar=tensor([0.1172, 0.0796, 0.0975, 0.0813, 0.0792, 0.1271, 0.1029, 0.0786], device='cuda:3'), in_proj_covar=tensor([0.0633, 0.0773, 0.0633, 0.0574, 0.0485, 0.0497, 0.0641, 0.0597], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:37:07,710 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 13:37:13,185 INFO [train.py:904] (3/8) Epoch 17, batch 6650, loss[loss=0.1998, simple_loss=0.2833, pruned_loss=0.05818, over 16942.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2976, pruned_loss=0.06404, over 3079507.37 frames. ], batch size: 109, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:37:59,841 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 6700, loss[loss=0.1917, simple_loss=0.2792, pruned_loss=0.05209, over 16547.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2961, pruned_loss=0.06388, over 3080473.34 frames. ], batch size: 68, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:38:53,629 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 13:39:34,958 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 6750, loss[loss=0.2367, simple_loss=0.3071, pruned_loss=0.08312, over 12406.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.294, pruned_loss=0.0628, over 3096751.30 frames. ], batch size: 248, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:39:58,663 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:40:02,625 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6414, 2.2527, 1.7368, 2.0633, 2.6335, 2.3020, 2.5605, 2.8341], device='cuda:3'), covar=tensor([0.0174, 0.0406, 0.0524, 0.0431, 0.0232, 0.0364, 0.0190, 0.0230], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0224, 0.0217, 0.0217, 0.0225, 0.0223, 0.0226, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:40:26,154 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:41:04,380 INFO [train.py:904] (3/8) Epoch 17, batch 6800, loss[loss=0.2335, simple_loss=0.3222, pruned_loss=0.07235, over 16224.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2938, pruned_loss=0.06261, over 3099293.78 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:41:33,047 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:42:02,685 INFO [optim.py:368] (3/8) 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 13:42:16,100 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4522, 4.5571, 4.3075, 4.0121, 3.9695, 4.4333, 4.1523, 4.0857], device='cuda:3'), covar=tensor([0.1134, 0.1328, 0.0507, 0.0476, 0.0973, 0.0897, 0.0961, 0.1177], device='cuda:3'), in_proj_covar=tensor([0.0272, 0.0384, 0.0322, 0.0309, 0.0331, 0.0360, 0.0219, 0.0385], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:42:23,105 INFO [train.py:904] (3/8) Epoch 17, batch 6850, loss[loss=0.1958, simple_loss=0.2933, pruned_loss=0.0492, over 16660.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2943, pruned_loss=0.06276, over 3101346.53 frames. ], batch size: 57, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:42:45,399 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 13:43:23,385 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0795, 2.3755, 2.4159, 2.8293, 2.0782, 3.2667, 1.8893, 2.7428], device='cuda:3'), covar=tensor([0.1010, 0.0539, 0.0913, 0.0169, 0.0119, 0.0371, 0.1254, 0.0613], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0167, 0.0190, 0.0175, 0.0203, 0.0211, 0.0193, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 13:43:37,896 INFO [train.py:904] (3/8) Epoch 17, batch 6900, loss[loss=0.2086, simple_loss=0.2983, pruned_loss=0.05946, over 16250.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2972, pruned_loss=0.06276, over 3100118.03 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:44:10,321 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:44:14,314 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:44:33,263 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.634e+02 3.268e+02 4.133e+02 7.362e+02, threshold=6.535e+02, percent-clipped=1.0 2023-04-30 13:44:55,623 INFO [train.py:904] (3/8) Epoch 17, batch 6950, loss[loss=0.2437, simple_loss=0.3195, pruned_loss=0.08398, over 11317.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.299, pruned_loss=0.06483, over 3070237.24 frames. ], batch size: 247, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:46:11,848 INFO [train.py:904] (3/8) Epoch 17, batch 7000, loss[loss=0.1903, simple_loss=0.2934, pruned_loss=0.04358, over 16848.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2985, pruned_loss=0.06402, over 3069445.02 frames. ], batch size: 83, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:46:12,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0929, 5.8175, 5.9982, 5.6847, 5.7458, 6.3429, 5.8010, 5.6586], device='cuda:3'), covar=tensor([0.0935, 0.1709, 0.1966, 0.1965, 0.2534, 0.0969, 0.1565, 0.2279], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0562, 0.0614, 0.0469, 0.0634, 0.0650, 0.0490, 0.0635], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 13:47:07,099 INFO [optim.py:368] (3/8) 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 13:47:07,436 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 7050, loss[loss=0.1984, simple_loss=0.2929, pruned_loss=0.05195, over 16479.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2983, pruned_loss=0.06249, over 3102994.56 frames. ], batch size: 75, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:47:29,370 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 13:48:00,524 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:48:34,024 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 13:48:47,182 INFO [train.py:904] (3/8) Epoch 17, batch 7100, loss[loss=0.2145, simple_loss=0.3008, pruned_loss=0.06407, over 16220.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2968, pruned_loss=0.06203, over 3111815.12 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:49:08,055 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:49:42,142 INFO [optim.py:368] (3/8) 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 13:50:02,970 INFO [train.py:904] (3/8) Epoch 17, batch 7150, loss[loss=0.2254, simple_loss=0.3088, pruned_loss=0.07103, over 16852.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2952, pruned_loss=0.06212, over 3107098.43 frames. ], batch size: 116, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:50:27,954 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 13:51:19,557 INFO [train.py:904] (3/8) Epoch 17, batch 7200, loss[loss=0.1778, simple_loss=0.2678, pruned_loss=0.0439, over 16640.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2931, pruned_loss=0.0607, over 3096386.06 frames. ], batch size: 134, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:51:40,698 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8725, 2.0602, 2.4041, 3.1702, 2.1591, 2.2776, 2.2468, 2.1531], device='cuda:3'), covar=tensor([0.1243, 0.3312, 0.2237, 0.0655, 0.3831, 0.2541, 0.3171, 0.3367], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0420, 0.0348, 0.0316, 0.0424, 0.0486, 0.0391, 0.0491], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:51:48,741 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:51,919 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:55,203 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 13:52:16,449 INFO [optim.py:368] (3/8) 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 13:52:21,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1014, 2.1390, 2.2382, 3.6416, 2.1385, 2.5067, 2.2594, 2.2735], device='cuda:3'), covar=tensor([0.1208, 0.3204, 0.2675, 0.0549, 0.3954, 0.2357, 0.3163, 0.3101], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0420, 0.0348, 0.0316, 0.0424, 0.0486, 0.0391, 0.0490], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:52:38,958 INFO [train.py:904] (3/8) Epoch 17, batch 7250, loss[loss=0.2319, simple_loss=0.2974, pruned_loss=0.08318, over 11447.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2909, pruned_loss=0.05946, over 3097705.00 frames. ], batch size: 247, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:53:08,594 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:53:12,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4182, 3.3775, 3.4073, 2.4698, 3.3082, 2.0694, 3.0426, 2.7211], device='cuda:3'), covar=tensor([0.0208, 0.0163, 0.0215, 0.0447, 0.0136, 0.2511, 0.0185, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0136, 0.0183, 0.0167, 0.0156, 0.0193, 0.0169, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:53:25,501 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:57,113 INFO [train.py:904] (3/8) Epoch 17, batch 7300, loss[loss=0.231, simple_loss=0.3051, pruned_loss=0.07851, over 11574.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2905, pruned_loss=0.05947, over 3084720.20 frames. ], batch size: 248, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:54:41,407 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:54:45,840 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0238, 3.0165, 3.0575, 5.1492, 4.0903, 4.4103, 1.9848, 3.4755], device='cuda:3'), covar=tensor([0.1299, 0.0759, 0.1151, 0.0111, 0.0362, 0.0362, 0.1510, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0169, 0.0192, 0.0175, 0.0204, 0.0214, 0.0196, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 13:54:47,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9899, 1.9996, 2.5602, 2.9183, 2.8566, 3.5306, 2.0417, 3.3405], device='cuda:3'), covar=tensor([0.0197, 0.0449, 0.0275, 0.0272, 0.0250, 0.0117, 0.0484, 0.0123], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0184, 0.0170, 0.0173, 0.0183, 0.0142, 0.0186, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:54:55,062 INFO [optim.py:368] (3/8) 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 13:54:55,538 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:10,621 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:16,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2067, 4.1153, 4.2974, 4.4171, 4.5177, 4.1331, 4.4565, 4.5544], device='cuda:3'), covar=tensor([0.1716, 0.1165, 0.1358, 0.0623, 0.0589, 0.1192, 0.0746, 0.0605], device='cuda:3'), in_proj_covar=tensor([0.0577, 0.0714, 0.0850, 0.0729, 0.0552, 0.0580, 0.0590, 0.0681], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:55:17,627 INFO [train.py:904] (3/8) Epoch 17, batch 7350, loss[loss=0.2163, simple_loss=0.3032, pruned_loss=0.06469, over 16815.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2911, pruned_loss=0.06042, over 3057538.21 frames. ], batch size: 124, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:55:49,800 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:11,082 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3634, 3.6035, 3.6659, 2.2285, 3.1333, 2.5167, 3.8252, 3.8630], device='cuda:3'), covar=tensor([0.0279, 0.0773, 0.0582, 0.1917, 0.0807, 0.0883, 0.0635, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0160, 0.0165, 0.0151, 0.0144, 0.0128, 0.0143, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 13:56:12,136 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,350 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:19,806 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:37,950 INFO [train.py:904] (3/8) Epoch 17, batch 7400, loss[loss=0.2203, simple_loss=0.3028, pruned_loss=0.06892, over 16668.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.293, pruned_loss=0.06129, over 3055685.49 frames. ], batch size: 57, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:56:48,596 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:59,827 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:07,984 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:36,685 INFO [optim.py:368] (3/8) 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 13:57:52,301 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:59,464 INFO [train.py:904] (3/8) Epoch 17, batch 7450, loss[loss=0.2284, simple_loss=0.2972, pruned_loss=0.07976, over 11410.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2942, pruned_loss=0.06251, over 3045455.33 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 13:58:18,999 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:58:25,441 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.61 vs. limit=5.0 2023-04-30 13:58:50,188 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7985, 2.7451, 2.1864, 2.4907, 3.2089, 2.8734, 3.4039, 3.4243], device='cuda:3'), covar=tensor([0.0100, 0.0401, 0.0538, 0.0415, 0.0229, 0.0348, 0.0228, 0.0229], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0220, 0.0213, 0.0214, 0.0220, 0.0219, 0.0221, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 13:59:20,327 INFO [train.py:904] (3/8) Epoch 17, batch 7500, loss[loss=0.1907, simple_loss=0.2787, pruned_loss=0.05136, over 16535.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2942, pruned_loss=0.06164, over 3049953.84 frames. ], batch size: 75, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:00:17,626 INFO [optim.py:368] (3/8) 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 14:00:23,203 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 14:00:39,256 INFO [train.py:904] (3/8) Epoch 17, batch 7550, loss[loss=0.1856, simple_loss=0.2882, pruned_loss=0.04153, over 16804.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2946, pruned_loss=0.06283, over 3005770.51 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:01:19,623 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 7600, loss[loss=0.1828, simple_loss=0.2701, pruned_loss=0.0478, over 17100.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2934, pruned_loss=0.06257, over 3018142.91 frames. ], batch size: 49, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:02:34,914 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-30 14:02:58,189 INFO [optim.py:368] (3/8) 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 14:03:08,020 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 7650, loss[loss=0.2027, simple_loss=0.2918, pruned_loss=0.05682, over 16748.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2939, pruned_loss=0.06298, over 3022413.05 frames. ], batch size: 124, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:07,612 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:33,133 INFO [train.py:904] (3/8) Epoch 17, batch 7700, loss[loss=0.2042, simple_loss=0.2852, pruned_loss=0.06166, over 16703.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2931, pruned_loss=0.06314, over 3036814.62 frames. ], batch size: 62, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:34,758 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:39,604 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.959e+02 3.624e+02 4.617e+02 9.999e+02, threshold=7.248e+02, percent-clipped=4.0 2023-04-30 14:05:34,958 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:50,026 INFO [train.py:904] (3/8) Epoch 17, batch 7750, loss[loss=0.1777, simple_loss=0.2748, pruned_loss=0.04023, over 16748.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2925, pruned_loss=0.06202, over 3060319.44 frames. ], batch size: 89, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:07:07,219 INFO [train.py:904] (3/8) Epoch 17, batch 7800, loss[loss=0.1927, simple_loss=0.2791, pruned_loss=0.05319, over 15314.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2939, pruned_loss=0.06284, over 3071058.06 frames. ], batch size: 190, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:07:53,480 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4271, 3.0379, 2.5874, 2.2231, 2.3197, 2.2251, 3.0203, 2.8890], device='cuda:3'), covar=tensor([0.2613, 0.0747, 0.1650, 0.2500, 0.2703, 0.2126, 0.0555, 0.1340], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0262, 0.0299, 0.0302, 0.0291, 0.0244, 0.0287, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:08:03,355 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 7850, loss[loss=0.2411, simple_loss=0.3102, pruned_loss=0.08598, over 11563.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2954, pruned_loss=0.06361, over 3049626.31 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:09:00,997 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 7900, loss[loss=0.1939, simple_loss=0.2849, pruned_loss=0.05149, over 16818.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2943, pruned_loss=0.06286, over 3057004.45 frames. ], batch size: 76, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:10:13,071 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:10:20,365 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0515, 3.3872, 3.4012, 1.9576, 2.9800, 2.2527, 3.5470, 3.6523], device='cuda:3'), covar=tensor([0.0250, 0.0789, 0.0617, 0.2049, 0.0839, 0.0989, 0.0607, 0.0839], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0157, 0.0163, 0.0149, 0.0142, 0.0126, 0.0141, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 14:10:26,394 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 14:10:30,093 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8150, 2.6721, 2.6063, 1.9159, 2.5486, 2.7140, 2.6271, 1.8073], device='cuda:3'), covar=tensor([0.0469, 0.0091, 0.0081, 0.0363, 0.0131, 0.0131, 0.0109, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0077, 0.0078, 0.0132, 0.0091, 0.0103, 0.0090, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:10:36,113 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 7950, loss[loss=0.1911, simple_loss=0.2725, pruned_loss=0.05483, over 16830.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2946, pruned_loss=0.06319, over 3068992.62 frames. ], batch size: 83, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:11:49,406 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:11:53,825 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6603, 3.6840, 2.2407, 4.2406, 2.7351, 4.1964, 2.4292, 2.9860], device='cuda:3'), covar=tensor([0.0254, 0.0375, 0.1582, 0.0157, 0.0838, 0.0566, 0.1495, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0170, 0.0192, 0.0150, 0.0172, 0.0211, 0.0200, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 14:12:13,320 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:14,140 INFO [train.py:904] (3/8) Epoch 17, batch 8000, loss[loss=0.2269, simple_loss=0.3166, pruned_loss=0.06859, over 16210.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2955, pruned_loss=0.06393, over 3065625.76 frames. ], batch size: 165, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:12:15,877 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:28,428 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 8050, loss[loss=0.2138, simple_loss=0.3025, pruned_loss=0.06253, over 15416.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2951, pruned_loss=0.06325, over 3061746.38 frames. ], batch size: 191, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:13:44,268 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:28,398 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5435, 3.5879, 2.7693, 2.1210, 2.3080, 2.2967, 3.7811, 3.2299], device='cuda:3'), covar=tensor([0.2997, 0.0711, 0.1843, 0.2808, 0.2674, 0.2121, 0.0471, 0.1283], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0261, 0.0300, 0.0301, 0.0291, 0.0244, 0.0287, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:14:37,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0308, 3.0919, 1.8606, 3.2149, 2.3758, 3.3343, 2.0950, 2.5865], device='cuda:3'), covar=tensor([0.0256, 0.0368, 0.1551, 0.0197, 0.0773, 0.0555, 0.1411, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0169, 0.0191, 0.0149, 0.0171, 0.0209, 0.0198, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 14:14:46,136 INFO [train.py:904] (3/8) Epoch 17, batch 8100, loss[loss=0.2119, simple_loss=0.2946, pruned_loss=0.06459, over 16929.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2949, pruned_loss=0.06288, over 3063000.13 frames. ], batch size: 109, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:14:50,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7500, 3.8551, 4.1419, 4.1141, 4.1162, 3.8567, 3.9102, 3.8866], device='cuda:3'), covar=tensor([0.0342, 0.0588, 0.0369, 0.0397, 0.0451, 0.0408, 0.0823, 0.0505], device='cuda:3'), in_proj_covar=tensor([0.0376, 0.0408, 0.0402, 0.0379, 0.0452, 0.0423, 0.0519, 0.0334], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 14:14:52,577 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-30 14:15:16,575 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:15:59,004 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2511, 6.0455, 6.1930, 5.8269, 5.9150, 6.4546, 5.9564, 5.7479], device='cuda:3'), covar=tensor([0.0850, 0.1557, 0.1943, 0.1647, 0.2213, 0.0782, 0.1441, 0.2129], device='cuda:3'), in_proj_covar=tensor([0.0384, 0.0556, 0.0611, 0.0466, 0.0626, 0.0647, 0.0483, 0.0627], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:16:00,978 INFO [train.py:904] (3/8) Epoch 17, batch 8150, loss[loss=0.1856, simple_loss=0.267, pruned_loss=0.05207, over 16765.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.292, pruned_loss=0.06177, over 3083931.32 frames. ], batch size: 89, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:16:27,575 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:17:11,457 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4495, 2.9448, 2.6831, 2.2302, 2.2371, 2.2620, 2.9048, 2.8565], device='cuda:3'), covar=tensor([0.2314, 0.0684, 0.1501, 0.2447, 0.2259, 0.1979, 0.0536, 0.1260], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0262, 0.0300, 0.0302, 0.0292, 0.0244, 0.0286, 0.0323], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:17:18,489 INFO [train.py:904] (3/8) Epoch 17, batch 8200, loss[loss=0.1897, simple_loss=0.2862, pruned_loss=0.04661, over 16313.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2891, pruned_loss=0.06085, over 3099969.10 frames. ], batch size: 68, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:18:05,889 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 17, batch 8250, loss[loss=0.1815, simple_loss=0.2664, pruned_loss=0.04837, over 11736.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2881, pruned_loss=0.05849, over 3078830.34 frames. ], batch size: 248, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:20:04,017 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:20:05,299 INFO [train.py:904] (3/8) Epoch 17, batch 8300, loss[loss=0.1808, simple_loss=0.2775, pruned_loss=0.0421, over 16200.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2847, pruned_loss=0.05518, over 3065653.60 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:21:09,571 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.327e+02 2.774e+02 3.132e+02 7.452e+02, threshold=5.547e+02, percent-clipped=1.0 2023-04-30 14:21:26,073 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:21:29,852 INFO [train.py:904] (3/8) Epoch 17, batch 8350, loss[loss=0.2019, simple_loss=0.2959, pruned_loss=0.05399, over 16372.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2835, pruned_loss=0.05294, over 3056854.96 frames. ], batch size: 146, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:21:36,771 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8302, 3.7239, 3.8960, 3.9945, 4.0936, 3.6659, 4.0216, 4.1018], device='cuda:3'), covar=tensor([0.1505, 0.1053, 0.1186, 0.0668, 0.0562, 0.1831, 0.0790, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0578, 0.0715, 0.0845, 0.0726, 0.0552, 0.0575, 0.0588, 0.0684], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:22:51,858 INFO [train.py:904] (3/8) Epoch 17, batch 8400, loss[loss=0.1864, simple_loss=0.2745, pruned_loss=0.04915, over 11784.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2809, pruned_loss=0.05107, over 3034428.31 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:23:16,979 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 17, batch 8450, loss[loss=0.1657, simple_loss=0.2616, pruned_loss=0.03484, over 15252.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2793, pruned_loss=0.04958, over 3035960.28 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:25:32,473 INFO [train.py:904] (3/8) Epoch 17, batch 8500, loss[loss=0.174, simple_loss=0.2541, pruned_loss=0.04701, over 11923.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2753, pruned_loss=0.04707, over 3027113.87 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:26:07,521 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0541, 2.3228, 2.3728, 3.0246, 1.9623, 3.2603, 1.8036, 2.8307], device='cuda:3'), covar=tensor([0.1126, 0.0618, 0.0935, 0.0178, 0.0116, 0.0389, 0.1446, 0.0618], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0165, 0.0186, 0.0171, 0.0199, 0.0209, 0.0192, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 14:26:10,252 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:14,622 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:19,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4788, 2.0589, 1.8108, 1.7462, 2.2851, 1.9886, 2.0834, 2.3768], device='cuda:3'), covar=tensor([0.0172, 0.0297, 0.0391, 0.0410, 0.0218, 0.0300, 0.0203, 0.0258], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0217, 0.0210, 0.0211, 0.0216, 0.0215, 0.0216, 0.0209], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:26:34,258 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 8550, loss[loss=0.1981, simple_loss=0.289, pruned_loss=0.05363, over 15239.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2732, pruned_loss=0.04622, over 3007722.41 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:27:27,455 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-30 14:28:08,376 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:28:15,909 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4267, 2.0474, 1.7837, 1.7688, 2.2789, 1.9142, 1.9237, 2.3687], device='cuda:3'), covar=tensor([0.0171, 0.0376, 0.0469, 0.0448, 0.0254, 0.0378, 0.0195, 0.0272], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0218, 0.0212, 0.0212, 0.0218, 0.0217, 0.0217, 0.0210], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:28:33,701 INFO [train.py:904] (3/8) Epoch 17, batch 8600, loss[loss=0.1848, simple_loss=0.2828, pruned_loss=0.04345, over 16736.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2739, pruned_loss=0.0451, over 3031089.48 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:29:52,152 INFO [optim.py:368] (3/8) 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 14:30:15,678 INFO [train.py:904] (3/8) Epoch 17, batch 8650, loss[loss=0.1613, simple_loss=0.2614, pruned_loss=0.03057, over 15425.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2719, pruned_loss=0.04368, over 3034051.24 frames. ], batch size: 191, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:31:43,989 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:31:47,950 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4174, 3.3555, 3.4712, 3.5347, 3.5940, 3.3162, 3.5721, 3.6372], device='cuda:3'), covar=tensor([0.1240, 0.0890, 0.1028, 0.0627, 0.0634, 0.2261, 0.0804, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0697, 0.0825, 0.0712, 0.0540, 0.0562, 0.0575, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:32:02,416 INFO [train.py:904] (3/8) Epoch 17, batch 8700, loss[loss=0.1646, simple_loss=0.2593, pruned_loss=0.03497, over 16524.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2694, pruned_loss=0.04252, over 3045071.55 frames. ], batch size: 68, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:32:31,753 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 17, batch 8750, loss[loss=0.184, simple_loss=0.2839, pruned_loss=0.04206, over 16725.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2696, pruned_loss=0.04207, over 3057708.55 frames. ], batch size: 83, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:33:43,343 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:34:13,579 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:03,188 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6129, 3.5692, 3.5139, 2.8106, 3.4731, 1.8512, 3.2875, 2.9632], device='cuda:3'), covar=tensor([0.0118, 0.0102, 0.0159, 0.0208, 0.0103, 0.2368, 0.0131, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0133, 0.0176, 0.0161, 0.0152, 0.0188, 0.0165, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:35:21,380 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 8800, loss[loss=0.166, simple_loss=0.2617, pruned_loss=0.03512, over 16430.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.268, pruned_loss=0.04088, over 3058769.58 frames. ], batch size: 146, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:36:22,032 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:36:54,974 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 8850, loss[loss=0.1754, simple_loss=0.2625, pruned_loss=0.04413, over 12571.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2709, pruned_loss=0.04088, over 3060120.14 frames. ], batch size: 249, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:37:28,383 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:03,152 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:27,261 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:33,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5293, 4.6955, 4.8822, 4.6840, 4.7145, 5.2574, 4.7268, 4.4791], device='cuda:3'), covar=tensor([0.1257, 0.1805, 0.1872, 0.2045, 0.2639, 0.1008, 0.1646, 0.2433], device='cuda:3'), in_proj_covar=tensor([0.0369, 0.0534, 0.0585, 0.0450, 0.0600, 0.0624, 0.0467, 0.0600], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:39:01,946 INFO [train.py:904] (3/8) Epoch 17, batch 8900, loss[loss=0.1804, simple_loss=0.276, pruned_loss=0.04235, over 16782.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2716, pruned_loss=0.04072, over 3051476.16 frames. ], batch size: 134, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:40:38,363 INFO [optim.py:368] (3/8) 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 14:40:50,848 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:41:04,757 INFO [train.py:904] (3/8) Epoch 17, batch 8950, loss[loss=0.1585, simple_loss=0.252, pruned_loss=0.03249, over 16449.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2707, pruned_loss=0.04046, over 3075666.98 frames. ], batch size: 68, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,011 INFO [train.py:904] (3/8) Epoch 17, batch 9000, loss[loss=0.1748, simple_loss=0.2623, pruned_loss=0.0436, over 16391.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2669, pruned_loss=0.0389, over 3084306.48 frames. ], batch size: 146, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,011 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 14:43:02,950 INFO [train.py:938] (3/8) Epoch 17, validation: loss=0.1478, simple_loss=0.2519, pruned_loss=0.02187, over 944034.00 frames. 2023-04-30 14:43:02,951 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 14:43:11,014 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:43:21,900 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5956, 3.5513, 3.5249, 2.8396, 3.4445, 2.0160, 3.3185, 2.9691], device='cuda:3'), covar=tensor([0.0118, 0.0115, 0.0154, 0.0209, 0.0101, 0.2235, 0.0128, 0.0199], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0133, 0.0177, 0.0161, 0.0152, 0.0190, 0.0165, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:43:31,374 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8000, 3.7680, 4.1525, 4.1102, 4.1234, 3.8761, 3.9049, 3.9343], device='cuda:3'), covar=tensor([0.0357, 0.0943, 0.0448, 0.0448, 0.0504, 0.0470, 0.0790, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0365, 0.0397, 0.0392, 0.0371, 0.0438, 0.0411, 0.0503, 0.0325], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 14:44:23,318 INFO [optim.py:368] (3/8) 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 14:44:40,990 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 9050, loss[loss=0.1777, simple_loss=0.2709, pruned_loss=0.04226, over 12889.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2681, pruned_loss=0.03948, over 3079937.46 frames. ], batch size: 247, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:45:46,645 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4862, 3.3991, 3.5180, 3.6072, 3.6458, 3.3169, 3.6330, 3.6851], device='cuda:3'), covar=tensor([0.1177, 0.0974, 0.1073, 0.0611, 0.0623, 0.2173, 0.0769, 0.0811], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0696, 0.0819, 0.0709, 0.0536, 0.0559, 0.0572, 0.0666], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:46:24,657 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 9100, loss[loss=0.1494, simple_loss=0.2449, pruned_loss=0.02691, over 17121.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2678, pruned_loss=0.04006, over 3061477.85 frames. ], batch size: 49, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:46:52,839 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7654, 3.4431, 3.8470, 1.7641, 3.9715, 4.0660, 3.0900, 3.0177], device='cuda:3'), covar=tensor([0.0594, 0.0225, 0.0159, 0.1229, 0.0053, 0.0105, 0.0340, 0.0417], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0100, 0.0088, 0.0132, 0.0071, 0.0112, 0.0119, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 14:46:55,490 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1084, 2.0399, 2.1436, 3.7160, 2.0300, 2.3395, 2.1740, 2.1465], device='cuda:3'), covar=tensor([0.1111, 0.3764, 0.2944, 0.0480, 0.4320, 0.2628, 0.3639, 0.3513], device='cuda:3'), in_proj_covar=tensor([0.0372, 0.0415, 0.0346, 0.0309, 0.0419, 0.0476, 0.0386, 0.0480], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:47:16,435 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9963, 4.2585, 4.0860, 4.1048, 3.7852, 3.8154, 3.8518, 4.2545], device='cuda:3'), covar=tensor([0.1054, 0.0903, 0.0900, 0.0818, 0.0745, 0.1823, 0.1002, 0.0899], device='cuda:3'), in_proj_covar=tensor([0.0604, 0.0739, 0.0597, 0.0545, 0.0462, 0.0475, 0.0612, 0.0573], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:48:02,850 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 9150, loss[loss=0.1613, simple_loss=0.2558, pruned_loss=0.03344, over 16288.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2684, pruned_loss=0.03977, over 3070446.39 frames. ], batch size: 146, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:48:28,875 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,345 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:49:07,021 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5467, 3.6601, 2.1839, 4.1300, 2.7196, 4.0342, 2.3364, 2.8832], device='cuda:3'), covar=tensor([0.0291, 0.0344, 0.1602, 0.0166, 0.0850, 0.0470, 0.1492, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0165, 0.0187, 0.0144, 0.0168, 0.0202, 0.0195, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 14:49:36,109 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:49:47,891 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6065, 3.6038, 3.5530, 3.0103, 3.4620, 2.0231, 3.3566, 3.0145], device='cuda:3'), covar=tensor([0.0143, 0.0127, 0.0191, 0.0206, 0.0119, 0.2200, 0.0133, 0.0184], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0134, 0.0177, 0.0160, 0.0153, 0.0190, 0.0165, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:50:13,333 INFO [train.py:904] (3/8) Epoch 17, batch 9200, loss[loss=0.1676, simple_loss=0.2623, pruned_loss=0.03648, over 16399.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2639, pruned_loss=0.03895, over 3060583.83 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:50:19,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0357, 3.0779, 1.9101, 3.2876, 2.2774, 3.2993, 2.0639, 2.5865], device='cuda:3'), covar=tensor([0.0330, 0.0373, 0.1471, 0.0310, 0.0826, 0.0584, 0.1384, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0165, 0.0187, 0.0144, 0.0167, 0.0202, 0.0195, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 14:51:10,342 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 17, batch 9250, loss[loss=0.1593, simple_loss=0.2586, pruned_loss=0.02998, over 15384.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2635, pruned_loss=0.03909, over 3046952.96 frames. ], batch size: 191, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:52:52,706 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6253, 4.4392, 4.6993, 4.8625, 5.0153, 4.5449, 5.0076, 5.0059], device='cuda:3'), covar=tensor([0.1693, 0.1184, 0.1477, 0.0652, 0.0453, 0.0749, 0.0489, 0.0574], device='cuda:3'), in_proj_covar=tensor([0.0564, 0.0691, 0.0817, 0.0707, 0.0535, 0.0557, 0.0571, 0.0665], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:53:42,075 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 9300, loss[loss=0.1574, simple_loss=0.2523, pruned_loss=0.03121, over 16363.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2621, pruned_loss=0.0383, over 3046885.26 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:54:41,745 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 14:55:09,819 INFO [optim.py:368] (3/8) 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 14:55:15,025 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 14:55:21,300 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 9350, loss[loss=0.1807, simple_loss=0.2703, pruned_loss=0.04552, over 16227.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.262, pruned_loss=0.03811, over 3074622.24 frames. ], batch size: 165, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:55:33,724 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 14:56:00,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5138, 4.6612, 4.8154, 4.6264, 4.7521, 5.1901, 4.6753, 4.3835], device='cuda:3'), covar=tensor([0.1173, 0.1744, 0.1658, 0.1828, 0.2094, 0.0860, 0.1533, 0.2305], device='cuda:3'), in_proj_covar=tensor([0.0361, 0.0529, 0.0580, 0.0444, 0.0594, 0.0615, 0.0463, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 14:56:57,422 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:57:07,474 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 14:57:07,782 INFO [train.py:904] (3/8) Epoch 17, batch 9400, loss[loss=0.1423, simple_loss=0.2322, pruned_loss=0.02619, over 12868.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2611, pruned_loss=0.03781, over 3049220.88 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:57:41,226 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 14:58:13,422 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9695, 3.8332, 4.0528, 4.1554, 4.2383, 3.8311, 4.2043, 4.2799], device='cuda:3'), covar=tensor([0.1581, 0.1069, 0.1195, 0.0654, 0.0530, 0.1512, 0.0675, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0558, 0.0686, 0.0808, 0.0700, 0.0530, 0.0552, 0.0566, 0.0661], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:58:29,671 INFO [optim.py:368] (3/8) 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 14:58:34,386 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2216, 2.1114, 2.0760, 3.8463, 2.0806, 2.4543, 2.1977, 2.2624], device='cuda:3'), covar=tensor([0.1144, 0.3898, 0.3070, 0.0467, 0.4229, 0.2611, 0.3672, 0.3512], device='cuda:3'), in_proj_covar=tensor([0.0371, 0.0413, 0.0346, 0.0309, 0.0417, 0.0474, 0.0384, 0.0479], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 14:58:48,377 INFO [train.py:904] (3/8) Epoch 17, batch 9450, loss[loss=0.1866, simple_loss=0.2684, pruned_loss=0.05238, over 12378.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2638, pruned_loss=0.03835, over 3043512.04 frames. ], batch size: 246, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:48,903 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171852.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:58:52,137 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:26,652 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:31,023 INFO [train.py:904] (3/8) Epoch 17, batch 9500, loss[loss=0.1666, simple_loss=0.2612, pruned_loss=0.03602, over 16675.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2634, pruned_loss=0.03797, over 3060577.74 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:01:41,169 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:01:45,634 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6334, 3.7037, 3.5049, 3.1960, 3.3032, 3.6134, 3.3588, 3.4454], device='cuda:3'), covar=tensor([0.0557, 0.0560, 0.0299, 0.0270, 0.0491, 0.0452, 0.1284, 0.0458], device='cuda:3'), in_proj_covar=tensor([0.0257, 0.0361, 0.0305, 0.0291, 0.0310, 0.0340, 0.0209, 0.0359], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-30 15:01:51,140 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.125e+02 2.584e+02 3.095e+02 6.778e+02, threshold=5.168e+02, percent-clipped=1.0 2023-04-30 15:02:14,949 INFO [train.py:904] (3/8) Epoch 17, batch 9550, loss[loss=0.202, simple_loss=0.295, pruned_loss=0.05449, over 15396.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2623, pruned_loss=0.03771, over 3070407.41 frames. ], batch size: 191, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:03:49,275 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171998.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:57,314 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 9600, loss[loss=0.1917, simple_loss=0.2874, pruned_loss=0.04799, over 15364.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2643, pruned_loss=0.03858, over 3071496.03 frames. ], batch size: 190, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:05:20,542 INFO [optim.py:368] (3/8) 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 15:05:38,177 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 17, batch 9650, loss[loss=0.1501, simple_loss=0.2509, pruned_loss=0.0246, over 16902.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2658, pruned_loss=0.03886, over 3068012.58 frames. ], batch size: 102, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:07:30,331 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0619, 4.0623, 4.4402, 4.3987, 4.4138, 4.1585, 4.1774, 4.1392], device='cuda:3'), covar=tensor([0.0356, 0.0775, 0.0535, 0.0578, 0.0522, 0.0580, 0.0878, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0362, 0.0392, 0.0387, 0.0366, 0.0432, 0.0405, 0.0493, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 15:07:32,480 INFO [train.py:904] (3/8) Epoch 17, batch 9700, loss[loss=0.1789, simple_loss=0.2809, pruned_loss=0.03843, over 16559.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2649, pruned_loss=0.03858, over 3075317.21 frames. ], batch size: 75, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:07:53,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3138, 3.6208, 3.9353, 2.1527, 3.3233, 2.4581, 3.7341, 3.7694], device='cuda:3'), covar=tensor([0.0239, 0.0737, 0.0453, 0.1929, 0.0667, 0.0943, 0.0537, 0.0939], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0148, 0.0158, 0.0145, 0.0136, 0.0123, 0.0136, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 15:08:05,522 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172118.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:08:57,397 INFO [optim.py:368] (3/8) 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 15:09:14,187 INFO [train.py:904] (3/8) Epoch 17, batch 9750, loss[loss=0.163, simple_loss=0.249, pruned_loss=0.0385, over 12428.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2639, pruned_loss=0.03886, over 3075208.48 frames. ], batch size: 247, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:09:18,961 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:09,721 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:54,093 INFO [train.py:904] (3/8) Epoch 17, batch 9800, loss[loss=0.1575, simple_loss=0.2414, pruned_loss=0.03683, over 11935.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2637, pruned_loss=0.03783, over 3082588.36 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:10:54,779 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:12:17,140 INFO [optim.py:368] (3/8) 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 15:12:38,993 INFO [train.py:904] (3/8) Epoch 17, batch 9850, loss[loss=0.151, simple_loss=0.2381, pruned_loss=0.03191, over 12403.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2646, pruned_loss=0.03727, over 3092481.24 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:12:50,626 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1712, 2.0822, 2.0863, 3.8211, 2.0457, 2.4065, 2.2139, 2.2029], device='cuda:3'), covar=tensor([0.1157, 0.3626, 0.3071, 0.0463, 0.4234, 0.2453, 0.3538, 0.3340], device='cuda:3'), in_proj_covar=tensor([0.0373, 0.0416, 0.0347, 0.0311, 0.0418, 0.0474, 0.0385, 0.0480], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:14:13,936 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:14:27,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0218, 2.0434, 2.1904, 3.5068, 2.0065, 2.3114, 2.1709, 2.1646], device='cuda:3'), covar=tensor([0.1189, 0.3738, 0.2770, 0.0513, 0.4184, 0.2635, 0.3562, 0.3645], device='cuda:3'), in_proj_covar=tensor([0.0372, 0.0414, 0.0346, 0.0309, 0.0416, 0.0472, 0.0384, 0.0479], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:14:32,002 INFO [train.py:904] (3/8) Epoch 17, batch 9900, loss[loss=0.1547, simple_loss=0.2427, pruned_loss=0.03337, over 12488.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2645, pruned_loss=0.03715, over 3072675.04 frames. ], batch size: 249, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:15:29,261 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 15:16:10,690 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 17, batch 9950, loss[loss=0.1825, simple_loss=0.271, pruned_loss=0.04702, over 12557.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2665, pruned_loss=0.03742, over 3091101.97 frames. ], batch size: 250, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:18:33,064 INFO [train.py:904] (3/8) Epoch 17, batch 10000, loss[loss=0.1695, simple_loss=0.2622, pruned_loss=0.03837, over 13204.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2653, pruned_loss=0.03684, over 3094231.89 frames. ], batch size: 250, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:18:40,708 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4611, 5.4723, 5.2725, 4.7773, 4.9567, 5.3386, 5.2834, 4.9429], device='cuda:3'), covar=tensor([0.0520, 0.0464, 0.0237, 0.0265, 0.0807, 0.0439, 0.0215, 0.0622], device='cuda:3'), in_proj_covar=tensor([0.0256, 0.0358, 0.0303, 0.0289, 0.0309, 0.0337, 0.0207, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-30 15:19:56,012 INFO [optim.py:368] (3/8) 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 15:19:56,869 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8768, 2.7624, 2.3824, 4.6272, 3.0788, 4.1288, 1.5542, 3.0124], device='cuda:3'), covar=tensor([0.1356, 0.0808, 0.1357, 0.0151, 0.0194, 0.0354, 0.1705, 0.0766], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0163, 0.0184, 0.0165, 0.0189, 0.0204, 0.0189, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:3') 2023-04-30 15:20:13,520 INFO [train.py:904] (3/8) Epoch 17, batch 10050, loss[loss=0.1977, simple_loss=0.2952, pruned_loss=0.05012, over 15443.00 frames. ], tot_loss[loss=0.17, simple_loss=0.266, pruned_loss=0.03705, over 3085539.35 frames. ], batch size: 190, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:20:58,575 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:21:17,525 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 15:21:47,147 INFO [train.py:904] (3/8) Epoch 17, batch 10100, loss[loss=0.1667, simple_loss=0.2561, pruned_loss=0.03865, over 15244.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2673, pruned_loss=0.03769, over 3109403.33 frames. ], batch size: 190, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:22:57,875 INFO [optim.py:368] (3/8) 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 15:23:33,083 INFO [train.py:904] (3/8) Epoch 18, batch 0, loss[loss=0.2309, simple_loss=0.3087, pruned_loss=0.07657, over 16882.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3087, pruned_loss=0.07657, over 16882.00 frames. ], batch size: 96, lr: 3.82e-03, grad_scale: 8.0 2023-04-30 15:23:33,084 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 15:23:40,344 INFO [train.py:938] (3/8) Epoch 18, validation: loss=0.148, simple_loss=0.2515, pruned_loss=0.02223, over 944034.00 frames. 2023-04-30 15:23:40,344 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 15:24:36,441 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172593.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:44,219 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 50, loss[loss=0.1752, simple_loss=0.2634, pruned_loss=0.0435, over 17074.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2723, pruned_loss=0.05188, over 754724.45 frames. ], batch size: 53, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:21,122 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:28,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6611, 3.7983, 4.0928, 2.9821, 3.6081, 4.0847, 3.7254, 2.5488], device='cuda:3'), covar=tensor([0.0511, 0.0309, 0.0051, 0.0349, 0.0117, 0.0089, 0.0099, 0.0400], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0077, 0.0076, 0.0132, 0.0091, 0.0100, 0.0088, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:25:42,961 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172641.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:48,588 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.462e+02 2.938e+02 3.654e+02 9.185e+02, threshold=5.877e+02, percent-clipped=7.0 2023-04-30 15:25:56,033 INFO [train.py:904] (3/8) Epoch 18, batch 100, loss[loss=0.1942, simple_loss=0.282, pruned_loss=0.05318, over 17054.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2681, pruned_loss=0.04857, over 1328572.16 frames. ], batch size: 53, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:26:05,930 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:16,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4946, 2.8243, 3.0304, 1.9438, 2.7192, 2.0560, 3.0535, 3.1504], device='cuda:3'), covar=tensor([0.0275, 0.0841, 0.0583, 0.1878, 0.0799, 0.1038, 0.0588, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0150, 0.0160, 0.0147, 0.0138, 0.0125, 0.0138, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 15:26:24,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4135, 4.2798, 4.3074, 4.0337, 4.0496, 4.3276, 4.1593, 4.0675], device='cuda:3'), covar=tensor([0.0654, 0.0773, 0.0312, 0.0296, 0.0798, 0.0476, 0.0549, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0262, 0.0370, 0.0310, 0.0297, 0.0318, 0.0345, 0.0212, 0.0367], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:26:44,059 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:46,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9216, 4.2540, 2.9570, 2.2743, 2.7550, 2.4063, 4.6251, 3.5420], device='cuda:3'), covar=tensor([0.2633, 0.0608, 0.1790, 0.2922, 0.2820, 0.2069, 0.0340, 0.1396], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0254, 0.0293, 0.0294, 0.0276, 0.0240, 0.0277, 0.0313], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:27:02,718 INFO [train.py:904] (3/8) Epoch 18, batch 150, loss[loss=0.1708, simple_loss=0.2568, pruned_loss=0.04245, over 17229.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2691, pruned_loss=0.04973, over 1770812.97 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:27:46,324 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:27:52,935 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 15:28:01,449 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.271e+02 2.721e+02 3.424e+02 7.505e+02, threshold=5.441e+02, percent-clipped=3.0 2023-04-30 15:28:09,990 INFO [train.py:904] (3/8) Epoch 18, batch 200, loss[loss=0.1637, simple_loss=0.2623, pruned_loss=0.03253, over 17261.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2671, pruned_loss=0.04862, over 2118310.34 frames. ], batch size: 52, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:28:36,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8694, 4.0681, 4.2811, 3.2262, 3.7064, 4.2669, 3.8979, 2.5372], device='cuda:3'), covar=tensor([0.0471, 0.0078, 0.0039, 0.0320, 0.0103, 0.0082, 0.0078, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0077, 0.0077, 0.0132, 0.0091, 0.0100, 0.0088, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:28:41,352 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:10,028 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:10,422 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 15:29:18,639 INFO [train.py:904] (3/8) Epoch 18, batch 250, loss[loss=0.176, simple_loss=0.2711, pruned_loss=0.04044, over 17123.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2659, pruned_loss=0.04845, over 2375870.36 frames. ], batch size: 47, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:29:29,820 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8633, 4.7868, 4.6841, 4.2227, 4.7696, 1.8587, 4.5168, 4.5007], device='cuda:3'), covar=tensor([0.0111, 0.0099, 0.0181, 0.0305, 0.0107, 0.2532, 0.0146, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0138, 0.0181, 0.0164, 0.0157, 0.0196, 0.0170, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:29:47,728 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:18,790 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.240e+02 2.726e+02 3.152e+02 7.045e+02, threshold=5.451e+02, percent-clipped=2.0 2023-04-30 15:30:23,673 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0755, 4.4723, 2.9639, 2.3246, 2.8151, 2.4310, 4.6681, 3.6227], device='cuda:3'), covar=tensor([0.2677, 0.0658, 0.2030, 0.2965, 0.3244, 0.2243, 0.0427, 0.1474], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0255, 0.0293, 0.0294, 0.0278, 0.0241, 0.0279, 0.0315], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:30:28,995 INFO [train.py:904] (3/8) Epoch 18, batch 300, loss[loss=0.1743, simple_loss=0.2542, pruned_loss=0.04719, over 15598.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.263, pruned_loss=0.0471, over 2584052.27 frames. ], batch size: 190, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:39,800 INFO [train.py:904] (3/8) Epoch 18, batch 350, loss[loss=0.1806, simple_loss=0.2773, pruned_loss=0.04189, over 17083.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2604, pruned_loss=0.04543, over 2751252.64 frames. ], batch size: 55, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:45,435 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172906.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:32:27,296 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0481, 2.5486, 2.6603, 1.8505, 2.8148, 2.8574, 2.4376, 2.3558], device='cuda:3'), covar=tensor([0.0752, 0.0233, 0.0219, 0.0962, 0.0110, 0.0247, 0.0462, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0105, 0.0091, 0.0138, 0.0074, 0.0118, 0.0125, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 15:32:40,483 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.138e+02 2.555e+02 2.961e+02 4.551e+02, threshold=5.109e+02, percent-clipped=0.0 2023-04-30 15:32:47,897 INFO [train.py:904] (3/8) Epoch 18, batch 400, loss[loss=0.1928, simple_loss=0.2666, pruned_loss=0.05951, over 15513.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2591, pruned_loss=0.04447, over 2876039.13 frames. ], batch size: 190, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:32:51,740 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:10,269 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0168, 4.9140, 4.6902, 4.0441, 4.7744, 1.7040, 4.5109, 4.5986], device='cuda:3'), covar=tensor([0.0084, 0.0093, 0.0230, 0.0478, 0.0121, 0.2852, 0.0162, 0.0249], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0139, 0.0183, 0.0166, 0.0159, 0.0198, 0.0172, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:33:30,367 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:44,198 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-04-30 15:33:59,531 INFO [train.py:904] (3/8) Epoch 18, batch 450, loss[loss=0.1684, simple_loss=0.2442, pruned_loss=0.04633, over 16907.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2569, pruned_loss=0.04355, over 2972012.96 frames. ], batch size: 116, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:00,082 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.073e+02 2.463e+02 2.995e+02 6.633e+02, threshold=4.925e+02, percent-clipped=1.0 2023-04-30 15:35:08,312 INFO [train.py:904] (3/8) Epoch 18, batch 500, loss[loss=0.1743, simple_loss=0.2593, pruned_loss=0.04466, over 16762.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2562, pruned_loss=0.04334, over 3045845.32 frames. ], batch size: 57, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:22,613 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6422, 3.6275, 2.0021, 3.8231, 2.7882, 3.8449, 2.0779, 2.8814], device='cuda:3'), covar=tensor([0.0237, 0.0355, 0.1779, 0.0372, 0.0734, 0.0675, 0.1634, 0.0723], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0173, 0.0197, 0.0154, 0.0175, 0.0214, 0.0205, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 15:35:28,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8883, 2.6699, 2.8465, 2.0920, 2.6917, 2.1215, 2.7177, 2.8748], device='cuda:3'), covar=tensor([0.0281, 0.0897, 0.0499, 0.1866, 0.0782, 0.0895, 0.0590, 0.0901], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0155, 0.0163, 0.0150, 0.0141, 0.0126, 0.0140, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 15:35:59,653 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:36:14,710 INFO [train.py:904] (3/8) Epoch 18, batch 550, loss[loss=0.1604, simple_loss=0.2497, pruned_loss=0.03559, over 17215.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2554, pruned_loss=0.043, over 3107166.65 frames. ], batch size: 44, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:37:14,571 INFO [optim.py:368] (3/8) 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 15:37:20,841 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173150.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:37:22,686 INFO [train.py:904] (3/8) Epoch 18, batch 600, loss[loss=0.168, simple_loss=0.2642, pruned_loss=0.03591, over 17049.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2548, pruned_loss=0.04323, over 3161209.31 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:37:39,525 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-30 15:38:30,581 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173201.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:31,377 INFO [train.py:904] (3/8) Epoch 18, batch 650, loss[loss=0.1665, simple_loss=0.2613, pruned_loss=0.03589, over 17090.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2536, pruned_loss=0.04316, over 3192112.16 frames. ], batch size: 47, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:38:45,496 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:58,319 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-30 15:39:30,715 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4643, 3.8209, 3.8170, 2.0091, 3.0386, 2.0522, 3.7990, 3.8835], device='cuda:3'), covar=tensor([0.0290, 0.0785, 0.0561, 0.2232, 0.0960, 0.1220, 0.0642, 0.1089], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0155, 0.0163, 0.0150, 0.0141, 0.0127, 0.0141, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 15:39:32,447 INFO [optim.py:368] (3/8) 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 15:39:38,125 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 15:39:40,834 INFO [train.py:904] (3/8) Epoch 18, batch 700, loss[loss=0.1819, simple_loss=0.2461, pruned_loss=0.05886, over 11997.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2525, pruned_loss=0.04282, over 3203672.94 frames. ], batch size: 247, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:39:44,097 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173254.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:04,142 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5241, 3.4543, 2.7845, 2.1381, 2.3056, 2.2448, 3.5242, 3.1296], device='cuda:3'), covar=tensor([0.2701, 0.0720, 0.1557, 0.2791, 0.2666, 0.2147, 0.0532, 0.1589], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0260, 0.0297, 0.0299, 0.0284, 0.0244, 0.0283, 0.0321], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:40:22,765 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:27,763 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 15:40:32,606 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2628, 4.7169, 4.2325, 4.5737, 4.2303, 4.1820, 4.2777, 4.7923], device='cuda:3'), covar=tensor([0.2684, 0.1992, 0.2758, 0.1738, 0.1914, 0.2390, 0.2649, 0.1944], device='cuda:3'), in_proj_covar=tensor([0.0644, 0.0788, 0.0635, 0.0585, 0.0497, 0.0499, 0.0654, 0.0607], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:40:50,345 INFO [train.py:904] (3/8) Epoch 18, batch 750, loss[loss=0.1552, simple_loss=0.2547, pruned_loss=0.02791, over 17094.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2536, pruned_loss=0.04337, over 3226095.33 frames. ], batch size: 47, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:40:51,630 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:00,115 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:23,354 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2420, 5.1486, 4.9144, 4.4030, 5.0315, 1.7237, 4.7723, 4.8644], device='cuda:3'), covar=tensor([0.0077, 0.0074, 0.0202, 0.0398, 0.0107, 0.2903, 0.0136, 0.0211], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0141, 0.0186, 0.0168, 0.0161, 0.0199, 0.0174, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:41:26,654 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:26,987 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7393, 2.3427, 2.4424, 4.6004, 2.4011, 2.7847, 2.5270, 2.6112], device='cuda:3'), covar=tensor([0.1107, 0.3716, 0.2938, 0.0413, 0.4087, 0.2637, 0.3308, 0.3532], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0427, 0.0356, 0.0323, 0.0428, 0.0491, 0.0397, 0.0499], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:41:49,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5297, 3.3493, 3.7063, 1.8282, 3.7707, 3.7950, 2.9463, 2.7780], device='cuda:3'), covar=tensor([0.0722, 0.0209, 0.0167, 0.1123, 0.0094, 0.0170, 0.0423, 0.0423], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0106, 0.0093, 0.0140, 0.0075, 0.0120, 0.0126, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 15:41:50,586 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 800, loss[loss=0.1628, simple_loss=0.2638, pruned_loss=0.03095, over 17258.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2531, pruned_loss=0.043, over 3253417.84 frames. ], batch size: 52, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:42:22,698 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:42:49,189 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 850, loss[loss=0.2004, simple_loss=0.2706, pruned_loss=0.06507, over 16875.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2531, pruned_loss=0.04269, over 3264204.23 frames. ], batch size: 116, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:43:34,070 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5592, 2.2365, 2.2544, 4.4259, 2.2879, 2.6930, 2.3255, 2.4067], device='cuda:3'), covar=tensor([0.1123, 0.3672, 0.3012, 0.0434, 0.4076, 0.2694, 0.3650, 0.3627], device='cuda:3'), in_proj_covar=tensor([0.0387, 0.0428, 0.0356, 0.0323, 0.0428, 0.0491, 0.0397, 0.0499], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:43:55,001 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:44:07,464 INFO [optim.py:368] (3/8) 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 15:44:15,664 INFO [train.py:904] (3/8) Epoch 18, batch 900, loss[loss=0.179, simple_loss=0.2545, pruned_loss=0.05177, over 16484.00 frames. ], tot_loss[loss=0.169, simple_loss=0.253, pruned_loss=0.04247, over 3260476.28 frames. ], batch size: 75, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:45:11,338 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173491.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:45:24,042 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:25,649 INFO [train.py:904] (3/8) Epoch 18, batch 950, loss[loss=0.1837, simple_loss=0.267, pruned_loss=0.05022, over 15598.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2523, pruned_loss=0.04209, over 3269514.71 frames. ], batch size: 190, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:45:31,835 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:46:26,062 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173549.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:46:35,247 INFO [train.py:904] (3/8) Epoch 18, batch 1000, loss[loss=0.1374, simple_loss=0.2267, pruned_loss=0.02408, over 16817.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2522, pruned_loss=0.0425, over 3283846.22 frames. ], batch size: 42, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:46:35,714 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173552.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:47:44,517 INFO [train.py:904] (3/8) Epoch 18, batch 1050, loss[loss=0.1566, simple_loss=0.2459, pruned_loss=0.03367, over 17222.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2516, pruned_loss=0.0425, over 3283375.02 frames. ], batch size: 45, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:48:02,457 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:48:46,279 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 1100, loss[loss=0.1798, simple_loss=0.276, pruned_loss=0.04178, over 17062.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2521, pruned_loss=0.04204, over 3294400.40 frames. ], batch size: 53, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:49:12,455 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:49:26,688 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173675.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:50:01,541 INFO [train.py:904] (3/8) Epoch 18, batch 1150, loss[loss=0.184, simple_loss=0.2531, pruned_loss=0.05741, over 16696.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2512, pruned_loss=0.04156, over 3303192.37 frames. ], batch size: 124, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:50:56,445 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1892, 3.2828, 3.4295, 2.1930, 2.8893, 2.4036, 3.6669, 3.6654], device='cuda:3'), covar=tensor([0.0253, 0.0900, 0.0623, 0.1910, 0.0861, 0.1019, 0.0495, 0.0855], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0157, 0.0164, 0.0151, 0.0141, 0.0126, 0.0142, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 15:51:02,213 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 1200, loss[loss=0.1752, simple_loss=0.2652, pruned_loss=0.04262, over 16611.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2501, pruned_loss=0.04112, over 3309223.26 frames. ], batch size: 57, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:19,913 INFO [train.py:904] (3/8) Epoch 18, batch 1250, loss[loss=0.1979, simple_loss=0.2909, pruned_loss=0.05246, over 17038.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2511, pruned_loss=0.04189, over 3320621.69 frames. ], batch size: 53, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:24,777 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:53:21,075 INFO [optim.py:368] (3/8) 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 15:53:22,579 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173847.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:53:28,943 INFO [train.py:904] (3/8) Epoch 18, batch 1300, loss[loss=0.1707, simple_loss=0.2681, pruned_loss=0.03661, over 17120.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2515, pruned_loss=0.04175, over 3312601.52 frames. ], batch size: 48, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:53:32,735 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:53:46,654 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9208, 4.3079, 4.2377, 3.0469, 3.6722, 4.1631, 3.8792, 2.0902], device='cuda:3'), covar=tensor([0.0530, 0.0104, 0.0067, 0.0443, 0.0150, 0.0174, 0.0143, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0079, 0.0079, 0.0134, 0.0093, 0.0104, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:53:57,098 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2151, 3.2823, 3.4738, 2.4572, 3.2035, 3.5556, 3.2681, 2.0014], device='cuda:3'), covar=tensor([0.0508, 0.0124, 0.0055, 0.0368, 0.0106, 0.0105, 0.0101, 0.0452], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0079, 0.0079, 0.0134, 0.0093, 0.0104, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:54:05,244 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2809, 2.0813, 2.2934, 3.9225, 2.2370, 2.4170, 2.1621, 2.2447], device='cuda:3'), covar=tensor([0.1520, 0.4206, 0.2935, 0.0761, 0.4290, 0.2922, 0.4250, 0.3631], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0429, 0.0358, 0.0325, 0.0429, 0.0493, 0.0399, 0.0501], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:54:37,695 INFO [train.py:904] (3/8) Epoch 18, batch 1350, loss[loss=0.1837, simple_loss=0.2512, pruned_loss=0.0581, over 16329.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2525, pruned_loss=0.04233, over 3321027.38 frames. ], batch size: 165, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:55:37,173 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 2.229e+02 2.739e+02 3.130e+02 5.090e+02, threshold=5.477e+02, percent-clipped=0.0 2023-04-30 15:55:44,243 INFO [train.py:904] (3/8) Epoch 18, batch 1400, loss[loss=0.1602, simple_loss=0.2579, pruned_loss=0.03126, over 17075.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2515, pruned_loss=0.04182, over 3307915.37 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:55:56,062 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 15:56:04,276 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:56:09,765 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173970.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 15:56:16,576 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 15:56:25,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9509, 1.9156, 2.3978, 2.7609, 2.7330, 2.8552, 1.8450, 3.0071], device='cuda:3'), covar=tensor([0.0171, 0.0455, 0.0310, 0.0280, 0.0283, 0.0228, 0.0527, 0.0166], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0190, 0.0175, 0.0178, 0.0187, 0.0147, 0.0192, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 15:56:56,559 INFO [train.py:904] (3/8) Epoch 18, batch 1450, loss[loss=0.1649, simple_loss=0.2366, pruned_loss=0.04666, over 16231.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.25, pruned_loss=0.04131, over 3311225.07 frames. ], batch size: 165, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:57:07,338 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-30 15:57:11,461 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:14,618 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:16,987 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:41,570 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6804, 4.7291, 4.8318, 4.7187, 4.6980, 5.3122, 4.8174, 4.5118], device='cuda:3'), covar=tensor([0.1583, 0.1973, 0.2314, 0.2243, 0.3076, 0.1174, 0.1774, 0.2831], device='cuda:3'), in_proj_covar=tensor([0.0394, 0.0576, 0.0630, 0.0477, 0.0646, 0.0659, 0.0496, 0.0647], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 15:57:54,753 INFO [optim.py:368] (3/8) 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 15:58:03,188 INFO [train.py:904] (3/8) Epoch 18, batch 1500, loss[loss=0.1735, simple_loss=0.2489, pruned_loss=0.04905, over 16830.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2502, pruned_loss=0.04161, over 3308899.07 frames. ], batch size: 102, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:58:36,119 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174076.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:58:38,866 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174078.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:59:10,940 INFO [train.py:904] (3/8) Epoch 18, batch 1550, loss[loss=0.1719, simple_loss=0.2637, pruned_loss=0.04012, over 17119.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2508, pruned_loss=0.04196, over 3318219.34 frames. ], batch size: 49, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:00:00,545 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:00:11,867 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174147.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:00:17,417 INFO [train.py:904] (3/8) Epoch 18, batch 1600, loss[loss=0.1931, simple_loss=0.2641, pruned_loss=0.06108, over 16719.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2527, pruned_loss=0.04291, over 3326498.49 frames. ], batch size: 124, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:00,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 16:01:16,760 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174195.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:01:23,701 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 1650, loss[loss=0.1576, simple_loss=0.2546, pruned_loss=0.03027, over 17120.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2549, pruned_loss=0.04324, over 3333627.12 frames. ], batch size: 49, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:02:23,837 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.550e+02 2.934e+02 3.695e+02 6.371e+02, threshold=5.868e+02, percent-clipped=1.0 2023-04-30 16:02:32,758 INFO [train.py:904] (3/8) Epoch 18, batch 1700, loss[loss=0.1933, simple_loss=0.2715, pruned_loss=0.05754, over 16731.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2569, pruned_loss=0.04379, over 3321081.52 frames. ], batch size: 83, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:02:57,235 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174270.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:03:15,567 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:25,128 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:40,505 INFO [train.py:904] (3/8) Epoch 18, batch 1750, loss[loss=0.1641, simple_loss=0.2635, pruned_loss=0.03232, over 17250.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2585, pruned_loss=0.04439, over 3313741.63 frames. ], batch size: 52, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:01,306 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:36,747 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 2.402e+02 2.710e+02 3.242e+02 7.254e+02, threshold=5.419e+02, percent-clipped=1.0 2023-04-30 16:04:45,368 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9799, 5.0715, 5.5566, 5.5199, 5.5359, 5.1694, 5.1385, 4.9216], device='cuda:3'), covar=tensor([0.0319, 0.0528, 0.0346, 0.0399, 0.0465, 0.0373, 0.0894, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0438, 0.0425, 0.0398, 0.0473, 0.0449, 0.0543, 0.0355], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 16:04:46,157 INFO [train.py:904] (3/8) Epoch 18, batch 1800, loss[loss=0.1541, simple_loss=0.2405, pruned_loss=0.03384, over 17243.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2595, pruned_loss=0.04487, over 3311414.45 frames. ], batch size: 45, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:46,588 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174352.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:05:10,904 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174371.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:05:13,934 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174373.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:33,471 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:51,520 INFO [train.py:904] (3/8) Epoch 18, batch 1850, loss[loss=0.1424, simple_loss=0.2311, pruned_loss=0.02686, over 15925.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.26, pruned_loss=0.04513, over 3309753.81 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:06:28,711 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9027, 4.4060, 3.1978, 2.3426, 2.7789, 2.5825, 4.6537, 3.6741], device='cuda:3'), covar=tensor([0.2747, 0.0620, 0.1790, 0.3079, 0.3077, 0.2047, 0.0373, 0.1410], device='cuda:3'), in_proj_covar=tensor([0.0319, 0.0264, 0.0301, 0.0303, 0.0290, 0.0247, 0.0286, 0.0326], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:06:34,324 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 16:06:51,029 INFO [optim.py:368] (3/8) 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 16:06:55,534 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174449.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:06:58,383 INFO [train.py:904] (3/8) Epoch 18, batch 1900, loss[loss=0.1568, simple_loss=0.2412, pruned_loss=0.03626, over 16896.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2591, pruned_loss=0.04436, over 3311371.55 frames. ], batch size: 96, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:07:06,792 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1164, 5.7625, 5.9181, 5.5477, 5.6252, 6.2345, 5.7474, 5.4235], device='cuda:3'), covar=tensor([0.0990, 0.1624, 0.1936, 0.1920, 0.2581, 0.0938, 0.1356, 0.2367], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0581, 0.0638, 0.0481, 0.0654, 0.0666, 0.0502, 0.0652], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:07:14,824 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0563, 4.4757, 4.3902, 3.1395, 3.7269, 4.3426, 4.0236, 2.5913], device='cuda:3'), covar=tensor([0.0436, 0.0063, 0.0048, 0.0365, 0.0132, 0.0111, 0.0093, 0.0446], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0080, 0.0080, 0.0134, 0.0093, 0.0104, 0.0091, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:07:39,349 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 16:07:57,452 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:08:06,701 INFO [train.py:904] (3/8) Epoch 18, batch 1950, loss[loss=0.1646, simple_loss=0.2454, pruned_loss=0.04184, over 16914.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2594, pruned_loss=0.04387, over 3310148.19 frames. ], batch size: 96, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:08:28,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6831, 2.6085, 2.2329, 2.3971, 2.9122, 2.6968, 3.3512, 3.2002], device='cuda:3'), covar=tensor([0.0127, 0.0427, 0.0510, 0.0452, 0.0287, 0.0371, 0.0245, 0.0261], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0232, 0.0222, 0.0223, 0.0232, 0.0231, 0.0234, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:08:29,002 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 16:08:45,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6170, 4.7870, 5.1510, 5.1020, 5.1530, 4.8256, 4.6448, 4.5724], device='cuda:3'), covar=tensor([0.0485, 0.0769, 0.0446, 0.0553, 0.0665, 0.0551, 0.1350, 0.0556], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0434, 0.0420, 0.0397, 0.0470, 0.0445, 0.0540, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 16:08:58,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6198, 3.6184, 2.8292, 2.1436, 2.3693, 2.2669, 3.6295, 3.2012], device='cuda:3'), covar=tensor([0.2609, 0.0741, 0.1743, 0.3075, 0.2695, 0.2068, 0.0619, 0.1449], device='cuda:3'), in_proj_covar=tensor([0.0317, 0.0263, 0.0299, 0.0300, 0.0289, 0.0246, 0.0285, 0.0326], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:09:05,466 INFO [optim.py:368] (3/8) 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 16:09:07,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 16:09:14,567 INFO [train.py:904] (3/8) Epoch 18, batch 2000, loss[loss=0.19, simple_loss=0.2812, pruned_loss=0.04937, over 16690.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2594, pruned_loss=0.04322, over 3297568.56 frames. ], batch size: 57, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:09:50,432 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 16:10:23,409 INFO [train.py:904] (3/8) Epoch 18, batch 2050, loss[loss=0.1644, simple_loss=0.262, pruned_loss=0.03346, over 17121.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2608, pruned_loss=0.04371, over 3300354.77 frames. ], batch size: 48, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:10:31,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7814, 2.8093, 2.6882, 4.7700, 3.6587, 4.3918, 1.5646, 3.0541], device='cuda:3'), covar=tensor([0.1437, 0.0844, 0.1257, 0.0244, 0.0317, 0.0407, 0.1747, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0167, 0.0188, 0.0177, 0.0199, 0.0214, 0.0193, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:10:40,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0447, 5.1759, 5.6045, 5.5572, 5.5880, 5.2305, 5.1676, 4.9673], device='cuda:3'), covar=tensor([0.0353, 0.0547, 0.0346, 0.0439, 0.0578, 0.0363, 0.0921, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0433, 0.0420, 0.0395, 0.0469, 0.0444, 0.0538, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 16:11:17,589 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:11:26,650 INFO [optim.py:368] (3/8) 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 16:11:28,713 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174647.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:11:35,057 INFO [train.py:904] (3/8) Epoch 18, batch 2100, loss[loss=0.1685, simple_loss=0.2489, pruned_loss=0.04405, over 16808.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.26, pruned_loss=0.0436, over 3315588.13 frames. ], batch size: 102, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:12:02,037 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174671.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:12:06,132 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174673.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:12:24,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0508, 4.9747, 4.7644, 3.5028, 4.8597, 1.6188, 4.3809, 4.5458], device='cuda:3'), covar=tensor([0.0145, 0.0144, 0.0296, 0.0847, 0.0175, 0.3925, 0.0262, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0146, 0.0192, 0.0174, 0.0167, 0.0202, 0.0182, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:12:45,802 INFO [train.py:904] (3/8) Epoch 18, batch 2150, loss[loss=0.1465, simple_loss=0.2339, pruned_loss=0.02958, over 16872.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2602, pruned_loss=0.04408, over 3312249.11 frames. ], batch size: 42, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:12:50,384 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174705.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:10,638 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:13,522 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:44,869 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 16:13:45,424 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:13:56,942 INFO [train.py:904] (3/8) Epoch 18, batch 2200, loss[loss=0.1474, simple_loss=0.2362, pruned_loss=0.02932, over 17202.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2607, pruned_loss=0.04457, over 3312878.52 frames. ], batch size: 43, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:14:16,317 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:14:56,018 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:15:06,367 INFO [train.py:904] (3/8) Epoch 18, batch 2250, loss[loss=0.1561, simple_loss=0.2446, pruned_loss=0.03382, over 16856.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2617, pruned_loss=0.04542, over 3317675.47 frames. ], batch size: 42, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:15:10,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9370, 4.8965, 4.7475, 4.2896, 4.8663, 1.9017, 4.5682, 4.6002], device='cuda:3'), covar=tensor([0.0145, 0.0100, 0.0230, 0.0322, 0.0108, 0.2714, 0.0163, 0.0193], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0146, 0.0192, 0.0173, 0.0167, 0.0202, 0.0181, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:15:25,061 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4138, 5.3915, 5.2890, 4.7742, 4.8697, 5.3309, 5.2453, 4.9103], device='cuda:3'), covar=tensor([0.0574, 0.0421, 0.0244, 0.0308, 0.0993, 0.0419, 0.0229, 0.0675], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0411, 0.0340, 0.0331, 0.0354, 0.0383, 0.0236, 0.0410], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:16:02,772 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:16:07,687 INFO [optim.py:368] (3/8) 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 16:16:14,118 INFO [train.py:904] (3/8) Epoch 18, batch 2300, loss[loss=0.2708, simple_loss=0.3352, pruned_loss=0.1033, over 12038.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2619, pruned_loss=0.04548, over 3315238.32 frames. ], batch size: 246, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:17:22,134 INFO [train.py:904] (3/8) Epoch 18, batch 2350, loss[loss=0.1639, simple_loss=0.2617, pruned_loss=0.033, over 17021.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2621, pruned_loss=0.04507, over 3324653.32 frames. ], batch size: 50, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:18:14,600 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:18:23,374 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 16:18:25,281 INFO [optim.py:368] (3/8) 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 16:18:25,679 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174947.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 16:18:32,022 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4200, 3.4138, 3.3823, 2.6130, 3.2707, 1.9870, 2.9029, 2.6536], device='cuda:3'), covar=tensor([0.0169, 0.0131, 0.0236, 0.0275, 0.0118, 0.2685, 0.0157, 0.0342], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0147, 0.0194, 0.0175, 0.0169, 0.0204, 0.0183, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:18:32,765 INFO [train.py:904] (3/8) Epoch 18, batch 2400, loss[loss=0.2385, simple_loss=0.3154, pruned_loss=0.08083, over 11382.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2637, pruned_loss=0.04575, over 3312049.73 frames. ], batch size: 246, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:19:20,826 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:33,333 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 2450, loss[loss=0.1818, simple_loss=0.2873, pruned_loss=0.03818, over 17268.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2634, pruned_loss=0.0456, over 3315919.81 frames. ], batch size: 52, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:20:12,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7628, 1.7944, 2.3490, 2.6704, 2.6627, 2.7433, 1.8988, 2.9453], device='cuda:3'), covar=tensor([0.0172, 0.0484, 0.0316, 0.0268, 0.0269, 0.0261, 0.0507, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:20:24,404 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0045, 4.9172, 4.8952, 4.4847, 4.5404, 4.9257, 4.7962, 4.6232], device='cuda:3'), covar=tensor([0.0618, 0.0633, 0.0294, 0.0333, 0.0998, 0.0473, 0.0358, 0.0667], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0413, 0.0341, 0.0333, 0.0355, 0.0385, 0.0236, 0.0411], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:20:40,479 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 18, batch 2500, loss[loss=0.1546, simple_loss=0.2479, pruned_loss=0.03069, over 17142.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2632, pruned_loss=0.0451, over 3319207.66 frames. ], batch size: 48, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:21:04,687 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 18, batch 2550, loss[loss=0.1585, simple_loss=0.2377, pruned_loss=0.03963, over 16834.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2638, pruned_loss=0.04523, over 3326751.92 frames. ], batch size: 102, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:22:30,179 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3599, 2.2519, 2.3626, 4.1757, 2.2470, 2.6772, 2.3099, 2.4576], device='cuda:3'), covar=tensor([0.1354, 0.3539, 0.2849, 0.0538, 0.3876, 0.2448, 0.3640, 0.3081], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0432, 0.0361, 0.0327, 0.0432, 0.0500, 0.0401, 0.0505], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:22:34,745 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 16:22:51,979 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:23:09,303 INFO [train.py:904] (3/8) Epoch 18, batch 2600, loss[loss=0.1581, simple_loss=0.2496, pruned_loss=0.03328, over 17021.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2631, pruned_loss=0.04496, over 3325164.53 frames. ], batch size: 53, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:17,693 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 2650, loss[loss=0.1562, simple_loss=0.2478, pruned_loss=0.03224, over 17241.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2626, pruned_loss=0.04385, over 3332510.21 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:27,556 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:25:22,162 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 2700, loss[loss=0.174, simple_loss=0.2598, pruned_loss=0.04412, over 16412.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2624, pruned_loss=0.04332, over 3334686.49 frames. ], batch size: 146, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:25:50,104 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175268.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:26:36,861 INFO [train.py:904] (3/8) Epoch 18, batch 2750, loss[loss=0.1677, simple_loss=0.2502, pruned_loss=0.04254, over 16836.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2628, pruned_loss=0.0434, over 3341362.30 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:27:40,024 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 2800, loss[loss=0.1765, simple_loss=0.2699, pruned_loss=0.0416, over 17084.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2629, pruned_loss=0.04344, over 3335751.91 frames. ], batch size: 49, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:27:58,787 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:11,808 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:44,387 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2600, 5.1609, 5.0659, 4.5870, 4.6802, 5.1220, 5.0994, 4.7705], device='cuda:3'), covar=tensor([0.0575, 0.0504, 0.0308, 0.0345, 0.1127, 0.0469, 0.0343, 0.0707], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0416, 0.0344, 0.0336, 0.0359, 0.0389, 0.0238, 0.0415], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:28:45,620 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 2850, loss[loss=0.2462, simple_loss=0.3118, pruned_loss=0.09037, over 11839.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2623, pruned_loss=0.04402, over 3324875.37 frames. ], batch size: 246, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:29:05,422 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175409.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:05,657 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9772, 2.0249, 2.5677, 2.8883, 2.7019, 3.3950, 2.1838, 3.3897], device='cuda:3'), covar=tensor([0.0228, 0.0501, 0.0296, 0.0310, 0.0308, 0.0182, 0.0508, 0.0163], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:29:35,302 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:30:04,918 INFO [train.py:904] (3/8) Epoch 18, batch 2900, loss[loss=0.1882, simple_loss=0.2507, pruned_loss=0.06283, over 16751.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2608, pruned_loss=0.04359, over 3331019.21 frames. ], batch size: 124, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:30:08,950 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175455.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:45,719 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:50,787 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8347, 2.8209, 2.3531, 2.6700, 3.1123, 2.9229, 3.4973, 3.3274], device='cuda:3'), covar=tensor([0.0140, 0.0364, 0.0489, 0.0424, 0.0265, 0.0362, 0.0256, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0235, 0.0225, 0.0225, 0.0236, 0.0234, 0.0240, 0.0228], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:31:04,471 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 2950, loss[loss=0.1905, simple_loss=0.2697, pruned_loss=0.0556, over 16734.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2608, pruned_loss=0.04481, over 3313936.06 frames. ], batch size: 124, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:31:50,283 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9810, 2.0797, 2.5754, 2.9511, 2.8131, 3.3821, 2.1881, 3.4371], device='cuda:3'), covar=tensor([0.0238, 0.0477, 0.0322, 0.0294, 0.0310, 0.0169, 0.0485, 0.0144], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0185, 0.0192, 0.0150, 0.0195, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:32:09,671 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:32:23,313 INFO [train.py:904] (3/8) Epoch 18, batch 3000, loss[loss=0.1616, simple_loss=0.2499, pruned_loss=0.03667, over 16839.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2615, pruned_loss=0.04498, over 3315595.50 frames. ], batch size: 42, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:23,314 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 16:32:32,125 INFO [train.py:938] (3/8) Epoch 18, validation: loss=0.1359, simple_loss=0.2415, pruned_loss=0.01516, over 944034.00 frames. 2023-04-30 16:32:32,126 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 16:32:48,740 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175563.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:33:12,011 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 16:33:42,215 INFO [train.py:904] (3/8) Epoch 18, batch 3050, loss[loss=0.1548, simple_loss=0.2467, pruned_loss=0.03141, over 17233.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2606, pruned_loss=0.04478, over 3316541.56 frames. ], batch size: 46, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:34:38,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1733, 2.6352, 2.1432, 2.4304, 2.9574, 2.6991, 3.1415, 3.0982], device='cuda:3'), covar=tensor([0.0179, 0.0357, 0.0505, 0.0409, 0.0217, 0.0318, 0.0225, 0.0251], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0234, 0.0224, 0.0223, 0.0235, 0.0233, 0.0238, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:34:46,185 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.265e+02 2.697e+02 3.394e+02 4.892e+02, threshold=5.395e+02, percent-clipped=0.0 2023-04-30 16:34:51,893 INFO [train.py:904] (3/8) Epoch 18, batch 3100, loss[loss=0.1782, simple_loss=0.2651, pruned_loss=0.04568, over 12721.00 frames. ], tot_loss[loss=0.174, simple_loss=0.26, pruned_loss=0.04402, over 3325032.63 frames. ], batch size: 247, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:01,369 INFO [train.py:904] (3/8) Epoch 18, batch 3150, loss[loss=0.1841, simple_loss=0.2655, pruned_loss=0.0514, over 16275.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2591, pruned_loss=0.04406, over 3314450.91 frames. ], batch size: 165, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:02,075 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 16:36:13,001 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175710.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:36,341 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:37:06,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2696, 5.2073, 5.0877, 4.6275, 4.7197, 5.1420, 5.1216, 4.7562], device='cuda:3'), covar=tensor([0.0620, 0.0492, 0.0299, 0.0354, 0.1176, 0.0458, 0.0292, 0.0816], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0423, 0.0349, 0.0341, 0.0365, 0.0396, 0.0241, 0.0421], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:37:09,267 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 3200, loss[loss=0.1636, simple_loss=0.2534, pruned_loss=0.03691, over 17222.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2582, pruned_loss=0.04329, over 3323273.94 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:37:17,453 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4370, 4.4197, 4.3456, 4.1367, 4.0921, 4.4540, 4.2115, 4.1831], device='cuda:3'), covar=tensor([0.0725, 0.0677, 0.0384, 0.0307, 0.0887, 0.0496, 0.0599, 0.0704], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0422, 0.0349, 0.0341, 0.0365, 0.0395, 0.0241, 0.0421], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:37:37,877 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175771.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:38,986 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8955, 4.0833, 2.5435, 4.6385, 3.1770, 4.5849, 2.5998, 3.2619], device='cuda:3'), covar=tensor([0.0270, 0.0348, 0.1484, 0.0215, 0.0727, 0.0447, 0.1473, 0.0679], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0178, 0.0196, 0.0163, 0.0177, 0.0222, 0.0205, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:37:56,524 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9798, 4.4945, 4.4235, 3.1993, 3.7373, 4.4356, 3.9852, 2.6657], device='cuda:3'), covar=tensor([0.0434, 0.0057, 0.0045, 0.0345, 0.0117, 0.0085, 0.0084, 0.0427], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0132, 0.0094, 0.0104, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:38:11,054 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:13,146 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8421, 3.0266, 2.7949, 5.0098, 4.1096, 4.4437, 1.8246, 3.1172], device='cuda:3'), covar=tensor([0.1527, 0.0762, 0.1182, 0.0219, 0.0249, 0.0381, 0.1675, 0.0838], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0180, 0.0203, 0.0216, 0.0193, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:38:20,675 INFO [train.py:904] (3/8) Epoch 18, batch 3250, loss[loss=0.222, simple_loss=0.2956, pruned_loss=0.07414, over 16742.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2587, pruned_loss=0.04363, over 3325892.86 frames. ], batch size: 134, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:09,521 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:17,730 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:23,829 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.233e+02 2.626e+02 2.962e+02 5.772e+02, threshold=5.253e+02, percent-clipped=1.0 2023-04-30 16:39:29,873 INFO [train.py:904] (3/8) Epoch 18, batch 3300, loss[loss=0.1703, simple_loss=0.2591, pruned_loss=0.04079, over 17122.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2598, pruned_loss=0.04405, over 3324896.95 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:31,548 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7043, 3.6738, 4.0340, 2.0791, 4.1796, 4.1479, 3.2431, 2.9796], device='cuda:3'), covar=tensor([0.0756, 0.0196, 0.0137, 0.1177, 0.0075, 0.0172, 0.0333, 0.0429], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0137, 0.0077, 0.0122, 0.0125, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 16:39:45,145 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:59,580 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0885, 4.1506, 4.5076, 4.4537, 4.4980, 4.1698, 4.2382, 4.0931], device='cuda:3'), covar=tensor([0.0411, 0.0688, 0.0412, 0.0447, 0.0480, 0.0467, 0.0839, 0.0655], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0441, 0.0431, 0.0404, 0.0477, 0.0453, 0.0552, 0.0359], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 16:40:10,934 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:40:38,822 INFO [train.py:904] (3/8) Epoch 18, batch 3350, loss[loss=0.1438, simple_loss=0.2271, pruned_loss=0.03027, over 16178.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2598, pruned_loss=0.04368, over 3316052.35 frames. ], batch size: 36, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:40:50,862 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:34,608 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:38,077 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:41:47,204 INFO [train.py:904] (3/8) Epoch 18, batch 3400, loss[loss=0.174, simple_loss=0.267, pruned_loss=0.04047, over 17137.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2592, pruned_loss=0.04334, over 3322636.26 frames. ], batch size: 48, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:42:24,136 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2978, 5.6126, 5.3511, 5.3931, 5.0610, 5.0197, 5.0727, 5.7118], device='cuda:3'), covar=tensor([0.1196, 0.0853, 0.0916, 0.0836, 0.0847, 0.0791, 0.1079, 0.0912], device='cuda:3'), in_proj_covar=tensor([0.0657, 0.0812, 0.0654, 0.0600, 0.0509, 0.0512, 0.0672, 0.0627], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:43:01,859 INFO [train.py:904] (3/8) Epoch 18, batch 3450, loss[loss=0.1812, simple_loss=0.275, pruned_loss=0.04375, over 17042.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2577, pruned_loss=0.04271, over 3324487.90 frames. ], batch size: 53, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:43:07,999 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176006.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:43:35,465 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:43:53,047 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1405, 2.4598, 2.5847, 1.9674, 2.8086, 2.7368, 2.4339, 2.3963], device='cuda:3'), covar=tensor([0.0669, 0.0264, 0.0259, 0.0872, 0.0109, 0.0251, 0.0439, 0.0397], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0107, 0.0096, 0.0139, 0.0078, 0.0124, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:44:06,111 INFO [optim.py:368] (3/8) 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 16:44:08,605 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 16:44:09,539 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:11,308 INFO [train.py:904] (3/8) Epoch 18, batch 3500, loss[loss=0.1952, simple_loss=0.2912, pruned_loss=0.04959, over 16561.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2576, pruned_loss=0.0431, over 3313205.41 frames. ], batch size: 62, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:44:31,793 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176074.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:09,982 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 16:45:14,940 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176098.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:20,519 INFO [train.py:904] (3/8) Epoch 18, batch 3550, loss[loss=0.1936, simple_loss=0.2644, pruned_loss=0.06144, over 11708.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2558, pruned_loss=0.04262, over 3304094.84 frames. ], batch size: 246, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:45:31,590 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3033, 4.2096, 4.5005, 2.5195, 4.7489, 4.7826, 3.5182, 3.9645], device='cuda:3'), covar=tensor([0.0675, 0.0225, 0.0222, 0.1079, 0.0116, 0.0201, 0.0389, 0.0330], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0108, 0.0096, 0.0140, 0.0078, 0.0125, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:46:07,643 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9113, 4.3844, 4.3583, 3.2529, 3.6288, 4.3148, 3.9074, 2.4674], device='cuda:3'), covar=tensor([0.0437, 0.0058, 0.0045, 0.0309, 0.0119, 0.0087, 0.0083, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0080, 0.0080, 0.0133, 0.0094, 0.0104, 0.0092, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:46:10,584 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 18, batch 3600, loss[loss=0.1652, simple_loss=0.2616, pruned_loss=0.0344, over 17058.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2552, pruned_loss=0.04242, over 3302552.47 frames. ], batch size: 50, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:47:18,727 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:47:27,494 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4555, 3.1837, 3.4135, 1.9485, 3.6431, 3.5861, 2.9292, 2.6643], device='cuda:3'), covar=tensor([0.0697, 0.0224, 0.0203, 0.1058, 0.0087, 0.0167, 0.0398, 0.0403], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0107, 0.0096, 0.0140, 0.0078, 0.0124, 0.0126, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:47:43,325 INFO [train.py:904] (3/8) Epoch 18, batch 3650, loss[loss=0.1599, simple_loss=0.2393, pruned_loss=0.04027, over 16708.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2543, pruned_loss=0.04272, over 3304521.48 frames. ], batch size: 134, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:48:14,883 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6886, 4.7342, 4.8622, 4.7732, 4.7210, 5.3738, 4.8545, 4.5430], device='cuda:3'), covar=tensor([0.1430, 0.2110, 0.2042, 0.2027, 0.2824, 0.1027, 0.1671, 0.2442], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0589, 0.0645, 0.0495, 0.0666, 0.0677, 0.0511, 0.0661], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 16:48:35,857 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176237.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:43,911 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:53,036 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 3700, loss[loss=0.168, simple_loss=0.2407, pruned_loss=0.04768, over 16907.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2528, pruned_loss=0.04395, over 3278349.19 frames. ], batch size: 116, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:49:05,194 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5119, 4.3614, 4.5897, 4.7514, 4.8546, 4.4360, 4.6646, 4.8246], device='cuda:3'), covar=tensor([0.1651, 0.1096, 0.1280, 0.0632, 0.0622, 0.1046, 0.2181, 0.0758], device='cuda:3'), in_proj_covar=tensor([0.0640, 0.0795, 0.0936, 0.0809, 0.0596, 0.0637, 0.0654, 0.0757], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:49:06,401 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:49:14,288 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9970, 2.3709, 2.3522, 2.4664, 2.0687, 3.1677, 1.8986, 2.6509], device='cuda:3'), covar=tensor([0.1100, 0.0661, 0.1094, 0.0169, 0.0138, 0.0353, 0.1330, 0.0735], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0181, 0.0203, 0.0215, 0.0193, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:50:10,766 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 3750, loss[loss=0.189, simple_loss=0.2532, pruned_loss=0.06243, over 16897.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2535, pruned_loss=0.04519, over 3279089.07 frames. ], batch size: 116, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:50:12,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8560, 2.0054, 2.4303, 2.6887, 2.7910, 2.6872, 1.9141, 3.0020], device='cuda:3'), covar=tensor([0.0174, 0.0425, 0.0301, 0.0255, 0.0245, 0.0271, 0.0495, 0.0134], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0188, 0.0176, 0.0181, 0.0188, 0.0149, 0.0192, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:50:13,732 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:35,174 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:53,959 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4169, 3.3555, 3.4422, 3.5144, 3.5522, 3.3004, 3.5165, 3.6174], device='cuda:3'), covar=tensor([0.1223, 0.0857, 0.1030, 0.0622, 0.0575, 0.2295, 0.1308, 0.0679], device='cuda:3'), in_proj_covar=tensor([0.0635, 0.0789, 0.0929, 0.0803, 0.0591, 0.0633, 0.0648, 0.0753], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:51:17,334 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 3800, loss[loss=0.1951, simple_loss=0.2709, pruned_loss=0.05968, over 16304.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2551, pruned_loss=0.047, over 3260631.36 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:51:44,171 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:51:58,905 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7694, 4.6590, 4.8493, 5.0186, 5.1768, 4.6785, 5.1398, 5.1699], device='cuda:3'), covar=tensor([0.1795, 0.1098, 0.1493, 0.0680, 0.0490, 0.0885, 0.0601, 0.0527], device='cuda:3'), in_proj_covar=tensor([0.0636, 0.0790, 0.0928, 0.0803, 0.0590, 0.0634, 0.0649, 0.0753], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:52:35,499 INFO [train.py:904] (3/8) Epoch 18, batch 3850, loss[loss=0.1684, simple_loss=0.2546, pruned_loss=0.04112, over 16567.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.256, pruned_loss=0.04793, over 3254873.22 frames. ], batch size: 62, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:52:53,510 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176414.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:53:42,933 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.342e+02 2.782e+02 3.184e+02 4.336e+02, threshold=5.564e+02, percent-clipped=0.0 2023-04-30 16:53:48,840 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8799, 2.9931, 2.7713, 4.5507, 3.7404, 4.2150, 1.7676, 2.9863], device='cuda:3'), covar=tensor([0.1293, 0.0682, 0.1077, 0.0150, 0.0302, 0.0387, 0.1485, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0182, 0.0204, 0.0215, 0.0194, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:53:49,440 INFO [train.py:904] (3/8) Epoch 18, batch 3900, loss[loss=0.1709, simple_loss=0.2486, pruned_loss=0.04661, over 16502.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2556, pruned_loss=0.04805, over 3250462.36 frames. ], batch size: 68, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:53:53,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 16:54:11,186 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:00,904 INFO [train.py:904] (3/8) Epoch 18, batch 3950, loss[loss=0.1633, simple_loss=0.2375, pruned_loss=0.04457, over 16761.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2553, pruned_loss=0.04903, over 3266281.62 frames. ], batch size: 124, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:55:35,296 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176527.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:50,449 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.227e+02 2.521e+02 3.093e+02 7.043e+02, threshold=5.041e+02, percent-clipped=2.0 2023-04-30 16:56:12,325 INFO [train.py:904] (3/8) Epoch 18, batch 4000, loss[loss=0.1642, simple_loss=0.2482, pruned_loss=0.04013, over 17222.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2553, pruned_loss=0.0492, over 3265615.47 frames. ], batch size: 43, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:00,441 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:18,434 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:22,921 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176601.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:24,388 INFO [train.py:904] (3/8) Epoch 18, batch 4050, loss[loss=0.1534, simple_loss=0.238, pruned_loss=0.03444, over 16454.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2559, pruned_loss=0.04845, over 3271943.34 frames. ], batch size: 146, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:32,897 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2044, 2.2130, 2.7937, 3.1982, 3.0467, 3.6454, 2.2257, 3.5364], device='cuda:3'), covar=tensor([0.0173, 0.0411, 0.0252, 0.0233, 0.0246, 0.0125, 0.0459, 0.0094], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0188, 0.0176, 0.0180, 0.0187, 0.0148, 0.0191, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:57:39,760 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:58,519 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:12,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5100, 4.8049, 4.5965, 4.5949, 4.3560, 4.2630, 4.2265, 4.8839], device='cuda:3'), covar=tensor([0.1181, 0.0894, 0.0975, 0.0786, 0.0772, 0.1401, 0.1086, 0.0905], device='cuda:3'), in_proj_covar=tensor([0.0650, 0.0803, 0.0651, 0.0599, 0.0503, 0.0512, 0.0668, 0.0622], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:58:30,486 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.366e+02 1.835e+02 2.096e+02 2.511e+02 3.496e+02, threshold=4.192e+02, percent-clipped=0.0 2023-04-30 16:58:33,594 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 4100, loss[loss=0.2191, simple_loss=0.2936, pruned_loss=0.07227, over 11735.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2568, pruned_loss=0.04764, over 3255940.52 frames. ], batch size: 247, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:58:57,243 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6800, 4.6714, 4.5104, 3.8331, 4.5855, 1.7198, 4.3418, 4.1371], device='cuda:3'), covar=tensor([0.0069, 0.0055, 0.0144, 0.0292, 0.0060, 0.2760, 0.0101, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0148, 0.0196, 0.0178, 0.0170, 0.0204, 0.0186, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 16:58:57,272 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:59,713 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5991, 2.6291, 2.3925, 3.5573, 2.9159, 3.7755, 1.4881, 2.6669], device='cuda:3'), covar=tensor([0.1374, 0.0748, 0.1253, 0.0184, 0.0237, 0.0415, 0.1646, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0182, 0.0205, 0.0215, 0.0195, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 16:59:28,854 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 4150, loss[loss=0.1735, simple_loss=0.2618, pruned_loss=0.04265, over 16420.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2634, pruned_loss=0.04976, over 3227017.18 frames. ], batch size: 68, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:00:27,860 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:01:00,610 INFO [optim.py:368] (3/8) 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 17:01:06,262 INFO [train.py:904] (3/8) Epoch 18, batch 4200, loss[loss=0.2407, simple_loss=0.3244, pruned_loss=0.07849, over 15465.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2709, pruned_loss=0.05167, over 3201019.35 frames. ], batch size: 190, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:01:06,909 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4946, 3.6179, 3.5924, 2.1459, 2.9494, 2.2573, 3.7728, 3.8136], device='cuda:3'), covar=tensor([0.0247, 0.0791, 0.0607, 0.1906, 0.0910, 0.0987, 0.0717, 0.0956], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0164, 0.0150, 0.0142, 0.0126, 0.0142, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:02:21,462 INFO [train.py:904] (3/8) Epoch 18, batch 4250, loss[loss=0.1904, simple_loss=0.2731, pruned_loss=0.05385, over 12148.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.274, pruned_loss=0.05147, over 3182974.84 frames. ], batch size: 246, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:51,598 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:03:15,956 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9888, 3.0227, 1.8124, 3.2154, 2.2968, 3.2855, 2.1099, 2.5148], device='cuda:3'), covar=tensor([0.0330, 0.0371, 0.1627, 0.0182, 0.0821, 0.0449, 0.1451, 0.0790], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0175, 0.0192, 0.0156, 0.0174, 0.0216, 0.0200, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:03:30,835 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 4300, loss[loss=0.2025, simple_loss=0.2976, pruned_loss=0.05371, over 16257.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2756, pruned_loss=0.05091, over 3192416.28 frames. ], batch size: 165, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:03:37,836 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:44,723 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176898.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:49,704 INFO [train.py:904] (3/8) Epoch 18, batch 4350, loss[loss=0.2165, simple_loss=0.3002, pruned_loss=0.0664, over 17067.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2788, pruned_loss=0.05195, over 3175488.61 frames. ], batch size: 53, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:04:56,564 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8742, 5.1445, 5.3950, 5.1675, 5.2436, 5.7900, 5.2448, 4.9095], device='cuda:3'), covar=tensor([0.0893, 0.1688, 0.1662, 0.1710, 0.2322, 0.0797, 0.1405, 0.2182], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0572, 0.0624, 0.0478, 0.0642, 0.0654, 0.0497, 0.0640], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:05:06,480 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176913.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:07,702 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176914.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:05:55,204 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:06:00,547 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9312, 3.2600, 3.2060, 2.0918, 3.0226, 3.1780, 2.9858, 1.8993], device='cuda:3'), covar=tensor([0.0488, 0.0048, 0.0058, 0.0391, 0.0096, 0.0093, 0.0107, 0.0424], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0078, 0.0079, 0.0132, 0.0094, 0.0104, 0.0091, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:06:03,298 INFO [train.py:904] (3/8) Epoch 18, batch 4400, loss[loss=0.214, simple_loss=0.2946, pruned_loss=0.06672, over 16956.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2813, pruned_loss=0.05333, over 3189930.37 frames. ], batch size: 109, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:06:17,358 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:06:45,569 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:07:06,072 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8684, 1.9460, 2.4540, 2.8292, 2.6757, 3.1810, 1.9241, 3.0534], device='cuda:3'), covar=tensor([0.0166, 0.0431, 0.0240, 0.0221, 0.0259, 0.0136, 0.0474, 0.0107], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0188, 0.0177, 0.0180, 0.0188, 0.0148, 0.0192, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:07:16,203 INFO [train.py:904] (3/8) Epoch 18, batch 4450, loss[loss=0.2254, simple_loss=0.3291, pruned_loss=0.06084, over 16230.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2851, pruned_loss=0.05483, over 3189352.96 frames. ], batch size: 165, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:07:29,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4968, 3.7045, 2.8875, 2.1933, 2.4641, 2.3484, 3.9896, 3.2844], device='cuda:3'), covar=tensor([0.2978, 0.0610, 0.1636, 0.2544, 0.2523, 0.1942, 0.0403, 0.1156], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0266, 0.0301, 0.0304, 0.0294, 0.0248, 0.0288, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:07:43,890 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:08:24,607 INFO [optim.py:368] (3/8) 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 17:08:30,153 INFO [train.py:904] (3/8) Epoch 18, batch 4500, loss[loss=0.188, simple_loss=0.2703, pruned_loss=0.05283, over 16469.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2849, pruned_loss=0.05508, over 3194189.09 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:43,612 INFO [train.py:904] (3/8) Epoch 18, batch 4550, loss[loss=0.2122, simple_loss=0.2978, pruned_loss=0.06327, over 15333.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2859, pruned_loss=0.05591, over 3208224.13 frames. ], batch size: 190, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:52,989 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177109.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:09:55,569 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:11,713 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:32,723 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 17:10:45,846 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7396, 1.8361, 2.3243, 2.6553, 2.5996, 2.9975, 1.8605, 2.9167], device='cuda:3'), covar=tensor([0.0166, 0.0463, 0.0258, 0.0251, 0.0268, 0.0143, 0.0511, 0.0108], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0181, 0.0189, 0.0148, 0.0192, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:10:48,535 INFO [optim.py:368] (3/8) 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 17:10:53,050 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0404, 5.1313, 5.4615, 5.4358, 5.4969, 5.0401, 5.0545, 4.7565], device='cuda:3'), covar=tensor([0.0268, 0.0336, 0.0302, 0.0321, 0.0447, 0.0332, 0.0941, 0.0428], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0413, 0.0405, 0.0380, 0.0450, 0.0426, 0.0519, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:10:53,843 INFO [train.py:904] (3/8) Epoch 18, batch 4600, loss[loss=0.1937, simple_loss=0.2838, pruned_loss=0.05186, over 17172.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2864, pruned_loss=0.0561, over 3205238.97 frames. ], batch size: 46, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:11:16,520 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1268, 3.0770, 1.8812, 3.4721, 2.4201, 3.4748, 2.0605, 2.6104], device='cuda:3'), covar=tensor([0.0296, 0.0412, 0.1652, 0.0149, 0.0843, 0.0479, 0.1453, 0.0750], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0174, 0.0192, 0.0153, 0.0173, 0.0214, 0.0200, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:11:19,982 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:20,118 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:22,868 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9528, 2.3651, 2.3869, 2.5484, 2.0146, 3.1844, 1.7868, 2.6098], device='cuda:3'), covar=tensor([0.1172, 0.0652, 0.0974, 0.0156, 0.0151, 0.0357, 0.1396, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0171, 0.0192, 0.0182, 0.0206, 0.0216, 0.0197, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:11:24,088 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:12:05,777 INFO [train.py:904] (3/8) Epoch 18, batch 4650, loss[loss=0.1785, simple_loss=0.2643, pruned_loss=0.04633, over 16469.00 frames. ], tot_loss[loss=0.199, simple_loss=0.286, pruned_loss=0.05603, over 3206075.94 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:12:16,098 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177209.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:13:10,932 INFO [optim.py:368] (3/8) 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 17:13:16,216 INFO [train.py:904] (3/8) Epoch 18, batch 4700, loss[loss=0.2014, simple_loss=0.2918, pruned_loss=0.05556, over 16658.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2832, pruned_loss=0.05488, over 3194185.52 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:13:59,447 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:14:17,188 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2952, 4.1943, 4.3673, 4.5285, 4.6669, 4.2513, 4.5824, 4.6789], device='cuda:3'), covar=tensor([0.1640, 0.1075, 0.1433, 0.0644, 0.0495, 0.1092, 0.0726, 0.0513], device='cuda:3'), in_proj_covar=tensor([0.0603, 0.0744, 0.0875, 0.0763, 0.0560, 0.0601, 0.0613, 0.0709], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:14:28,060 INFO [train.py:904] (3/8) Epoch 18, batch 4750, loss[loss=0.1645, simple_loss=0.2484, pruned_loss=0.04026, over 17121.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2784, pruned_loss=0.05252, over 3202043.47 frames. ], batch size: 49, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:14:56,077 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:15:27,348 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8977, 3.9653, 4.5109, 2.0412, 4.8925, 4.7933, 3.3187, 3.4475], device='cuda:3'), covar=tensor([0.0780, 0.0243, 0.0134, 0.1243, 0.0041, 0.0077, 0.0384, 0.0432], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0107, 0.0095, 0.0139, 0.0076, 0.0123, 0.0126, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:15:34,634 INFO [optim.py:368] (3/8) 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 17:15:40,560 INFO [train.py:904] (3/8) Epoch 18, batch 4800, loss[loss=0.19, simple_loss=0.2783, pruned_loss=0.05081, over 16837.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2747, pruned_loss=0.05041, over 3214005.66 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:15:47,119 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 17:16:07,490 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177369.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:16:57,034 INFO [train.py:904] (3/8) Epoch 18, batch 4850, loss[loss=0.1719, simple_loss=0.2596, pruned_loss=0.04207, over 17118.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2749, pruned_loss=0.04936, over 3214040.22 frames. ], batch size: 47, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:17:19,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8799, 4.9512, 5.2716, 5.2543, 5.3087, 4.8791, 4.8639, 4.5492], device='cuda:3'), covar=tensor([0.0262, 0.0476, 0.0345, 0.0355, 0.0471, 0.0358, 0.1001, 0.0432], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0412, 0.0406, 0.0380, 0.0451, 0.0425, 0.0520, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:17:28,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0552, 3.0018, 1.7899, 3.2051, 2.2741, 3.2623, 2.0021, 2.6001], device='cuda:3'), covar=tensor([0.0278, 0.0399, 0.1638, 0.0142, 0.0891, 0.0527, 0.1582, 0.0671], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0188, 0.0150, 0.0171, 0.0210, 0.0197, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:17:36,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5684, 3.5005, 3.5335, 2.7572, 3.3209, 1.9717, 3.0943, 2.8127], device='cuda:3'), covar=tensor([0.0152, 0.0162, 0.0150, 0.0275, 0.0107, 0.2486, 0.0148, 0.0267], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0144, 0.0188, 0.0173, 0.0163, 0.0199, 0.0179, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:18:05,879 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.952e+02 2.254e+02 2.679e+02 4.935e+02, threshold=4.508e+02, percent-clipped=3.0 2023-04-30 17:18:11,208 INFO [train.py:904] (3/8) Epoch 18, batch 4900, loss[loss=0.1718, simple_loss=0.2596, pruned_loss=0.04198, over 16470.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2743, pruned_loss=0.0481, over 3202040.73 frames. ], batch size: 75, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:18:29,119 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4579, 3.4908, 2.0897, 3.9784, 2.5984, 3.9279, 2.3029, 2.8141], device='cuda:3'), covar=tensor([0.0304, 0.0375, 0.1683, 0.0119, 0.0902, 0.0490, 0.1550, 0.0750], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0170, 0.0188, 0.0149, 0.0170, 0.0209, 0.0196, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:18:30,786 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177467.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:18:44,930 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 17:19:07,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 17:19:25,468 INFO [train.py:904] (3/8) Epoch 18, batch 4950, loss[loss=0.2044, simple_loss=0.298, pruned_loss=0.05544, over 17014.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2745, pruned_loss=0.04779, over 3207941.01 frames. ], batch size: 55, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:19:34,717 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177509.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:20:14,881 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3135, 4.1717, 4.3814, 4.5357, 4.7154, 4.2933, 4.6421, 4.7192], device='cuda:3'), covar=tensor([0.1711, 0.1276, 0.1518, 0.0727, 0.0504, 0.1036, 0.0700, 0.0543], device='cuda:3'), in_proj_covar=tensor([0.0600, 0.0743, 0.0872, 0.0761, 0.0558, 0.0598, 0.0611, 0.0706], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:20:30,520 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 5000, loss[loss=0.2167, simple_loss=0.3089, pruned_loss=0.06222, over 11809.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2758, pruned_loss=0.04802, over 3211135.42 frames. ], batch size: 247, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:20:37,350 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177553.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:20:41,848 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:21:47,170 INFO [train.py:904] (3/8) Epoch 18, batch 5050, loss[loss=0.196, simple_loss=0.2815, pruned_loss=0.05531, over 17041.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2758, pruned_loss=0.0476, over 3233529.66 frames. ], batch size: 53, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:21:49,819 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9168, 5.2729, 5.5000, 5.2086, 5.2938, 5.8716, 5.3594, 5.0839], device='cuda:3'), covar=tensor([0.0961, 0.1658, 0.1673, 0.1875, 0.2343, 0.0840, 0.1235, 0.2081], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0562, 0.0615, 0.0474, 0.0633, 0.0645, 0.0485, 0.0635], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:22:04,614 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.030e+02 2.394e+02 2.756e+02 4.573e+02, threshold=4.789e+02, percent-clipped=0.0 2023-04-30 17:22:58,430 INFO [train.py:904] (3/8) Epoch 18, batch 5100, loss[loss=0.1647, simple_loss=0.2518, pruned_loss=0.03885, over 16735.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2743, pruned_loss=0.04699, over 3231645.27 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:23:16,201 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177664.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:23:40,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2661, 5.2498, 5.0922, 4.5935, 4.6855, 5.0976, 5.1131, 4.8557], device='cuda:3'), covar=tensor([0.0575, 0.0343, 0.0313, 0.0344, 0.1200, 0.0453, 0.0245, 0.0657], device='cuda:3'), in_proj_covar=tensor([0.0279, 0.0395, 0.0329, 0.0319, 0.0341, 0.0371, 0.0224, 0.0390], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:24:05,198 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 17:24:10,476 INFO [train.py:904] (3/8) Epoch 18, batch 5150, loss[loss=0.188, simple_loss=0.2739, pruned_loss=0.05112, over 17035.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2753, pruned_loss=0.0464, over 3216829.09 frames. ], batch size: 55, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:24:38,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7247, 4.7951, 5.1393, 5.1172, 5.0818, 4.7682, 4.7061, 4.5431], device='cuda:3'), covar=tensor([0.0280, 0.0407, 0.0289, 0.0285, 0.0427, 0.0340, 0.0914, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0379, 0.0413, 0.0407, 0.0380, 0.0453, 0.0427, 0.0522, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:24:46,063 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:04,668 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9321, 5.0187, 5.3572, 5.3383, 5.3336, 4.9771, 4.9336, 4.6900], device='cuda:3'), covar=tensor([0.0309, 0.0442, 0.0335, 0.0369, 0.0500, 0.0328, 0.0930, 0.0438], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0413, 0.0408, 0.0381, 0.0453, 0.0427, 0.0522, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:25:20,503 INFO [optim.py:368] (3/8) 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 17:25:24,682 INFO [train.py:904] (3/8) Epoch 18, batch 5200, loss[loss=0.1905, simple_loss=0.2757, pruned_loss=0.05264, over 16669.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2735, pruned_loss=0.04579, over 3209974.54 frames. ], batch size: 134, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:25:44,165 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:47,811 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:06,674 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:07,061 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 17:26:21,237 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 17:26:31,588 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:39,155 INFO [train.py:904] (3/8) Epoch 18, batch 5250, loss[loss=0.1816, simple_loss=0.2741, pruned_loss=0.04451, over 16400.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2708, pruned_loss=0.04539, over 3212205.08 frames. ], batch size: 146, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:26:55,360 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:57,994 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5759, 3.6383, 3.4540, 3.1706, 3.2388, 3.5488, 3.2926, 3.3964], device='cuda:3'), covar=tensor([0.0549, 0.0506, 0.0314, 0.0267, 0.0609, 0.0453, 0.1475, 0.0476], device='cuda:3'), in_proj_covar=tensor([0.0280, 0.0395, 0.0329, 0.0320, 0.0340, 0.0370, 0.0224, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:26:58,850 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:37,109 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7433, 4.9409, 5.0084, 4.8758, 4.9433, 5.4277, 4.9767, 4.6728], device='cuda:3'), covar=tensor([0.1060, 0.1875, 0.1576, 0.1937, 0.2180, 0.0883, 0.1370, 0.2473], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0561, 0.0614, 0.0473, 0.0632, 0.0643, 0.0485, 0.0633], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:27:37,234 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 1.963e+02 2.196e+02 2.536e+02 4.359e+02, threshold=4.392e+02, percent-clipped=1.0 2023-04-30 17:27:50,018 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6872, 2.6337, 1.8968, 2.7572, 2.1117, 2.8113, 2.1225, 2.3747], device='cuda:3'), covar=tensor([0.0283, 0.0343, 0.1293, 0.0193, 0.0677, 0.0498, 0.1210, 0.0625], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0151, 0.0173, 0.0211, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:27:52,846 INFO [train.py:904] (3/8) Epoch 18, batch 5300, loss[loss=0.1882, simple_loss=0.2768, pruned_loss=0.0498, over 16518.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.268, pruned_loss=0.04454, over 3205553.01 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:28:02,061 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177858.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:28:28,997 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7179, 2.6293, 2.6426, 4.3714, 3.3311, 4.1029, 1.5904, 2.9751], device='cuda:3'), covar=tensor([0.1319, 0.0773, 0.1142, 0.0137, 0.0265, 0.0348, 0.1579, 0.0804], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0171, 0.0192, 0.0181, 0.0205, 0.0214, 0.0197, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:28:36,919 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 17:29:05,780 INFO [train.py:904] (3/8) Epoch 18, batch 5350, loss[loss=0.1802, simple_loss=0.2741, pruned_loss=0.04318, over 15425.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2666, pruned_loss=0.04372, over 3226281.51 frames. ], batch size: 190, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:29:17,489 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:29:17,645 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4686, 5.4043, 5.3356, 4.9552, 5.0234, 5.3222, 5.3767, 5.1155], device='cuda:3'), covar=tensor([0.0609, 0.0537, 0.0274, 0.0279, 0.1056, 0.0526, 0.0220, 0.0524], device='cuda:3'), in_proj_covar=tensor([0.0284, 0.0400, 0.0333, 0.0324, 0.0344, 0.0375, 0.0227, 0.0394], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:30:15,239 INFO [optim.py:368] (3/8) 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 17:30:19,773 INFO [train.py:904] (3/8) Epoch 18, batch 5400, loss[loss=0.1735, simple_loss=0.2648, pruned_loss=0.04112, over 16533.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2691, pruned_loss=0.0443, over 3215710.94 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:30:36,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5706, 3.6225, 3.4093, 3.0754, 3.2355, 3.5028, 3.3490, 3.3630], device='cuda:3'), covar=tensor([0.0547, 0.0508, 0.0280, 0.0251, 0.0559, 0.0409, 0.1363, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0398, 0.0332, 0.0323, 0.0343, 0.0374, 0.0225, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:31:39,766 INFO [train.py:904] (3/8) Epoch 18, batch 5450, loss[loss=0.2466, simple_loss=0.3267, pruned_loss=0.08323, over 15418.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2728, pruned_loss=0.04641, over 3185482.20 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:32:08,121 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:32:46,785 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0395, 3.1925, 3.1843, 2.0537, 2.9804, 3.2206, 3.0179, 1.8488], device='cuda:3'), covar=tensor([0.0481, 0.0063, 0.0063, 0.0410, 0.0115, 0.0093, 0.0095, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0132, 0.0094, 0.0103, 0.0091, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:32:52,433 INFO [optim.py:368] (3/8) 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 17:32:56,917 INFO [train.py:904] (3/8) Epoch 18, batch 5500, loss[loss=0.22, simple_loss=0.3066, pruned_loss=0.06669, over 16866.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.28, pruned_loss=0.0509, over 3158218.23 frames. ], batch size: 90, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:33:23,718 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-30 17:34:14,385 INFO [train.py:904] (3/8) Epoch 18, batch 5550, loss[loss=0.2132, simple_loss=0.2941, pruned_loss=0.06619, over 16201.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2872, pruned_loss=0.05664, over 3111131.21 frames. ], batch size: 165, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:09,025 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:35:31,150 INFO [optim.py:368] (3/8) 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 17:35:34,160 INFO [train.py:904] (3/8) Epoch 18, batch 5600, loss[loss=0.2679, simple_loss=0.3238, pruned_loss=0.106, over 10976.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2928, pruned_loss=0.06148, over 3070757.83 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:36,604 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:36:18,297 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2289, 3.1603, 3.3994, 1.6454, 3.6114, 3.6242, 2.8789, 2.6497], device='cuda:3'), covar=tensor([0.0846, 0.0249, 0.0182, 0.1265, 0.0075, 0.0192, 0.0415, 0.0473], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0105, 0.0094, 0.0137, 0.0076, 0.0121, 0.0124, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 17:36:58,586 INFO [train.py:904] (3/8) Epoch 18, batch 5650, loss[loss=0.2054, simple_loss=0.2878, pruned_loss=0.06154, over 16593.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2964, pruned_loss=0.06454, over 3053737.93 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:37:09,709 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 18, batch 5700, loss[loss=0.2309, simple_loss=0.3103, pruned_loss=0.07569, over 15480.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2986, pruned_loss=0.06681, over 3024555.38 frames. ], batch size: 191, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:38:20,137 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8736, 2.6768, 2.5913, 1.9190, 2.4984, 2.6515, 2.5028, 1.8815], device='cuda:3'), covar=tensor([0.0401, 0.0079, 0.0096, 0.0359, 0.0142, 0.0114, 0.0117, 0.0385], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0130, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:38:25,290 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:39:39,090 INFO [train.py:904] (3/8) Epoch 18, batch 5750, loss[loss=0.2129, simple_loss=0.2948, pruned_loss=0.06551, over 17041.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3025, pruned_loss=0.0697, over 2983265.16 frames. ], batch size: 53, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:40:08,177 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:40:56,660 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-30 17:40:59,316 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 18, batch 5800, loss[loss=0.2168, simple_loss=0.2912, pruned_loss=0.07122, over 11677.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3023, pruned_loss=0.06849, over 2994895.47 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:41:28,155 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:42:19,739 INFO [train.py:904] (3/8) Epoch 18, batch 5850, loss[loss=0.2099, simple_loss=0.2917, pruned_loss=0.06403, over 16253.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2996, pruned_loss=0.0666, over 3015278.57 frames. ], batch size: 165, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:16,108 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 18, batch 5900, loss[loss=0.2021, simple_loss=0.2898, pruned_loss=0.0572, over 16699.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2982, pruned_loss=0.06499, over 3048167.12 frames. ], batch size: 57, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:42,224 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:44,123 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178484.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:51,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1397, 3.2315, 3.3685, 2.3217, 3.1388, 3.3780, 3.2185, 1.8201], device='cuda:3'), covar=tensor([0.0490, 0.0078, 0.0058, 0.0375, 0.0098, 0.0108, 0.0087, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0079, 0.0080, 0.0132, 0.0094, 0.0105, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:44:58,646 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:59,548 INFO [train.py:904] (3/8) Epoch 18, batch 5950, loss[loss=0.2205, simple_loss=0.3068, pruned_loss=0.06713, over 16678.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2984, pruned_loss=0.06368, over 3054801.97 frames. ], batch size: 124, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:45:18,613 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:45:21,265 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4415, 3.4638, 2.1766, 3.8837, 2.6961, 3.8587, 2.1604, 2.7170], device='cuda:3'), covar=tensor([0.0257, 0.0363, 0.1517, 0.0208, 0.0745, 0.0616, 0.1530, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0190, 0.0151, 0.0173, 0.0211, 0.0198, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:45:29,116 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9708, 5.4960, 5.6669, 5.3929, 5.4595, 6.0322, 5.4410, 5.2696], device='cuda:3'), covar=tensor([0.0928, 0.1764, 0.2249, 0.1885, 0.2311, 0.0813, 0.1534, 0.2314], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0561, 0.0616, 0.0471, 0.0631, 0.0642, 0.0488, 0.0636], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 17:46:17,845 INFO [optim.py:368] (3/8) 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 17:46:19,094 INFO [train.py:904] (3/8) Epoch 18, batch 6000, loss[loss=0.2119, simple_loss=0.2925, pruned_loss=0.0656, over 16660.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2974, pruned_loss=0.063, over 3086631.56 frames. ], batch size: 62, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:46:19,095 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 17:46:29,951 INFO [train.py:938] (3/8) Epoch 18, validation: loss=0.1531, simple_loss=0.2661, pruned_loss=0.02007, over 944034.00 frames. 2023-04-30 17:46:29,952 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 17:47:48,044 INFO [train.py:904] (3/8) Epoch 18, batch 6050, loss[loss=0.2119, simple_loss=0.2837, pruned_loss=0.07009, over 11761.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2959, pruned_loss=0.06249, over 3097538.85 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:48:14,273 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178619.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:48:55,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8501, 2.3945, 1.9017, 2.1859, 2.7937, 2.4174, 2.6967, 2.9165], device='cuda:3'), covar=tensor([0.0156, 0.0357, 0.0485, 0.0388, 0.0211, 0.0310, 0.0191, 0.0209], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0224, 0.0217, 0.0216, 0.0226, 0.0224, 0.0226, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 17:49:06,484 INFO [optim.py:368] (3/8) 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 17:49:06,499 INFO [train.py:904] (3/8) Epoch 18, batch 6100, loss[loss=0.1868, simple_loss=0.2672, pruned_loss=0.0532, over 16576.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2954, pruned_loss=0.06168, over 3095408.78 frames. ], batch size: 57, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:49:24,117 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6575, 2.4917, 2.4234, 3.6921, 2.7646, 3.8635, 1.5272, 2.7545], device='cuda:3'), covar=tensor([0.1419, 0.0829, 0.1261, 0.0181, 0.0263, 0.0431, 0.1714, 0.0890], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0180, 0.0206, 0.0215, 0.0196, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:49:51,403 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:12,343 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:16,778 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178697.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:50:18,842 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4920, 3.3839, 3.8732, 1.7920, 4.1087, 4.1047, 3.0514, 2.9689], device='cuda:3'), covar=tensor([0.0809, 0.0269, 0.0164, 0.1299, 0.0053, 0.0133, 0.0378, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0106, 0.0095, 0.0139, 0.0076, 0.0122, 0.0125, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:50:23,986 INFO [train.py:904] (3/8) Epoch 18, batch 6150, loss[loss=0.2166, simple_loss=0.3025, pruned_loss=0.06535, over 16241.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2938, pruned_loss=0.06104, over 3110809.68 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:34,601 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 17:51:39,656 INFO [optim.py:368] (3/8) 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 17:51:39,671 INFO [train.py:904] (3/8) Epoch 18, batch 6200, loss[loss=0.2198, simple_loss=0.2972, pruned_loss=0.07121, over 11655.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2917, pruned_loss=0.05979, over 3124586.47 frames. ], batch size: 246, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:45,817 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178755.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:51:49,671 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178758.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:52:08,987 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:52:10,503 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0540, 2.4379, 2.5505, 1.8386, 2.7245, 2.8226, 2.5056, 2.3817], device='cuda:3'), covar=tensor([0.0738, 0.0267, 0.0211, 0.1035, 0.0095, 0.0242, 0.0413, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0106, 0.0094, 0.0139, 0.0076, 0.0121, 0.0125, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 17:52:56,567 INFO [train.py:904] (3/8) Epoch 18, batch 6250, loss[loss=0.2094, simple_loss=0.3048, pruned_loss=0.05696, over 16351.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.291, pruned_loss=0.05912, over 3128713.29 frames. ], batch size: 146, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:53:07,990 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:53:41,388 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:54:15,682 INFO [train.py:904] (3/8) Epoch 18, batch 6300, loss[loss=0.2213, simple_loss=0.3046, pruned_loss=0.06906, over 16274.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.291, pruned_loss=0.05833, over 3135947.59 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:54:17,523 INFO [optim.py:368] (3/8) 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 17:54:45,822 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2933, 3.4215, 3.6032, 3.5735, 3.5734, 3.3849, 3.4087, 3.4646], device='cuda:3'), covar=tensor([0.0405, 0.0711, 0.0458, 0.0450, 0.0528, 0.0576, 0.0891, 0.0579], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0420, 0.0412, 0.0387, 0.0460, 0.0434, 0.0531, 0.0346], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 17:55:22,817 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178894.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:55:34,244 INFO [train.py:904] (3/8) Epoch 18, batch 6350, loss[loss=0.2283, simple_loss=0.31, pruned_loss=0.07323, over 15359.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2918, pruned_loss=0.05959, over 3118284.56 frames. ], batch size: 190, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:56:52,036 INFO [train.py:904] (3/8) Epoch 18, batch 6400, loss[loss=0.204, simple_loss=0.278, pruned_loss=0.06503, over 16165.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2919, pruned_loss=0.06062, over 3098052.14 frames. ], batch size: 35, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:56:53,837 INFO [optim.py:368] (3/8) 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 17:56:57,535 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178955.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 17:57:21,014 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-30 17:57:28,085 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 6450, loss[loss=0.1872, simple_loss=0.2834, pruned_loss=0.04548, over 16883.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2918, pruned_loss=0.05971, over 3104672.63 frames. ], batch size: 90, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:22,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9255, 3.9974, 2.4751, 4.7437, 3.1015, 4.6473, 2.6841, 3.2034], device='cuda:3'), covar=tensor([0.0244, 0.0341, 0.1642, 0.0171, 0.0765, 0.0442, 0.1450, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0191, 0.0152, 0.0174, 0.0212, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 17:59:24,078 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179050.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:59:26,156 INFO [train.py:904] (3/8) Epoch 18, batch 6500, loss[loss=0.2159, simple_loss=0.303, pruned_loss=0.06442, over 16924.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2906, pruned_loss=0.05951, over 3107809.23 frames. ], batch size: 109, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:27,319 INFO [optim.py:368] (3/8) 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 17:59:28,317 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179053.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:59:59,156 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 18:00:13,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0791, 3.3768, 3.5298, 1.8904, 3.0196, 2.3482, 3.4877, 3.6285], device='cuda:3'), covar=tensor([0.0243, 0.0748, 0.0549, 0.2052, 0.0787, 0.0926, 0.0611, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0149, 0.0141, 0.0127, 0.0141, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:00:44,327 INFO [train.py:904] (3/8) Epoch 18, batch 6550, loss[loss=0.1869, simple_loss=0.2967, pruned_loss=0.03856, over 16697.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2937, pruned_loss=0.06027, over 3116958.64 frames. ], batch size: 89, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:00:52,473 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 18:00:54,404 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179108.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:20,181 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179126.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:58,890 INFO [train.py:904] (3/8) Epoch 18, batch 6600, loss[loss=0.203, simple_loss=0.2933, pruned_loss=0.05628, over 16507.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2954, pruned_loss=0.06052, over 3120829.61 frames. ], batch size: 75, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:02:00,668 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.804e+02 3.237e+02 3.890e+02 6.888e+02, threshold=6.473e+02, percent-clipped=0.0 2023-04-30 18:02:05,101 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179156.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:02:22,932 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-04-30 18:03:17,504 INFO [train.py:904] (3/8) Epoch 18, batch 6650, loss[loss=0.2161, simple_loss=0.3021, pruned_loss=0.06502, over 15252.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2957, pruned_loss=0.06158, over 3111733.88 frames. ], batch size: 190, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:03:24,582 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 18:04:30,597 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179250.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:04:32,473 INFO [train.py:904] (3/8) Epoch 18, batch 6700, loss[loss=0.2362, simple_loss=0.3045, pruned_loss=0.08394, over 11589.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2945, pruned_loss=0.06234, over 3085188.28 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:34,183 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.690e+02 3.432e+02 4.163e+02 9.246e+02, threshold=6.864e+02, percent-clipped=3.0 2023-04-30 18:05:09,369 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:05:46,463 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5589, 1.8261, 2.2811, 2.5419, 2.6156, 2.8435, 1.8742, 2.7690], device='cuda:3'), covar=tensor([0.0193, 0.0438, 0.0266, 0.0291, 0.0246, 0.0179, 0.0497, 0.0123], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0187, 0.0173, 0.0177, 0.0186, 0.0145, 0.0190, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:05:48,861 INFO [train.py:904] (3/8) Epoch 18, batch 6750, loss[loss=0.237, simple_loss=0.3074, pruned_loss=0.0833, over 11876.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2935, pruned_loss=0.06272, over 3081303.11 frames. ], batch size: 247, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:06:19,477 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:01,041 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:03,104 INFO [train.py:904] (3/8) Epoch 18, batch 6800, loss[loss=0.2257, simple_loss=0.3062, pruned_loss=0.07254, over 15657.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.294, pruned_loss=0.06274, over 3094940.41 frames. ], batch size: 191, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:07:04,927 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179353.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:07:18,271 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:18,579 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 18:07:48,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4590, 2.9706, 3.0712, 1.9016, 2.7480, 2.1502, 3.0183, 3.1331], device='cuda:3'), covar=tensor([0.0279, 0.0729, 0.0554, 0.1959, 0.0795, 0.0953, 0.0682, 0.0863], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0160, 0.0165, 0.0150, 0.0142, 0.0128, 0.0142, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:08:15,640 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179398.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:20,458 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179401.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:08:21,234 INFO [train.py:904] (3/8) Epoch 18, batch 6850, loss[loss=0.2505, simple_loss=0.3188, pruned_loss=0.09112, over 11629.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.295, pruned_loss=0.06275, over 3103815.58 frames. ], batch size: 246, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:08:23,192 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 18:08:36,426 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 18:08:48,905 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1955, 5.1921, 5.6290, 5.5945, 5.5784, 5.2575, 5.1452, 4.9641], device='cuda:3'), covar=tensor([0.0318, 0.0565, 0.0422, 0.0405, 0.0434, 0.0448, 0.1038, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0389, 0.0425, 0.0414, 0.0391, 0.0464, 0.0436, 0.0534, 0.0349], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 18:08:50,176 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:51,667 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 18:08:56,183 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:09:24,178 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179444.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:09:34,746 INFO [train.py:904] (3/8) Epoch 18, batch 6900, loss[loss=0.2278, simple_loss=0.3106, pruned_loss=0.07249, over 15385.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2969, pruned_loss=0.06217, over 3099812.05 frames. ], batch size: 190, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:09:38,468 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:10:48,211 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2585, 3.4166, 3.5706, 3.5411, 3.5453, 3.3450, 3.4061, 3.4335], device='cuda:3'), covar=tensor([0.0426, 0.0726, 0.0476, 0.0434, 0.0575, 0.0595, 0.0861, 0.0588], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0429, 0.0417, 0.0394, 0.0468, 0.0440, 0.0539, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 18:10:53,461 INFO [train.py:904] (3/8) Epoch 18, batch 6950, loss[loss=0.1973, simple_loss=0.285, pruned_loss=0.05481, over 16550.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2985, pruned_loss=0.06377, over 3088526.82 frames. ], batch size: 75, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:10:58,836 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179505.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:11:34,360 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8502, 2.9082, 2.5241, 4.5401, 3.2623, 4.0252, 1.6266, 3.0026], device='cuda:3'), covar=tensor([0.1403, 0.0775, 0.1265, 0.0167, 0.0386, 0.0496, 0.1726, 0.0866], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0176, 0.0203, 0.0211, 0.0194, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:12:07,073 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179550.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:12:09,855 INFO [train.py:904] (3/8) Epoch 18, batch 7000, loss[loss=0.1986, simple_loss=0.2965, pruned_loss=0.05038, over 17068.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2992, pruned_loss=0.06375, over 3068489.76 frames. ], batch size: 53, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:12:12,175 INFO [optim.py:368] (3/8) 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 18:12:55,824 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 18:13:16,940 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179598.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:13:21,682 INFO [train.py:904] (3/8) Epoch 18, batch 7050, loss[loss=0.2065, simple_loss=0.2923, pruned_loss=0.06035, over 17065.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2991, pruned_loss=0.06316, over 3075346.44 frames. ], batch size: 55, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:13:35,899 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4151, 2.9253, 2.6862, 2.2787, 2.2814, 2.2755, 2.8606, 2.8749], device='cuda:3'), covar=tensor([0.2236, 0.0763, 0.1481, 0.2325, 0.2210, 0.2030, 0.0451, 0.1297], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0265, 0.0300, 0.0306, 0.0292, 0.0249, 0.0287, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:14:12,919 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 7100, loss[loss=0.2014, simple_loss=0.2905, pruned_loss=0.05618, over 16782.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2986, pruned_loss=0.06368, over 3048961.85 frames. ], batch size: 83, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:14:40,299 INFO [optim.py:368] (3/8) 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 18:15:33,583 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1678, 1.4552, 1.9622, 2.0804, 2.1772, 2.3670, 1.6554, 2.2857], device='cuda:3'), covar=tensor([0.0228, 0.0464, 0.0249, 0.0302, 0.0272, 0.0184, 0.0483, 0.0123], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0178, 0.0187, 0.0145, 0.0191, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:15:46,875 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:15:53,678 INFO [train.py:904] (3/8) Epoch 18, batch 7150, loss[loss=0.2079, simple_loss=0.2982, pruned_loss=0.05879, over 16820.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2965, pruned_loss=0.0629, over 3068322.92 frames. ], batch size: 116, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:15:55,624 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-04-30 18:16:09,049 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6199, 3.6588, 2.7060, 2.2162, 2.4262, 2.2573, 3.9638, 3.2679], device='cuda:3'), covar=tensor([0.2870, 0.0769, 0.1954, 0.2529, 0.2706, 0.2146, 0.0473, 0.1303], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0292, 0.0249, 0.0287, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:16:16,355 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179717.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:16:27,339 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 18:17:00,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5771, 3.5509, 2.7000, 2.1647, 2.2873, 2.2422, 3.7232, 3.1939], device='cuda:3'), covar=tensor([0.2941, 0.0751, 0.1945, 0.2908, 0.2941, 0.2223, 0.0526, 0.1388], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0291, 0.0249, 0.0286, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:17:08,213 INFO [train.py:904] (3/8) Epoch 18, batch 7200, loss[loss=0.1679, simple_loss=0.253, pruned_loss=0.04135, over 16603.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2936, pruned_loss=0.06117, over 3065205.34 frames. ], batch size: 57, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:17:10,635 INFO [optim.py:368] (3/8) 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 18:17:29,446 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 18:17:30,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6904, 1.7886, 1.6323, 1.4205, 1.8952, 1.5102, 1.6110, 1.8324], device='cuda:3'), covar=tensor([0.0150, 0.0275, 0.0376, 0.0339, 0.0195, 0.0268, 0.0171, 0.0187], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:17:54,652 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-30 18:18:17,290 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179800.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:18:27,049 INFO [train.py:904] (3/8) Epoch 18, batch 7250, loss[loss=0.1905, simple_loss=0.2778, pruned_loss=0.0516, over 16409.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2911, pruned_loss=0.05991, over 3070142.12 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:18:29,302 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1091, 4.2061, 4.5330, 4.4682, 4.4953, 4.2184, 4.2534, 4.1114], device='cuda:3'), covar=tensor([0.0321, 0.0500, 0.0382, 0.0443, 0.0440, 0.0380, 0.0784, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0426, 0.0416, 0.0393, 0.0467, 0.0438, 0.0534, 0.0351], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 18:19:41,445 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3779, 1.5599, 2.1125, 2.2629, 2.3250, 2.6793, 1.7529, 2.5880], device='cuda:3'), covar=tensor([0.0200, 0.0479, 0.0257, 0.0323, 0.0280, 0.0163, 0.0477, 0.0122], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0186, 0.0173, 0.0176, 0.0185, 0.0143, 0.0189, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:19:42,142 INFO [train.py:904] (3/8) Epoch 18, batch 7300, loss[loss=0.2137, simple_loss=0.3026, pruned_loss=0.06238, over 16351.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2902, pruned_loss=0.05947, over 3077388.85 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:19:45,256 INFO [optim.py:368] (3/8) 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 18:19:50,389 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:20:33,698 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4204, 4.6326, 4.7798, 4.6384, 4.7041, 5.1617, 4.6996, 4.4379], device='cuda:3'), covar=tensor([0.1437, 0.1786, 0.2013, 0.1831, 0.2163, 0.0972, 0.1603, 0.2518], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0559, 0.0618, 0.0470, 0.0628, 0.0644, 0.0486, 0.0631], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:20:58,393 INFO [train.py:904] (3/8) Epoch 18, batch 7350, loss[loss=0.209, simple_loss=0.2957, pruned_loss=0.0611, over 16443.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2921, pruned_loss=0.06135, over 3037582.61 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:21:05,532 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179906.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:21:32,218 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 7400, loss[loss=0.2043, simple_loss=0.2905, pruned_loss=0.0591, over 16222.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2927, pruned_loss=0.06127, over 3057033.94 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:22:19,965 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.864e+02 3.444e+02 4.186e+02 8.589e+02, threshold=6.889e+02, percent-clipped=1.0 2023-04-30 18:22:30,130 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2282, 2.0684, 1.7226, 1.7218, 2.3186, 1.9689, 2.0174, 2.3821], device='cuda:3'), covar=tensor([0.0169, 0.0317, 0.0478, 0.0446, 0.0211, 0.0345, 0.0180, 0.0230], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:22:41,117 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179967.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:23:09,676 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:21,175 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 7450, loss[loss=0.1958, simple_loss=0.2881, pruned_loss=0.05177, over 16908.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2941, pruned_loss=0.06264, over 3054061.45 frames. ], batch size: 109, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:24:05,903 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:10,827 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3025, 3.1694, 3.4129, 1.7273, 3.5698, 3.5796, 2.8067, 2.6607], device='cuda:3'), covar=tensor([0.0777, 0.0248, 0.0216, 0.1256, 0.0093, 0.0257, 0.0423, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0105, 0.0094, 0.0138, 0.0075, 0.0120, 0.0124, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 18:24:19,814 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:01,488 INFO [train.py:904] (3/8) Epoch 18, batch 7500, loss[loss=0.2035, simple_loss=0.2947, pruned_loss=0.05608, over 16426.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2946, pruned_loss=0.06215, over 3065574.41 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:25:04,520 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:26,263 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1303, 3.1623, 1.9188, 3.3973, 2.4007, 3.4602, 2.1244, 2.6501], device='cuda:3'), covar=tensor([0.0301, 0.0427, 0.1658, 0.0209, 0.0856, 0.0619, 0.1451, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0172, 0.0190, 0.0150, 0.0175, 0.0212, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:25:56,694 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180087.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:26:16,422 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180100.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:26:19,022 INFO [train.py:904] (3/8) Epoch 18, batch 7550, loss[loss=0.1955, simple_loss=0.2825, pruned_loss=0.05424, over 16673.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2935, pruned_loss=0.06228, over 3061920.94 frames. ], batch size: 76, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:26:47,577 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-30 18:27:18,248 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2329, 4.3706, 4.4785, 4.2819, 4.3837, 4.8453, 4.3310, 4.0672], device='cuda:3'), covar=tensor([0.1558, 0.1863, 0.2264, 0.2069, 0.2262, 0.1014, 0.1760, 0.2692], device='cuda:3'), in_proj_covar=tensor([0.0387, 0.0559, 0.0619, 0.0469, 0.0627, 0.0644, 0.0488, 0.0631], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:27:30,330 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180148.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:27:36,553 INFO [train.py:904] (3/8) Epoch 18, batch 7600, loss[loss=0.2034, simple_loss=0.2867, pruned_loss=0.06006, over 16534.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2923, pruned_loss=0.06201, over 3069767.02 frames. ], batch size: 75, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:27:36,885 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180152.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:27:39,406 INFO [optim.py:368] (3/8) 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 18:28:55,650 INFO [train.py:904] (3/8) Epoch 18, batch 7650, loss[loss=0.2173, simple_loss=0.3036, pruned_loss=0.06544, over 15305.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2935, pruned_loss=0.06263, over 3080087.72 frames. ], batch size: 191, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:29:00,859 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:29:05,860 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1044, 2.3574, 2.3610, 2.7616, 1.9118, 3.1390, 1.8128, 2.7300], device='cuda:3'), covar=tensor([0.1160, 0.0639, 0.1013, 0.0206, 0.0135, 0.0367, 0.1514, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0178, 0.0206, 0.0214, 0.0196, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:30:13,289 INFO [train.py:904] (3/8) Epoch 18, batch 7700, loss[loss=0.2561, simple_loss=0.3182, pruned_loss=0.09701, over 11489.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2931, pruned_loss=0.06275, over 3079068.83 frames. ], batch size: 247, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:30:18,205 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 3.108e+02 3.616e+02 4.495e+02 6.527e+02, threshold=7.232e+02, percent-clipped=0.0 2023-04-30 18:30:20,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7605, 3.7333, 4.0903, 1.9840, 4.3268, 4.3224, 3.2085, 3.1381], device='cuda:3'), covar=tensor([0.0756, 0.0230, 0.0187, 0.1319, 0.0061, 0.0153, 0.0394, 0.0458], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0105, 0.0094, 0.0139, 0.0075, 0.0121, 0.0125, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 18:30:29,223 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180262.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:30:35,646 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:56,908 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:15,924 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 7750, loss[loss=0.2152, simple_loss=0.2871, pruned_loss=0.07161, over 11191.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2926, pruned_loss=0.06206, over 3086739.69 frames. ], batch size: 250, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:23,766 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180336.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:29,343 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 7800, loss[loss=0.2607, simple_loss=0.3232, pruned_loss=0.09914, over 11351.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2943, pruned_loss=0.06332, over 3071897.07 frames. ], batch size: 246, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:51,024 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.996e+02 3.455e+02 4.260e+02 9.369e+02, threshold=6.911e+02, percent-clipped=1.0 2023-04-30 18:32:55,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2772, 4.3042, 4.1973, 3.4446, 4.2276, 1.7326, 4.0131, 3.9679], device='cuda:3'), covar=tensor([0.0132, 0.0118, 0.0204, 0.0381, 0.0125, 0.2784, 0.0177, 0.0242], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0138, 0.0184, 0.0168, 0.0159, 0.0195, 0.0173, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:33:33,618 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:55,563 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180397.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:56,869 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0942, 2.4975, 2.5985, 1.8969, 2.7166, 2.7885, 2.4700, 2.4292], device='cuda:3'), covar=tensor([0.0685, 0.0243, 0.0219, 0.0970, 0.0109, 0.0260, 0.0430, 0.0428], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0105, 0.0094, 0.0138, 0.0075, 0.0120, 0.0124, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 18:34:02,022 INFO [train.py:904] (3/8) Epoch 18, batch 7850, loss[loss=0.2197, simple_loss=0.3078, pruned_loss=0.06586, over 16654.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2943, pruned_loss=0.06238, over 3092074.40 frames. ], batch size: 62, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:34:09,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3135, 5.6302, 5.3391, 5.3786, 5.0566, 5.0055, 5.0202, 5.7031], device='cuda:3'), covar=tensor([0.1090, 0.0762, 0.0994, 0.0810, 0.0819, 0.0729, 0.1158, 0.0907], device='cuda:3'), in_proj_covar=tensor([0.0618, 0.0760, 0.0624, 0.0569, 0.0475, 0.0490, 0.0636, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:34:59,660 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9470, 4.2035, 4.0074, 4.0598, 3.7618, 3.8108, 3.8169, 4.1808], device='cuda:3'), covar=tensor([0.1073, 0.0859, 0.0958, 0.0807, 0.0753, 0.1592, 0.0955, 0.1036], device='cuda:3'), in_proj_covar=tensor([0.0619, 0.0761, 0.0626, 0.0570, 0.0475, 0.0490, 0.0637, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:35:16,217 INFO [train.py:904] (3/8) Epoch 18, batch 7900, loss[loss=0.1996, simple_loss=0.2948, pruned_loss=0.05218, over 16772.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2933, pruned_loss=0.06181, over 3081470.57 frames. ], batch size: 102, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:35:17,224 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:35:20,361 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.825e+02 3.445e+02 4.253e+02 7.547e+02, threshold=6.890e+02, percent-clipped=0.0 2023-04-30 18:36:25,746 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 18:36:31,922 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 7950, loss[loss=0.2748, simple_loss=0.3299, pruned_loss=0.1098, over 11875.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2939, pruned_loss=0.06235, over 3075812.41 frames. ], batch size: 248, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:37:35,933 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 18:37:52,807 INFO [train.py:904] (3/8) Epoch 18, batch 8000, loss[loss=0.2653, simple_loss=0.3242, pruned_loss=0.1033, over 11124.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2948, pruned_loss=0.06275, over 3083466.18 frames. ], batch size: 246, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:37:57,084 INFO [optim.py:368] (3/8) 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 18:38:07,594 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:08,933 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180562.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:38:21,354 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:36,571 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:10,322 INFO [train.py:904] (3/8) Epoch 18, batch 8050, loss[loss=0.2067, simple_loss=0.2935, pruned_loss=0.05997, over 15400.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2941, pruned_loss=0.06217, over 3089451.55 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:39:11,806 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 18:39:23,082 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180610.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:39:50,350 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:54,565 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:40:08,472 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7603, 1.7882, 1.6005, 1.5333, 1.9068, 1.6392, 1.6267, 1.9291], device='cuda:3'), covar=tensor([0.0160, 0.0233, 0.0342, 0.0288, 0.0184, 0.0216, 0.0150, 0.0178], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:40:26,588 INFO [train.py:904] (3/8) Epoch 18, batch 8100, loss[loss=0.2193, simple_loss=0.2866, pruned_loss=0.07604, over 11810.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2934, pruned_loss=0.06113, over 3109827.33 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:40:32,030 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 2.839e+02 3.233e+02 3.954e+02 8.532e+02, threshold=6.466e+02, percent-clipped=3.0 2023-04-30 18:40:41,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5214, 3.5127, 3.4814, 2.6656, 3.3868, 2.0946, 3.1247, 2.8296], device='cuda:3'), covar=tensor([0.0150, 0.0116, 0.0175, 0.0221, 0.0108, 0.2152, 0.0134, 0.0228], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0139, 0.0185, 0.0170, 0.0160, 0.0196, 0.0174, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:41:00,845 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5642, 3.4278, 2.7260, 2.2198, 2.3103, 2.2768, 3.6180, 3.1860], device='cuda:3'), covar=tensor([0.2919, 0.0777, 0.1846, 0.2754, 0.2714, 0.2153, 0.0513, 0.1324], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0265, 0.0301, 0.0306, 0.0294, 0.0250, 0.0288, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:41:12,000 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:26,889 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:41,226 INFO [train.py:904] (3/8) Epoch 18, batch 8150, loss[loss=0.1549, simple_loss=0.2457, pruned_loss=0.03203, over 16667.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2913, pruned_loss=0.06045, over 3120273.78 frames. ], batch size: 89, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:42:03,609 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-30 18:42:24,164 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180730.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:25,766 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-30 18:42:26,698 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 8200, loss[loss=0.169, simple_loss=0.2555, pruned_loss=0.04125, over 16447.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2885, pruned_loss=0.06025, over 3099745.38 frames. ], batch size: 35, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:43:02,071 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.834e+02 3.334e+02 4.232e+02 1.685e+03, threshold=6.669e+02, percent-clipped=1.0 2023-04-30 18:43:12,594 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7977, 1.3246, 1.7014, 1.6462, 1.8010, 1.9243, 1.5984, 1.7714], device='cuda:3'), covar=tensor([0.0226, 0.0358, 0.0210, 0.0275, 0.0244, 0.0159, 0.0367, 0.0119], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0175, 0.0184, 0.0143, 0.0188, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:43:51,531 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0249, 4.0987, 3.9201, 3.6904, 3.6234, 4.0168, 3.6557, 3.7919], device='cuda:3'), covar=tensor([0.0584, 0.0526, 0.0303, 0.0286, 0.0774, 0.0424, 0.0924, 0.0576], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0392, 0.0320, 0.0310, 0.0332, 0.0362, 0.0220, 0.0382], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:44:01,826 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180793.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:44:03,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3154, 4.3915, 4.2097, 3.9160, 3.8627, 4.2944, 4.0091, 4.0255], device='cuda:3'), covar=tensor([0.0562, 0.0473, 0.0301, 0.0316, 0.0902, 0.0444, 0.0606, 0.0650], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0392, 0.0320, 0.0310, 0.0332, 0.0362, 0.0220, 0.0382], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:44:09,456 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-30 18:44:14,893 INFO [train.py:904] (3/8) Epoch 18, batch 8250, loss[loss=0.1879, simple_loss=0.2693, pruned_loss=0.05326, over 12133.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2877, pruned_loss=0.0584, over 3066261.60 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:44:44,053 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4292, 2.8637, 3.2357, 1.9554, 2.8559, 2.1489, 3.1497, 3.0828], device='cuda:3'), covar=tensor([0.0279, 0.0825, 0.0472, 0.2023, 0.0743, 0.0995, 0.0605, 0.0842], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0158, 0.0163, 0.0149, 0.0141, 0.0127, 0.0140, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:44:53,343 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:45:20,849 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 18:45:37,343 INFO [train.py:904] (3/8) Epoch 18, batch 8300, loss[loss=0.1617, simple_loss=0.258, pruned_loss=0.03271, over 15307.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2848, pruned_loss=0.05534, over 3066742.70 frames. ], batch size: 191, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:45:43,843 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.270e+02 2.749e+02 3.299e+02 7.574e+02, threshold=5.499e+02, percent-clipped=1.0 2023-04-30 18:45:44,474 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8460, 5.1898, 5.3111, 5.1119, 5.1368, 5.7249, 5.2207, 4.9460], device='cuda:3'), covar=tensor([0.0879, 0.1764, 0.2125, 0.1995, 0.2418, 0.0908, 0.1349, 0.2286], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0556, 0.0620, 0.0468, 0.0624, 0.0643, 0.0488, 0.0630], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:45:52,356 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180861.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:26,768 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2024, 3.2790, 3.3271, 2.3640, 3.1139, 3.3128, 3.1667, 1.9910], device='cuda:3'), covar=tensor([0.0473, 0.0066, 0.0056, 0.0347, 0.0095, 0.0093, 0.0088, 0.0461], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0079, 0.0079, 0.0132, 0.0093, 0.0104, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:46:32,571 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:44,242 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6438, 3.0659, 3.3255, 2.0792, 2.8490, 2.1529, 3.3587, 3.3061], device='cuda:3'), covar=tensor([0.0278, 0.0812, 0.0485, 0.1943, 0.0806, 0.1002, 0.0579, 0.0803], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0156, 0.0161, 0.0147, 0.0140, 0.0126, 0.0139, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 18:46:54,465 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6986, 4.6207, 5.0752, 5.0441, 5.0463, 4.7984, 4.6753, 4.6414], device='cuda:3'), covar=tensor([0.0320, 0.0739, 0.0430, 0.0384, 0.0436, 0.0392, 0.1228, 0.0489], device='cuda:3'), in_proj_covar=tensor([0.0383, 0.0420, 0.0408, 0.0385, 0.0454, 0.0430, 0.0524, 0.0343], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 18:46:58,357 INFO [train.py:904] (3/8) Epoch 18, batch 8350, loss[loss=0.1627, simple_loss=0.2602, pruned_loss=0.03259, over 16478.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2838, pruned_loss=0.05303, over 3069510.28 frames. ], batch size: 68, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:47:09,894 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:47:35,927 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 8400, loss[loss=0.1768, simple_loss=0.2691, pruned_loss=0.04227, over 16492.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2811, pruned_loss=0.05083, over 3053544.70 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:48:22,151 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.220e+02 2.634e+02 3.245e+02 6.969e+02, threshold=5.268e+02, percent-clipped=3.0 2023-04-30 18:49:13,721 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 18:49:16,575 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:49:31,280 INFO [train.py:904] (3/8) Epoch 18, batch 8450, loss[loss=0.1661, simple_loss=0.2617, pruned_loss=0.03527, over 16260.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2791, pruned_loss=0.04896, over 3057029.05 frames. ], batch size: 165, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:50:10,756 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-04-30 18:50:31,761 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181040.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:50:50,496 INFO [train.py:904] (3/8) Epoch 18, batch 8500, loss[loss=0.1701, simple_loss=0.2494, pruned_loss=0.04543, over 11737.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2761, pruned_loss=0.04697, over 3063039.72 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:50:58,702 INFO [optim.py:368] (3/8) 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 18:51:26,461 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5470, 3.7982, 3.8277, 2.7577, 3.4139, 3.7670, 3.5133, 2.1746], device='cuda:3'), covar=tensor([0.0431, 0.0049, 0.0040, 0.0306, 0.0099, 0.0089, 0.0075, 0.0459], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0077, 0.0077, 0.0129, 0.0091, 0.0102, 0.0089, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:51:48,654 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 8550, loss[loss=0.1912, simple_loss=0.2864, pruned_loss=0.04804, over 16758.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2737, pruned_loss=0.04582, over 3052238.91 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:53:50,704 INFO [train.py:904] (3/8) Epoch 18, batch 8600, loss[loss=0.1667, simple_loss=0.2542, pruned_loss=0.03957, over 12185.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.274, pruned_loss=0.04537, over 3040243.32 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:54:01,233 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.216e+02 2.624e+02 3.286e+02 8.410e+02, threshold=5.248e+02, percent-clipped=1.0 2023-04-30 18:54:49,988 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181181.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:55:07,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7926, 4.1603, 3.1082, 2.3891, 2.6367, 2.5416, 4.4003, 3.5755], device='cuda:3'), covar=tensor([0.2852, 0.0574, 0.1750, 0.2935, 0.2775, 0.2060, 0.0389, 0.1229], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0258, 0.0293, 0.0298, 0.0285, 0.0244, 0.0281, 0.0319], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:55:24,922 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4042, 3.0603, 2.6949, 2.2603, 2.1372, 2.2384, 3.0293, 2.8605], device='cuda:3'), covar=tensor([0.2597, 0.0686, 0.1613, 0.2635, 0.2600, 0.2129, 0.0477, 0.1350], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0258, 0.0293, 0.0298, 0.0285, 0.0243, 0.0280, 0.0319], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 18:55:29,852 INFO [train.py:904] (3/8) Epoch 18, batch 8650, loss[loss=0.1675, simple_loss=0.2579, pruned_loss=0.03854, over 12441.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2719, pruned_loss=0.04416, over 3014029.41 frames. ], batch size: 247, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:55:59,190 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3271, 4.3396, 4.2266, 3.5417, 4.2321, 1.6436, 4.0215, 4.0140], device='cuda:3'), covar=tensor([0.0106, 0.0097, 0.0162, 0.0289, 0.0102, 0.2859, 0.0142, 0.0226], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0138, 0.0182, 0.0166, 0.0158, 0.0195, 0.0172, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:56:16,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6381, 4.6659, 4.4922, 4.1660, 4.1881, 4.5815, 4.3586, 4.2877], device='cuda:3'), covar=tensor([0.0572, 0.0616, 0.0319, 0.0303, 0.0936, 0.0582, 0.0409, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0273, 0.0388, 0.0318, 0.0307, 0.0327, 0.0360, 0.0220, 0.0379], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 18:56:25,651 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181226.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:57:15,663 INFO [train.py:904] (3/8) Epoch 18, batch 8700, loss[loss=0.1653, simple_loss=0.2606, pruned_loss=0.035, over 16866.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2694, pruned_loss=0.04273, over 3033379.91 frames. ], batch size: 102, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:57:25,074 INFO [optim.py:368] (3/8) 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 18:57:56,765 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:39,661 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181297.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:48,988 INFO [train.py:904] (3/8) Epoch 18, batch 8750, loss[loss=0.1834, simple_loss=0.2776, pruned_loss=0.04458, over 16537.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.269, pruned_loss=0.04204, over 3049406.19 frames. ], batch size: 62, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:59:12,701 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 19:00:06,965 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 19:00:41,584 INFO [train.py:904] (3/8) Epoch 18, batch 8800, loss[loss=0.1729, simple_loss=0.2691, pruned_loss=0.0384, over 16660.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2674, pruned_loss=0.04077, over 3063462.38 frames. ], batch size: 89, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:00:51,212 INFO [optim.py:368] (3/8) 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 19:00:54,104 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181358.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:01:57,961 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:02:18,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5048, 1.9419, 1.6112, 1.5936, 2.2437, 1.8929, 1.9116, 2.3224], device='cuda:3'), covar=tensor([0.0153, 0.0416, 0.0536, 0.0480, 0.0300, 0.0383, 0.0190, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0219, 0.0213, 0.0213, 0.0221, 0.0218, 0.0220, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:02:27,191 INFO [train.py:904] (3/8) Epoch 18, batch 8850, loss[loss=0.1914, simple_loss=0.2947, pruned_loss=0.04401, over 16370.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2702, pruned_loss=0.04032, over 3046844.89 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:03:44,879 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:04:15,168 INFO [train.py:904] (3/8) Epoch 18, batch 8900, loss[loss=0.1746, simple_loss=0.2718, pruned_loss=0.03869, over 16799.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2703, pruned_loss=0.03958, over 3058453.85 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:04:25,752 INFO [optim.py:368] (3/8) 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 19:05:23,203 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:06:10,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6770, 2.5328, 1.8596, 2.1301, 2.9983, 2.6721, 3.3195, 3.3482], device='cuda:3'), covar=tensor([0.0131, 0.0549, 0.0767, 0.0618, 0.0334, 0.0472, 0.0254, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0221, 0.0215, 0.0215, 0.0222, 0.0220, 0.0222, 0.0214], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:06:18,289 INFO [train.py:904] (3/8) Epoch 18, batch 8950, loss[loss=0.17, simple_loss=0.259, pruned_loss=0.04045, over 12690.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2697, pruned_loss=0.03948, over 3078619.23 frames. ], batch size: 250, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:07:11,512 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5868, 3.6459, 3.4313, 3.0986, 3.2838, 3.5724, 3.3732, 3.3858], device='cuda:3'), covar=tensor([0.0544, 0.0471, 0.0261, 0.0225, 0.0461, 0.0391, 0.1048, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0382, 0.0314, 0.0302, 0.0321, 0.0354, 0.0218, 0.0374], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:07:17,051 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:19,769 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181530.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:08:08,280 INFO [train.py:904] (3/8) Epoch 18, batch 9000, loss[loss=0.1689, simple_loss=0.2533, pruned_loss=0.0423, over 12067.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2664, pruned_loss=0.03853, over 3054919.24 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:08:08,281 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 19:08:17,833 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 19:08:18,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6576, 2.3083, 2.2737, 4.2789, 2.3276, 2.6774, 2.3967, 2.4760], device='cuda:3'), covar=tensor([0.0995, 0.3658, 0.2954, 0.0410, 0.4072, 0.2608, 0.3495, 0.3721], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0424, 0.0352, 0.0314, 0.0425, 0.0485, 0.0394, 0.0493], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:08:27,942 INFO [optim.py:368] (3/8) 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 19:09:39,050 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 9050, loss[loss=0.1634, simple_loss=0.2533, pruned_loss=0.03671, over 12556.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2673, pruned_loss=0.03901, over 3064975.33 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:46,766 INFO [train.py:904] (3/8) Epoch 18, batch 9100, loss[loss=0.186, simple_loss=0.2921, pruned_loss=0.03993, over 15485.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2679, pruned_loss=0.03971, over 3072427.87 frames. ], batch size: 191, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:50,071 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.401e+02 2.871e+02 3.524e+02 6.480e+02, threshold=5.743e+02, percent-clipped=6.0 2023-04-30 19:13:34,509 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3756, 2.1739, 2.1813, 4.0434, 2.1566, 2.5211, 2.2618, 2.3346], device='cuda:3'), covar=tensor([0.1124, 0.3739, 0.3138, 0.0440, 0.4387, 0.2683, 0.3696, 0.3609], device='cuda:3'), in_proj_covar=tensor([0.0378, 0.0420, 0.0348, 0.0311, 0.0421, 0.0481, 0.0390, 0.0488], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:13:43,589 INFO [train.py:904] (3/8) Epoch 18, batch 9150, loss[loss=0.1636, simple_loss=0.265, pruned_loss=0.03105, over 17231.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2683, pruned_loss=0.03926, over 3079224.94 frames. ], batch size: 44, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:14:10,432 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5490, 3.5504, 3.5601, 2.8023, 3.4443, 1.9569, 3.2503, 2.9255], device='cuda:3'), covar=tensor([0.0170, 0.0152, 0.0181, 0.0213, 0.0137, 0.2454, 0.0153, 0.0245], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0135, 0.0178, 0.0162, 0.0156, 0.0192, 0.0169, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:15:27,777 INFO [train.py:904] (3/8) Epoch 18, batch 9200, loss[loss=0.1642, simple_loss=0.2477, pruned_loss=0.04035, over 12092.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2636, pruned_loss=0.03837, over 3062235.74 frames. ], batch size: 246, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:36,945 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.118e+02 2.449e+02 2.968e+02 5.097e+02, threshold=4.898e+02, percent-clipped=0.0 2023-04-30 19:15:57,060 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-30 19:16:44,306 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9904, 3.8721, 4.0698, 4.1955, 4.2696, 3.8446, 4.2247, 4.2927], device='cuda:3'), covar=tensor([0.1587, 0.1081, 0.1269, 0.0613, 0.0527, 0.1515, 0.0677, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0566, 0.0697, 0.0816, 0.0715, 0.0534, 0.0565, 0.0582, 0.0670], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:17:05,940 INFO [train.py:904] (3/8) Epoch 18, batch 9250, loss[loss=0.1689, simple_loss=0.2614, pruned_loss=0.0382, over 15454.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2631, pruned_loss=0.03827, over 3055989.26 frames. ], batch size: 193, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:18:57,425 INFO [train.py:904] (3/8) Epoch 18, batch 9300, loss[loss=0.1791, simple_loss=0.2662, pruned_loss=0.04602, over 16329.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2612, pruned_loss=0.0378, over 3038110.55 frames. ], batch size: 166, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:19:06,079 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:20:30,343 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9022, 2.2384, 2.2771, 2.9785, 1.7877, 3.2650, 1.7052, 2.6921], device='cuda:3'), covar=tensor([0.1413, 0.0783, 0.1207, 0.0182, 0.0104, 0.0379, 0.1678, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0166, 0.0188, 0.0173, 0.0197, 0.0209, 0.0193, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 19:20:41,425 INFO [train.py:904] (3/8) Epoch 18, batch 9350, loss[loss=0.1595, simple_loss=0.2472, pruned_loss=0.03593, over 12424.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2605, pruned_loss=0.03749, over 3059368.26 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:24,027 INFO [train.py:904] (3/8) Epoch 18, batch 9400, loss[loss=0.1502, simple_loss=0.2364, pruned_loss=0.032, over 12911.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2601, pruned_loss=0.03711, over 3059008.29 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:28,036 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:22:33,105 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.111e+02 2.455e+02 3.021e+02 7.291e+02, threshold=4.911e+02, percent-clipped=2.0 2023-04-30 19:23:16,407 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:24:05,162 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 9450, loss[loss=0.1662, simple_loss=0.2617, pruned_loss=0.03534, over 16681.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2624, pruned_loss=0.03753, over 3058308.93 frames. ], batch size: 76, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:19,112 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182039.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:25:41,959 INFO [train.py:904] (3/8) Epoch 18, batch 9500, loss[loss=0.1631, simple_loss=0.26, pruned_loss=0.03312, over 16790.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2615, pruned_loss=0.03701, over 3060292.52 frames. ], batch size: 83, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:55,087 INFO [optim.py:368] (3/8) 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 19:27:27,071 INFO [train.py:904] (3/8) Epoch 18, batch 9550, loss[loss=0.1985, simple_loss=0.2908, pruned_loss=0.05317, over 16831.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2612, pruned_loss=0.0372, over 3070880.10 frames. ], batch size: 116, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:06,816 INFO [train.py:904] (3/8) Epoch 18, batch 9600, loss[loss=0.1878, simple_loss=0.2665, pruned_loss=0.0546, over 12155.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2633, pruned_loss=0.03868, over 3044188.66 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:15,386 INFO [optim.py:368] (3/8) 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 19:30:14,222 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 18, batch 9650, loss[loss=0.1839, simple_loss=0.2684, pruned_loss=0.04968, over 12295.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2655, pruned_loss=0.03912, over 3031829.11 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:31:08,691 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-30 19:32:00,667 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182234.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:32:37,721 INFO [train.py:904] (3/8) Epoch 18, batch 9700, loss[loss=0.1864, simple_loss=0.2719, pruned_loss=0.05043, over 16793.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2642, pruned_loss=0.03868, over 3050988.90 frames. ], batch size: 124, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:45,794 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.275e+02 2.741e+02 3.246e+02 5.302e+02, threshold=5.483e+02, percent-clipped=0.0 2023-04-30 19:32:58,903 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-30 19:33:37,411 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:33:44,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0115, 2.4031, 2.0854, 2.0758, 2.6788, 2.4361, 2.6223, 2.8582], device='cuda:3'), covar=tensor([0.0130, 0.0368, 0.0463, 0.0466, 0.0270, 0.0346, 0.0215, 0.0252], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0221, 0.0215, 0.0215, 0.0222, 0.0220, 0.0218, 0.0212], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:33:58,328 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5348, 3.7633, 3.8630, 2.3493, 3.3399, 2.5007, 3.9941, 3.7558], device='cuda:3'), covar=tensor([0.0200, 0.0751, 0.0550, 0.1875, 0.0676, 0.0964, 0.0566, 0.0972], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0151, 0.0159, 0.0146, 0.0138, 0.0124, 0.0137, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 19:34:17,889 INFO [train.py:904] (3/8) Epoch 18, batch 9750, loss[loss=0.1777, simple_loss=0.2765, pruned_loss=0.03943, over 16403.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2625, pruned_loss=0.03855, over 3050091.72 frames. ], batch size: 166, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:35:19,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8123, 1.3401, 1.7618, 1.6851, 1.8598, 1.9164, 1.6439, 1.9048], device='cuda:3'), covar=tensor([0.0230, 0.0399, 0.0198, 0.0317, 0.0281, 0.0208, 0.0392, 0.0127], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0180, 0.0167, 0.0168, 0.0178, 0.0137, 0.0183, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:35:24,326 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:38,989 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:56,050 INFO [train.py:904] (3/8) Epoch 18, batch 9800, loss[loss=0.1812, simple_loss=0.2806, pruned_loss=0.04092, over 16813.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2626, pruned_loss=0.03793, over 3053421.34 frames. ], batch size: 124, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:36:05,429 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.961e+02 2.421e+02 2.847e+02 6.201e+02, threshold=4.843e+02, percent-clipped=1.0 2023-04-30 19:37:39,079 INFO [train.py:904] (3/8) Epoch 18, batch 9850, loss[loss=0.1536, simple_loss=0.2511, pruned_loss=0.02809, over 16372.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2635, pruned_loss=0.03718, over 3051986.37 frames. ], batch size: 68, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:38:34,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8350, 3.2129, 3.4799, 1.9692, 2.8696, 2.1375, 3.3251, 3.3235], device='cuda:3'), covar=tensor([0.0249, 0.0812, 0.0494, 0.2037, 0.0822, 0.1033, 0.0667, 0.0958], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0151, 0.0159, 0.0146, 0.0137, 0.0123, 0.0137, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 19:38:34,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5260, 3.7053, 2.8562, 2.1620, 2.3140, 2.2843, 3.9294, 3.3513], device='cuda:3'), covar=tensor([0.2901, 0.0590, 0.1675, 0.2937, 0.2897, 0.2095, 0.0320, 0.1163], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0252, 0.0287, 0.0291, 0.0274, 0.0239, 0.0275, 0.0310], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:39:04,982 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7528, 2.6076, 2.3064, 3.6207, 2.2504, 3.7808, 1.3822, 2.9643], device='cuda:3'), covar=tensor([0.1419, 0.0705, 0.1263, 0.0137, 0.0100, 0.0370, 0.1785, 0.0677], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0166, 0.0189, 0.0172, 0.0195, 0.0208, 0.0193, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 19:39:29,921 INFO [train.py:904] (3/8) Epoch 18, batch 9900, loss[loss=0.1661, simple_loss=0.27, pruned_loss=0.03109, over 16914.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2639, pruned_loss=0.03736, over 3046128.55 frames. ], batch size: 116, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:40,074 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6940, 2.6269, 2.5439, 4.4886, 2.8526, 4.1880, 1.4702, 2.9968], device='cuda:3'), covar=tensor([0.1493, 0.0876, 0.1220, 0.0187, 0.0142, 0.0354, 0.1776, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0167, 0.0189, 0.0173, 0.0195, 0.0209, 0.0193, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 19:39:40,680 INFO [optim.py:368] (3/8) 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 19:39:42,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2935, 3.0895, 3.3480, 1.7419, 3.5055, 3.5696, 2.8312, 2.7195], device='cuda:3'), covar=tensor([0.0709, 0.0272, 0.0199, 0.1200, 0.0073, 0.0150, 0.0424, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0101, 0.0089, 0.0134, 0.0072, 0.0114, 0.0120, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 19:41:28,674 INFO [train.py:904] (3/8) Epoch 18, batch 9950, loss[loss=0.1752, simple_loss=0.2676, pruned_loss=0.0414, over 12527.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2666, pruned_loss=0.03793, over 3058231.67 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:41:35,642 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3170, 5.3190, 5.1960, 4.0294, 5.2138, 1.8221, 4.8289, 4.9203], device='cuda:3'), covar=tensor([0.0168, 0.0119, 0.0240, 0.0635, 0.0125, 0.3174, 0.0184, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0134, 0.0175, 0.0158, 0.0154, 0.0190, 0.0166, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:41:58,887 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:43:31,283 INFO [train.py:904] (3/8) Epoch 18, batch 10000, loss[loss=0.1717, simple_loss=0.256, pruned_loss=0.04373, over 12691.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2652, pruned_loss=0.03724, over 3089557.93 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:43:42,213 INFO [optim.py:368] (3/8) 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 19:44:16,785 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:45:02,967 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-04-30 19:45:14,019 INFO [train.py:904] (3/8) Epoch 18, batch 10050, loss[loss=0.1625, simple_loss=0.2591, pruned_loss=0.03297, over 16707.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2651, pruned_loss=0.03684, over 3102080.87 frames. ], batch size: 76, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:15,010 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:20,804 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182637.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:24,003 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 19:46:46,717 INFO [train.py:904] (3/8) Epoch 18, batch 10100, loss[loss=0.1513, simple_loss=0.2458, pruned_loss=0.02841, over 16752.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2651, pruned_loss=0.03675, over 3112359.00 frames. ], batch size: 124, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:55,747 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.131e+02 2.554e+02 3.108e+02 6.507e+02, threshold=5.108e+02, percent-clipped=7.0 2023-04-30 19:47:44,811 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 0, loss[loss=0.2305, simple_loss=0.316, pruned_loss=0.07253, over 16633.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.316, pruned_loss=0.07253, over 16633.00 frames. ], batch size: 62, lr: 3.61e-03, grad_scale: 8.0 2023-04-30 19:48:32,142 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 19:48:39,774 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 19:49:37,143 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2886, 5.6717, 5.3919, 5.4480, 5.0510, 4.9954, 5.0246, 5.7945], device='cuda:3'), covar=tensor([0.1210, 0.0883, 0.1293, 0.0889, 0.0872, 0.0771, 0.1257, 0.0929], device='cuda:3'), in_proj_covar=tensor([0.0612, 0.0750, 0.0615, 0.0557, 0.0471, 0.0483, 0.0626, 0.0580], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:49:50,013 INFO [train.py:904] (3/8) Epoch 19, batch 50, loss[loss=0.2069, simple_loss=0.2836, pruned_loss=0.06506, over 16338.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2756, pruned_loss=0.05049, over 752416.08 frames. ], batch size: 146, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:49:59,030 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.571e+02 3.042e+02 3.790e+02 6.505e+02, threshold=6.085e+02, percent-clipped=6.0 2023-04-30 19:50:55,412 INFO [train.py:904] (3/8) Epoch 19, batch 100, loss[loss=0.2078, simple_loss=0.291, pruned_loss=0.06232, over 17019.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2697, pruned_loss=0.04892, over 1324865.52 frames. ], batch size: 55, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:03,139 INFO [train.py:904] (3/8) Epoch 19, batch 150, loss[loss=0.1795, simple_loss=0.264, pruned_loss=0.04754, over 16445.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2666, pruned_loss=0.04745, over 1769700.69 frames. ], batch size: 68, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:10,247 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182856.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:15,179 INFO [optim.py:368] (3/8) 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 19:52:18,033 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 19, batch 200, loss[loss=0.176, simple_loss=0.2539, pruned_loss=0.04906, over 16777.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2673, pruned_loss=0.04786, over 2111430.65 frames. ], batch size: 124, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:53:31,069 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2744, 4.0623, 4.2986, 4.4467, 4.5557, 4.1570, 4.3132, 4.5524], device='cuda:3'), covar=tensor([0.1678, 0.1130, 0.1401, 0.0794, 0.0655, 0.1224, 0.3257, 0.0719], device='cuda:3'), in_proj_covar=tensor([0.0583, 0.0719, 0.0839, 0.0737, 0.0549, 0.0576, 0.0599, 0.0688], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:53:34,351 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:42,613 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:00,397 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:21,114 INFO [train.py:904] (3/8) Epoch 19, batch 250, loss[loss=0.163, simple_loss=0.2466, pruned_loss=0.03973, over 16518.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.264, pruned_loss=0.04645, over 2387256.67 frames. ], batch size: 75, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:54:32,986 INFO [optim.py:368] (3/8) 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 19:54:55,238 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6396, 3.5516, 4.3306, 2.2400, 3.4031, 2.6004, 4.1263, 3.7971], device='cuda:3'), covar=tensor([0.0221, 0.0981, 0.0368, 0.2013, 0.0722, 0.0945, 0.0587, 0.1238], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0153, 0.0161, 0.0148, 0.0139, 0.0125, 0.0138, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 19:55:08,834 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:55:23,889 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 19:55:30,551 INFO [train.py:904] (3/8) Epoch 19, batch 300, loss[loss=0.1821, simple_loss=0.2629, pruned_loss=0.05068, over 16528.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2612, pruned_loss=0.04536, over 2600999.94 frames. ], batch size: 68, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:41,128 INFO [train.py:904] (3/8) Epoch 19, batch 350, loss[loss=0.1397, simple_loss=0.2269, pruned_loss=0.02628, over 17189.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2584, pruned_loss=0.04416, over 2760492.25 frames. ], batch size: 46, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:48,027 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6907, 5.0416, 4.7643, 4.7660, 4.5632, 4.5631, 4.5229, 5.1064], device='cuda:3'), covar=tensor([0.1163, 0.0890, 0.1028, 0.0805, 0.0828, 0.1121, 0.1134, 0.0901], device='cuda:3'), in_proj_covar=tensor([0.0632, 0.0776, 0.0637, 0.0577, 0.0487, 0.0497, 0.0646, 0.0600], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 19:56:52,338 INFO [optim.py:368] (3/8) 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 19:57:51,264 INFO [train.py:904] (3/8) Epoch 19, batch 400, loss[loss=0.1668, simple_loss=0.2501, pruned_loss=0.04171, over 16605.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2576, pruned_loss=0.0441, over 2883365.54 frames. ], batch size: 68, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:58:18,170 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1893, 5.8547, 5.9336, 5.6121, 5.7718, 6.3004, 5.8242, 5.5466], device='cuda:3'), covar=tensor([0.0877, 0.1871, 0.2107, 0.2356, 0.2298, 0.0924, 0.1546, 0.2367], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0560, 0.0623, 0.0472, 0.0628, 0.0653, 0.0491, 0.0631], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 19:59:03,157 INFO [train.py:904] (3/8) Epoch 19, batch 450, loss[loss=0.1477, simple_loss=0.2405, pruned_loss=0.02742, over 17137.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2563, pruned_loss=0.04378, over 2980453.85 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:59:14,112 INFO [optim.py:368] (3/8) 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 19:59:28,272 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:08,067 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 20:00:12,283 INFO [train.py:904] (3/8) Epoch 19, batch 500, loss[loss=0.1579, simple_loss=0.2408, pruned_loss=0.03749, over 15530.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2554, pruned_loss=0.04287, over 3062174.96 frames. ], batch size: 190, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:00:28,367 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,914 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,928 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:01:21,462 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-30 20:01:23,249 INFO [train.py:904] (3/8) Epoch 19, batch 550, loss[loss=0.1846, simple_loss=0.2719, pruned_loss=0.04871, over 16532.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2549, pruned_loss=0.04218, over 3119107.24 frames. ], batch size: 62, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:01:30,023 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9988, 3.0276, 3.1501, 2.1347, 2.9505, 3.2011, 3.0436, 1.8389], device='cuda:3'), covar=tensor([0.0539, 0.0118, 0.0063, 0.0410, 0.0125, 0.0117, 0.0105, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0079, 0.0079, 0.0131, 0.0094, 0.0103, 0.0090, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 20:01:34,915 INFO [optim.py:368] (3/8) 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 20:01:48,856 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7571, 2.5013, 1.9761, 2.3096, 2.8770, 2.6549, 2.8260, 2.8940], device='cuda:3'), covar=tensor([0.0249, 0.0380, 0.0529, 0.0452, 0.0236, 0.0325, 0.0234, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0232, 0.0223, 0.0224, 0.0232, 0.0231, 0.0233, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:02:32,735 INFO [train.py:904] (3/8) Epoch 19, batch 600, loss[loss=0.1409, simple_loss=0.2277, pruned_loss=0.02704, over 16825.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2547, pruned_loss=0.04221, over 3172997.23 frames. ], batch size: 42, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:19,864 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1250, 5.1040, 4.8342, 4.3207, 4.9047, 1.8628, 4.6690, 4.7117], device='cuda:3'), covar=tensor([0.0094, 0.0079, 0.0209, 0.0405, 0.0107, 0.2738, 0.0156, 0.0253], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0142, 0.0185, 0.0168, 0.0162, 0.0199, 0.0176, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:03:42,736 INFO [train.py:904] (3/8) Epoch 19, batch 650, loss[loss=0.1673, simple_loss=0.2435, pruned_loss=0.0455, over 16760.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2538, pruned_loss=0.04214, over 3206046.32 frames. ], batch size: 134, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:54,610 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.241e+02 2.597e+02 3.196e+02 6.464e+02, threshold=5.194e+02, percent-clipped=2.0 2023-04-30 20:04:10,589 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3826, 5.7431, 5.5073, 5.5497, 5.1545, 5.1836, 5.2352, 5.8639], device='cuda:3'), covar=tensor([0.1308, 0.0884, 0.1003, 0.0793, 0.0946, 0.0773, 0.1042, 0.0922], device='cuda:3'), in_proj_covar=tensor([0.0654, 0.0799, 0.0658, 0.0593, 0.0501, 0.0512, 0.0665, 0.0617], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:04:23,151 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 20:04:24,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8797, 3.0594, 2.9713, 5.0580, 4.1629, 4.4968, 1.7496, 3.2688], device='cuda:3'), covar=tensor([0.1318, 0.0738, 0.1045, 0.0217, 0.0227, 0.0397, 0.1567, 0.0743], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0180, 0.0199, 0.0213, 0.0195, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:04:53,075 INFO [train.py:904] (3/8) Epoch 19, batch 700, loss[loss=0.1685, simple_loss=0.2542, pruned_loss=0.04134, over 16871.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2535, pruned_loss=0.04179, over 3228384.08 frames. ], batch size: 96, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:05:00,998 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:05:59,948 INFO [train.py:904] (3/8) Epoch 19, batch 750, loss[loss=0.1605, simple_loss=0.2543, pruned_loss=0.03333, over 16673.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2536, pruned_loss=0.0419, over 3249869.56 frames. ], batch size: 57, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:06:11,898 INFO [optim.py:368] (3/8) 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 20:06:25,272 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183469.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:06:27,667 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4337, 3.4978, 3.9260, 2.0491, 3.1167, 2.4838, 3.9063, 3.7615], device='cuda:3'), covar=tensor([0.0243, 0.0853, 0.0498, 0.1970, 0.0799, 0.0918, 0.0559, 0.1072], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0156, 0.0163, 0.0150, 0.0141, 0.0126, 0.0141, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:07:10,060 INFO [train.py:904] (3/8) Epoch 19, batch 800, loss[loss=0.1745, simple_loss=0.2535, pruned_loss=0.04779, over 16891.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2528, pruned_loss=0.04173, over 3264047.27 frames. ], batch size: 96, lr: 3.61e-03, grad_scale: 4.0 2023-04-30 20:07:24,164 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:07:25,703 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 20:07:32,562 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:19,002 INFO [train.py:904] (3/8) Epoch 19, batch 850, loss[loss=0.1598, simple_loss=0.2388, pruned_loss=0.04045, over 16699.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2533, pruned_loss=0.04245, over 3284677.48 frames. ], batch size: 134, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:08:29,786 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.149e+02 2.547e+02 3.022e+02 7.911e+02, threshold=5.094e+02, percent-clipped=1.0 2023-04-30 20:08:30,865 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183560.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:34,708 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9162, 2.9374, 2.7422, 4.6019, 3.8258, 4.2464, 1.7286, 3.0371], device='cuda:3'), covar=tensor([0.1327, 0.0707, 0.1132, 0.0196, 0.0238, 0.0406, 0.1522, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0180, 0.0200, 0.0214, 0.0196, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:08:38,500 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:51,343 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 20:09:13,027 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 20:09:28,459 INFO [train.py:904] (3/8) Epoch 19, batch 900, loss[loss=0.1508, simple_loss=0.2311, pruned_loss=0.03521, over 15866.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2527, pruned_loss=0.04197, over 3277768.27 frames. ], batch size: 35, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:09:44,182 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 950, loss[loss=0.1867, simple_loss=0.2512, pruned_loss=0.0611, over 16873.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2526, pruned_loss=0.042, over 3279828.06 frames. ], batch size: 109, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:10:45,885 INFO [optim.py:368] (3/8) 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 20:11:08,179 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:11:42,836 INFO [train.py:904] (3/8) Epoch 19, batch 1000, loss[loss=0.1614, simple_loss=0.2341, pruned_loss=0.04439, over 16449.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2514, pruned_loss=0.04248, over 3276016.94 frames. ], batch size: 75, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:12:16,630 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6319, 4.6699, 5.0168, 5.0096, 5.0311, 4.6958, 4.6986, 4.5020], device='cuda:3'), covar=tensor([0.0375, 0.0808, 0.0454, 0.0456, 0.0466, 0.0515, 0.0874, 0.0627], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0437, 0.0425, 0.0401, 0.0470, 0.0448, 0.0538, 0.0357], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 20:12:50,599 INFO [train.py:904] (3/8) Epoch 19, batch 1050, loss[loss=0.1404, simple_loss=0.2348, pruned_loss=0.02301, over 17176.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2514, pruned_loss=0.04151, over 3289266.90 frames. ], batch size: 44, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:13:01,769 INFO [optim.py:368] (3/8) 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 20:13:06,550 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:14:00,540 INFO [train.py:904] (3/8) Epoch 19, batch 1100, loss[loss=0.1634, simple_loss=0.2422, pruned_loss=0.04236, over 16815.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2513, pruned_loss=0.04112, over 3303101.84 frames. ], batch size: 102, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:14:06,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9142, 4.3673, 4.4328, 3.1998, 3.6891, 4.4257, 3.8730, 2.4654], device='cuda:3'), covar=tensor([0.0440, 0.0062, 0.0039, 0.0326, 0.0124, 0.0073, 0.0075, 0.0446], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0080, 0.0079, 0.0131, 0.0095, 0.0103, 0.0091, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 20:14:09,354 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8744, 4.7991, 4.7880, 4.4603, 4.4451, 4.7980, 4.6462, 4.5244], device='cuda:3'), covar=tensor([0.0652, 0.0837, 0.0299, 0.0259, 0.0854, 0.0470, 0.0449, 0.0673], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0415, 0.0341, 0.0330, 0.0351, 0.0385, 0.0234, 0.0404], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:14:38,232 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-30 20:14:55,801 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 1150, loss[loss=0.1621, simple_loss=0.2467, pruned_loss=0.03875, over 15409.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2509, pruned_loss=0.04104, over 3285594.15 frames. ], batch size: 190, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:15:20,304 INFO [optim.py:368] (3/8) 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 20:16:18,544 INFO [train.py:904] (3/8) Epoch 19, batch 1200, loss[loss=0.156, simple_loss=0.2523, pruned_loss=0.02981, over 17098.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.25, pruned_loss=0.04058, over 3287304.51 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:16:20,084 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183903.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:16:48,249 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-30 20:16:54,883 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4424, 4.3217, 4.4661, 4.6398, 4.7268, 4.2928, 4.5290, 4.7190], device='cuda:3'), covar=tensor([0.1638, 0.1118, 0.1379, 0.0743, 0.0610, 0.1168, 0.2412, 0.0753], device='cuda:3'), in_proj_covar=tensor([0.0627, 0.0772, 0.0903, 0.0791, 0.0586, 0.0618, 0.0639, 0.0736], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:17:25,271 INFO [train.py:904] (3/8) Epoch 19, batch 1250, loss[loss=0.1922, simple_loss=0.2569, pruned_loss=0.06371, over 16875.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.25, pruned_loss=0.04138, over 3291477.39 frames. ], batch size: 116, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:17:35,899 INFO [optim.py:368] (3/8) 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 20:17:50,865 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183970.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:17:51,119 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2406, 3.3290, 3.5883, 2.2166, 3.0499, 2.4767, 3.6560, 3.6118], device='cuda:3'), covar=tensor([0.0230, 0.0962, 0.0548, 0.1870, 0.0851, 0.0980, 0.0576, 0.0967], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:18:02,997 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3462, 2.6258, 2.1702, 2.3755, 2.9797, 2.7059, 3.1353, 3.1294], device='cuda:3'), covar=tensor([0.0209, 0.0406, 0.0549, 0.0466, 0.0265, 0.0342, 0.0255, 0.0259], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0235, 0.0226, 0.0226, 0.0236, 0.0233, 0.0237, 0.0228], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:18:37,489 INFO [train.py:904] (3/8) Epoch 19, batch 1300, loss[loss=0.1694, simple_loss=0.2469, pruned_loss=0.04596, over 16534.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.25, pruned_loss=0.04129, over 3299293.56 frames. ], batch size: 75, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:44,262 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:19:46,321 INFO [train.py:904] (3/8) Epoch 19, batch 1350, loss[loss=0.1812, simple_loss=0.2705, pruned_loss=0.04598, over 16764.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2503, pruned_loss=0.04112, over 3287566.88 frames. ], batch size: 57, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:55,436 INFO [optim.py:368] (3/8) 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 20:20:03,920 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:20:55,153 INFO [train.py:904] (3/8) Epoch 19, batch 1400, loss[loss=0.1671, simple_loss=0.2608, pruned_loss=0.03669, over 16722.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2501, pruned_loss=0.0407, over 3295263.98 frames. ], batch size: 62, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:21:01,509 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 20:21:09,579 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:10,484 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184112.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:30,128 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2528, 5.2461, 4.9359, 4.4350, 5.0471, 1.8669, 4.7715, 4.9436], device='cuda:3'), covar=tensor([0.0087, 0.0083, 0.0219, 0.0414, 0.0110, 0.2994, 0.0167, 0.0208], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0148, 0.0193, 0.0176, 0.0170, 0.0206, 0.0185, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:22:05,088 INFO [train.py:904] (3/8) Epoch 19, batch 1450, loss[loss=0.1986, simple_loss=0.2667, pruned_loss=0.06527, over 16701.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2497, pruned_loss=0.04078, over 3306393.38 frames. ], batch size: 89, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:22:15,457 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.293e+02 2.595e+02 3.219e+02 5.738e+02, threshold=5.191e+02, percent-clipped=1.0 2023-04-30 20:22:30,596 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3770, 5.2937, 5.1750, 4.6770, 4.8104, 5.2401, 5.1724, 4.8439], device='cuda:3'), covar=tensor([0.0549, 0.0441, 0.0278, 0.0375, 0.1059, 0.0403, 0.0280, 0.0730], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0417, 0.0344, 0.0332, 0.0352, 0.0388, 0.0235, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:23:08,677 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184198.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:23:13,296 INFO [train.py:904] (3/8) Epoch 19, batch 1500, loss[loss=0.1754, simple_loss=0.2526, pruned_loss=0.04906, over 16589.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2495, pruned_loss=0.04155, over 3300033.23 frames. ], batch size: 68, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:22,396 INFO [train.py:904] (3/8) Epoch 19, batch 1550, loss[loss=0.1645, simple_loss=0.2481, pruned_loss=0.04042, over 16841.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2508, pruned_loss=0.04302, over 3287007.62 frames. ], batch size: 42, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:34,823 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.353e+02 2.741e+02 3.128e+02 4.649e+02, threshold=5.482e+02, percent-clipped=0.0 2023-04-30 20:24:47,420 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:25:28,674 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9818, 2.9015, 2.7267, 4.6454, 3.8386, 4.1912, 1.7510, 3.1626], device='cuda:3'), covar=tensor([0.1257, 0.0709, 0.1087, 0.0192, 0.0233, 0.0421, 0.1498, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0172, 0.0193, 0.0184, 0.0203, 0.0215, 0.0197, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:25:31,172 INFO [train.py:904] (3/8) Epoch 19, batch 1600, loss[loss=0.1722, simple_loss=0.2626, pruned_loss=0.04093, over 17235.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2521, pruned_loss=0.04286, over 3299542.90 frames. ], batch size: 45, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:25:53,346 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:39,948 INFO [train.py:904] (3/8) Epoch 19, batch 1650, loss[loss=0.1953, simple_loss=0.2733, pruned_loss=0.05867, over 16608.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.253, pruned_loss=0.04297, over 3304392.55 frames. ], batch size: 134, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:26:50,258 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:51,988 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 20:26:52,324 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.220e+02 2.674e+02 3.187e+02 5.456e+02, threshold=5.348e+02, percent-clipped=0.0 2023-04-30 20:27:49,666 INFO [train.py:904] (3/8) Epoch 19, batch 1700, loss[loss=0.1935, simple_loss=0.2914, pruned_loss=0.04781, over 16582.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2549, pruned_loss=0.04346, over 3309657.91 frames. ], batch size: 62, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:27:55,021 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:05,314 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 20:28:14,012 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:58,451 INFO [train.py:904] (3/8) Epoch 19, batch 1750, loss[loss=0.1929, simple_loss=0.2844, pruned_loss=0.05063, over 16728.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2555, pruned_loss=0.04325, over 3314708.58 frames. ], batch size: 62, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:29:10,989 INFO [optim.py:368] (3/8) 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 20:29:45,830 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184486.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:02,695 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 1800, loss[loss=0.165, simple_loss=0.2544, pruned_loss=0.03778, over 17180.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2574, pruned_loss=0.04371, over 3307134.09 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:31:05,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9590, 4.4605, 4.4743, 3.2222, 3.6864, 4.4162, 3.9275, 2.4468], device='cuda:3'), covar=tensor([0.0435, 0.0063, 0.0032, 0.0326, 0.0127, 0.0074, 0.0076, 0.0457], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0106, 0.0093, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 20:31:07,262 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184546.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:09,255 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 1850, loss[loss=0.1402, simple_loss=0.2257, pruned_loss=0.02737, over 16967.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2587, pruned_loss=0.04408, over 3308459.50 frames. ], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:31:29,851 INFO [optim.py:368] (3/8) 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 20:31:33,857 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-30 20:32:11,451 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7077, 3.7263, 2.3773, 4.3378, 2.8545, 4.2834, 2.6167, 3.0869], device='cuda:3'), covar=tensor([0.0295, 0.0437, 0.1588, 0.0305, 0.0838, 0.0490, 0.1378, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0161, 0.0175, 0.0216, 0.0202, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:32:19,140 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 20:32:26,186 INFO [train.py:904] (3/8) Epoch 19, batch 1900, loss[loss=0.1903, simple_loss=0.2678, pruned_loss=0.05643, over 16669.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2583, pruned_loss=0.04346, over 3304817.64 frames. ], batch size: 134, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:28,683 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3853, 3.7107, 4.0012, 2.1893, 3.2127, 2.5196, 3.8730, 3.8576], device='cuda:3'), covar=tensor([0.0262, 0.0874, 0.0482, 0.1937, 0.0812, 0.0938, 0.0633, 0.1011], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0151, 0.0142, 0.0128, 0.0143, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:33:35,921 INFO [train.py:904] (3/8) Epoch 19, batch 1950, loss[loss=0.2061, simple_loss=0.2912, pruned_loss=0.06048, over 12223.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2576, pruned_loss=0.04286, over 3304186.03 frames. ], batch size: 247, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:48,882 INFO [optim.py:368] (3/8) 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 20:34:47,042 INFO [train.py:904] (3/8) Epoch 19, batch 2000, loss[loss=0.1817, simple_loss=0.2729, pruned_loss=0.0453, over 17314.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2565, pruned_loss=0.04236, over 3313979.93 frames. ], batch size: 52, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:34:53,085 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184706.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:06,419 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184715.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:10,151 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5433, 3.6093, 4.2225, 2.2293, 3.3842, 2.6922, 4.0771, 3.7983], device='cuda:3'), covar=tensor([0.0252, 0.0968, 0.0414, 0.1917, 0.0761, 0.0908, 0.0553, 0.1064], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0151, 0.0142, 0.0127, 0.0143, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:35:56,926 INFO [train.py:904] (3/8) Epoch 19, batch 2050, loss[loss=0.1955, simple_loss=0.27, pruned_loss=0.06047, over 16868.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2564, pruned_loss=0.04236, over 3320577.95 frames. ], batch size: 116, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:35:59,928 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184754.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:36:10,010 INFO [optim.py:368] (3/8) 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 20:36:43,538 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 20:37:10,295 INFO [train.py:904] (3/8) Epoch 19, batch 2100, loss[loss=0.2099, simple_loss=0.2902, pruned_loss=0.06477, over 12117.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.258, pruned_loss=0.04346, over 3309821.35 frames. ], batch size: 247, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:37:36,119 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1614, 2.0778, 1.7301, 1.7779, 2.3805, 2.0565, 2.1434, 2.4241], device='cuda:3'), covar=tensor([0.0270, 0.0343, 0.0489, 0.0459, 0.0218, 0.0315, 0.0205, 0.0248], device='cuda:3'), in_proj_covar=tensor([0.0204, 0.0236, 0.0225, 0.0227, 0.0236, 0.0236, 0.0239, 0.0230], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:38:05,008 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 2150, loss[loss=0.1826, simple_loss=0.2606, pruned_loss=0.05232, over 16167.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.259, pruned_loss=0.04394, over 3300046.16 frames. ], batch size: 165, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:38:30,739 INFO [optim.py:368] (3/8) 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 20:38:38,199 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184866.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:39:27,103 INFO [train.py:904] (3/8) Epoch 19, batch 2200, loss[loss=0.1887, simple_loss=0.2583, pruned_loss=0.05961, over 16886.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2595, pruned_loss=0.04448, over 3300375.52 frames. ], batch size: 109, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:01,201 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:17,942 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:35,942 INFO [train.py:904] (3/8) Epoch 19, batch 2250, loss[loss=0.2421, simple_loss=0.3243, pruned_loss=0.07992, over 12102.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.26, pruned_loss=0.04445, over 3301448.50 frames. ], batch size: 246, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:48,428 INFO [optim.py:368] (3/8) 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 20:41:27,647 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-30 20:41:36,801 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2639, 2.1825, 2.3561, 4.0669, 2.2059, 2.6338, 2.2725, 2.3840], device='cuda:3'), covar=tensor([0.1401, 0.3779, 0.2796, 0.0576, 0.4008, 0.2381, 0.3867, 0.3228], device='cuda:3'), in_proj_covar=tensor([0.0397, 0.0441, 0.0363, 0.0327, 0.0435, 0.0506, 0.0409, 0.0514], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:41:38,066 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-30 20:41:45,014 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185000.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:41:46,800 INFO [train.py:904] (3/8) Epoch 19, batch 2300, loss[loss=0.1559, simple_loss=0.2592, pruned_loss=0.02634, over 17270.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2606, pruned_loss=0.04484, over 3308213.32 frames. ], batch size: 52, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:42:05,833 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:42:15,415 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 20:42:16,312 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9073, 4.5468, 4.5205, 5.0653, 5.3565, 4.5902, 5.2846, 5.3257], device='cuda:3'), covar=tensor([0.1738, 0.1562, 0.2842, 0.1142, 0.0739, 0.1267, 0.0871, 0.0876], device='cuda:3'), in_proj_covar=tensor([0.0647, 0.0801, 0.0936, 0.0816, 0.0607, 0.0640, 0.0656, 0.0761], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:42:56,744 INFO [train.py:904] (3/8) Epoch 19, batch 2350, loss[loss=0.2074, simple_loss=0.2776, pruned_loss=0.06863, over 16774.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2602, pruned_loss=0.04492, over 3307636.90 frames. ], batch size: 83, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:43:08,920 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:09,639 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:37,457 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4440, 3.3125, 3.6093, 1.8478, 3.6757, 3.7363, 3.0447, 2.7539], device='cuda:3'), covar=tensor([0.0810, 0.0226, 0.0180, 0.1227, 0.0110, 0.0203, 0.0422, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0109, 0.0097, 0.0142, 0.0079, 0.0126, 0.0129, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:44:06,544 INFO [train.py:904] (3/8) Epoch 19, batch 2400, loss[loss=0.1936, simple_loss=0.2897, pruned_loss=0.04877, over 16656.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2619, pruned_loss=0.04557, over 3307949.47 frames. ], batch size: 57, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:44:33,454 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:02,146 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 2450, loss[loss=0.1849, simple_loss=0.287, pruned_loss=0.04138, over 17032.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2629, pruned_loss=0.04559, over 3311800.61 frames. ], batch size: 50, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:45:27,049 INFO [optim.py:368] (3/8) 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 20:45:46,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1504, 5.5395, 5.7276, 5.4301, 5.5096, 6.1069, 5.5855, 5.3204], device='cuda:3'), covar=tensor([0.0859, 0.1994, 0.2400, 0.2092, 0.2618, 0.1013, 0.1513, 0.2415], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0594, 0.0656, 0.0495, 0.0665, 0.0695, 0.0515, 0.0665], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 20:46:08,236 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:08,421 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:23,230 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4518, 5.3860, 5.2681, 4.8194, 4.9249, 5.3420, 5.3421, 4.9193], device='cuda:3'), covar=tensor([0.0612, 0.0484, 0.0295, 0.0334, 0.1074, 0.0471, 0.0291, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0427, 0.0351, 0.0341, 0.0361, 0.0396, 0.0239, 0.0417], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 20:46:23,968 INFO [train.py:904] (3/8) Epoch 19, batch 2500, loss[loss=0.1441, simple_loss=0.2316, pruned_loss=0.02829, over 16865.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2629, pruned_loss=0.0448, over 3316098.37 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:46:51,192 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185222.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:47:32,254 INFO [train.py:904] (3/8) Epoch 19, batch 2550, loss[loss=0.1581, simple_loss=0.2421, pruned_loss=0.03709, over 16796.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2629, pruned_loss=0.04519, over 3305739.10 frames. ], batch size: 83, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:47:44,039 INFO [optim.py:368] (3/8) 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 20:48:04,399 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6564, 4.0301, 4.0175, 2.7688, 3.5906, 4.0085, 3.7645, 1.9812], device='cuda:3'), covar=tensor([0.0529, 0.0109, 0.0075, 0.0453, 0.0142, 0.0143, 0.0139, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0134, 0.0096, 0.0106, 0.0093, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 20:48:31,912 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:48:40,161 INFO [train.py:904] (3/8) Epoch 19, batch 2600, loss[loss=0.1657, simple_loss=0.2561, pruned_loss=0.0377, over 17235.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2623, pruned_loss=0.04437, over 3316969.33 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:49:50,076 INFO [train.py:904] (3/8) Epoch 19, batch 2650, loss[loss=0.1986, simple_loss=0.2935, pruned_loss=0.05185, over 17060.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2621, pruned_loss=0.04324, over 3319698.65 frames. ], batch size: 53, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:50:03,487 INFO [optim.py:368] (3/8) 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 20:50:58,285 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1708, 3.8952, 4.5165, 2.3768, 4.8736, 4.8118, 3.2597, 3.7605], device='cuda:3'), covar=tensor([0.0705, 0.0260, 0.0197, 0.1169, 0.0063, 0.0134, 0.0470, 0.0372], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0141, 0.0079, 0.0126, 0.0128, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 20:50:59,021 INFO [train.py:904] (3/8) Epoch 19, batch 2700, loss[loss=0.171, simple_loss=0.2668, pruned_loss=0.03761, over 16231.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2624, pruned_loss=0.04294, over 3314987.74 frames. ], batch size: 35, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:51:19,123 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185416.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 20:51:56,419 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6068, 5.9058, 5.6663, 5.7337, 5.3391, 5.2451, 5.3147, 6.0597], device='cuda:3'), covar=tensor([0.1254, 0.0905, 0.1101, 0.0826, 0.0895, 0.0743, 0.1308, 0.0901], device='cuda:3'), in_proj_covar=tensor([0.0666, 0.0820, 0.0677, 0.0609, 0.0514, 0.0522, 0.0682, 0.0636], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:52:08,642 INFO [train.py:904] (3/8) Epoch 19, batch 2750, loss[loss=0.1787, simple_loss=0.2733, pruned_loss=0.04209, over 16996.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2622, pruned_loss=0.0423, over 3316944.81 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:52:20,528 INFO [optim.py:368] (3/8) 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 20:52:30,895 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:54,424 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185485.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:03,470 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 20:53:10,253 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 20:53:18,343 INFO [train.py:904] (3/8) Epoch 19, batch 2800, loss[loss=0.1517, simple_loss=0.2473, pruned_loss=0.02802, over 17110.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2621, pruned_loss=0.04268, over 3316137.91 frames. ], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:53:46,643 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185522.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:56,322 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 2850, loss[loss=0.1833, simple_loss=0.2605, pruned_loss=0.05304, over 16719.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.261, pruned_loss=0.04273, over 3311568.40 frames. ], batch size: 134, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:54:41,475 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:28,839 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185595.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:34,566 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6086, 5.9528, 5.7336, 5.8429, 5.3485, 5.4170, 5.3972, 6.1304], device='cuda:3'), covar=tensor([0.1521, 0.1100, 0.1216, 0.0825, 0.0989, 0.0690, 0.1230, 0.0982], device='cuda:3'), in_proj_covar=tensor([0.0673, 0.0828, 0.0684, 0.0614, 0.0520, 0.0527, 0.0690, 0.0640], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 20:55:39,056 INFO [train.py:904] (3/8) Epoch 19, batch 2900, loss[loss=0.1459, simple_loss=0.2315, pruned_loss=0.03014, over 16755.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2594, pruned_loss=0.04297, over 3313782.95 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:09,341 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 20:56:36,416 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:56:47,750 INFO [train.py:904] (3/8) Epoch 19, batch 2950, loss[loss=0.1535, simple_loss=0.233, pruned_loss=0.037, over 16841.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2591, pruned_loss=0.04368, over 3304590.85 frames. ], batch size: 102, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:00,932 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.308e+02 2.753e+02 3.431e+02 9.141e+02, threshold=5.506e+02, percent-clipped=4.0 2023-04-30 20:57:58,854 INFO [train.py:904] (3/8) Epoch 19, batch 3000, loss[loss=0.1975, simple_loss=0.2744, pruned_loss=0.06035, over 16849.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.26, pruned_loss=0.04409, over 3302887.20 frames. ], batch size: 90, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:58,854 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 20:58:07,639 INFO [train.py:938] (3/8) Epoch 19, validation: loss=0.1362, simple_loss=0.2416, pruned_loss=0.01538, over 944034.00 frames. 2023-04-30 20:58:07,640 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 20:58:28,261 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185716.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 20:58:57,118 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-30 20:59:16,277 INFO [train.py:904] (3/8) Epoch 19, batch 3050, loss[loss=0.1461, simple_loss=0.2355, pruned_loss=0.0284, over 16839.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2599, pruned_loss=0.04386, over 3305475.01 frames. ], batch size: 42, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:59:28,147 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 20:59:35,050 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:03,979 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:23,314 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185799.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:00:26,857 INFO [train.py:904] (3/8) Epoch 19, batch 3100, loss[loss=0.1703, simple_loss=0.2469, pruned_loss=0.04685, over 16815.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2597, pruned_loss=0.0439, over 3306052.21 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:00:51,303 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185820.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:00:56,478 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185824.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:58,615 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 21:01:09,058 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185833.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:34,486 INFO [train.py:904] (3/8) Epoch 19, batch 3150, loss[loss=0.1877, simple_loss=0.2623, pruned_loss=0.05651, over 16871.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2584, pruned_loss=0.04337, over 3307496.17 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:01:46,031 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185860.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:01:47,104 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.254e+02 2.659e+02 3.271e+02 6.332e+02, threshold=5.317e+02, percent-clipped=3.0 2023-04-30 21:02:18,038 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2222, 3.3484, 3.6918, 2.1625, 3.1125, 2.3865, 3.6773, 3.6996], device='cuda:3'), covar=tensor([0.0238, 0.0952, 0.0529, 0.1950, 0.0837, 0.1005, 0.0552, 0.0930], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0161, 0.0165, 0.0151, 0.0143, 0.0127, 0.0142, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:02:43,074 INFO [train.py:904] (3/8) Epoch 19, batch 3200, loss[loss=0.1507, simple_loss=0.2346, pruned_loss=0.03338, over 16768.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2575, pruned_loss=0.04301, over 3316363.90 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:02:43,582 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8728, 2.8093, 2.5403, 4.1330, 3.4028, 4.1011, 1.5727, 3.0048], device='cuda:3'), covar=tensor([0.1303, 0.0672, 0.1178, 0.0170, 0.0169, 0.0393, 0.1575, 0.0768], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0184, 0.0204, 0.0216, 0.0197, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:03:25,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8075, 3.9924, 3.0296, 2.3824, 2.6482, 2.4722, 4.0889, 3.4880], device='cuda:3'), covar=tensor([0.2458, 0.0539, 0.1581, 0.2775, 0.2454, 0.1900, 0.0461, 0.1398], device='cuda:3'), in_proj_covar=tensor([0.0321, 0.0266, 0.0301, 0.0305, 0.0296, 0.0252, 0.0290, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:03:53,534 INFO [train.py:904] (3/8) Epoch 19, batch 3250, loss[loss=0.1884, simple_loss=0.2669, pruned_loss=0.05494, over 16904.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2586, pruned_loss=0.0435, over 3322414.60 frames. ], batch size: 109, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:04:06,305 INFO [optim.py:368] (3/8) 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 21:05:07,481 INFO [train.py:904] (3/8) Epoch 19, batch 3300, loss[loss=0.1824, simple_loss=0.2631, pruned_loss=0.05081, over 16889.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2592, pruned_loss=0.04367, over 3322501.85 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:05:25,475 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6882, 1.7765, 2.2597, 2.5461, 2.6145, 2.6433, 1.9026, 2.8060], device='cuda:3'), covar=tensor([0.0177, 0.0473, 0.0336, 0.0273, 0.0296, 0.0270, 0.0519, 0.0164], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0193, 0.0180, 0.0183, 0.0193, 0.0152, 0.0197, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:06:15,807 INFO [train.py:904] (3/8) Epoch 19, batch 3350, loss[loss=0.1631, simple_loss=0.2631, pruned_loss=0.03158, over 17261.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2592, pruned_loss=0.04317, over 3328147.96 frames. ], batch size: 52, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:27,998 INFO [optim.py:368] (3/8) 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 21:06:59,999 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1606, 3.2955, 3.0513, 5.2369, 4.2404, 4.6007, 2.1847, 3.4420], device='cuda:3'), covar=tensor([0.1230, 0.0709, 0.1044, 0.0172, 0.0239, 0.0364, 0.1398, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0185, 0.0205, 0.0217, 0.0198, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:07:23,694 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 3400, loss[loss=0.1721, simple_loss=0.2623, pruned_loss=0.04096, over 16605.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2587, pruned_loss=0.0432, over 3324172.94 frames. ], batch size: 62, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:07:33,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5243, 2.2929, 1.7815, 2.0524, 2.6338, 2.4037, 2.5193, 2.7482], device='cuda:3'), covar=tensor([0.0211, 0.0368, 0.0523, 0.0409, 0.0211, 0.0289, 0.0184, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0206, 0.0236, 0.0225, 0.0227, 0.0237, 0.0235, 0.0243, 0.0230], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:07:44,670 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186115.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:07:57,584 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186124.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:02,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3382, 5.8171, 5.9210, 5.6670, 5.7167, 6.2666, 5.7989, 5.4753], device='cuda:3'), covar=tensor([0.0788, 0.2027, 0.2261, 0.1960, 0.2634, 0.0936, 0.1414, 0.2469], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0593, 0.0659, 0.0497, 0.0667, 0.0693, 0.0511, 0.0666], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:08:34,723 INFO [train.py:904] (3/8) Epoch 19, batch 3450, loss[loss=0.1813, simple_loss=0.2549, pruned_loss=0.05387, over 16895.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2579, pruned_loss=0.04297, over 3322305.21 frames. ], batch size: 90, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:08:38,533 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186155.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:08:44,140 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:08:48,334 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:02,954 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:45,133 INFO [train.py:904] (3/8) Epoch 19, batch 3500, loss[loss=0.1816, simple_loss=0.2635, pruned_loss=0.04991, over 16290.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2573, pruned_loss=0.04293, over 3326677.91 frames. ], batch size: 165, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:10:08,975 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:10:22,023 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8486, 1.9422, 2.3393, 2.6899, 2.7281, 2.6935, 1.9721, 2.9324], device='cuda:3'), covar=tensor([0.0152, 0.0426, 0.0297, 0.0240, 0.0256, 0.0262, 0.0459, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0192, 0.0179, 0.0182, 0.0192, 0.0152, 0.0195, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:10:55,186 INFO [train.py:904] (3/8) Epoch 19, batch 3550, loss[loss=0.152, simple_loss=0.2427, pruned_loss=0.03061, over 17227.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2564, pruned_loss=0.04256, over 3316792.96 frames. ], batch size: 44, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:11:01,612 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9805, 2.4550, 1.8257, 2.2736, 2.8116, 2.6831, 2.9525, 3.0161], device='cuda:3'), covar=tensor([0.0181, 0.0403, 0.0595, 0.0463, 0.0261, 0.0324, 0.0221, 0.0281], device='cuda:3'), in_proj_covar=tensor([0.0207, 0.0237, 0.0226, 0.0227, 0.0238, 0.0235, 0.0243, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:11:05,156 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:11:29,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0264, 2.0579, 2.5862, 2.9475, 2.8037, 3.4862, 2.4109, 3.4840], device='cuda:3'), covar=tensor([0.0240, 0.0459, 0.0303, 0.0312, 0.0297, 0.0190, 0.0427, 0.0151], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0191, 0.0179, 0.0182, 0.0192, 0.0152, 0.0195, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:11:30,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9264, 4.4207, 3.1591, 2.4128, 2.8049, 2.6125, 4.7535, 3.7258], device='cuda:3'), covar=tensor([0.2798, 0.0528, 0.1762, 0.2878, 0.2905, 0.2019, 0.0391, 0.1275], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0268, 0.0301, 0.0306, 0.0297, 0.0253, 0.0290, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:11:41,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3370, 3.3143, 3.5991, 2.5220, 3.3868, 3.7003, 3.4240, 2.1936], device='cuda:3'), covar=tensor([0.0479, 0.0165, 0.0053, 0.0363, 0.0097, 0.0084, 0.0082, 0.0431], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0133, 0.0095, 0.0106, 0.0093, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:11:54,377 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 21:12:03,397 INFO [train.py:904] (3/8) Epoch 19, batch 3600, loss[loss=0.1502, simple_loss=0.2474, pruned_loss=0.02647, over 17097.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2561, pruned_loss=0.04225, over 3320294.39 frames. ], batch size: 49, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:12:28,384 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:13:13,925 INFO [train.py:904] (3/8) Epoch 19, batch 3650, loss[loss=0.1825, simple_loss=0.2536, pruned_loss=0.05564, over 16744.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2547, pruned_loss=0.04305, over 3304256.32 frames. ], batch size: 124, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:13:27,661 INFO [optim.py:368] (3/8) 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 21:14:16,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3216, 3.5316, 3.7509, 2.4787, 3.4909, 3.8388, 3.5735, 2.1389], device='cuda:3'), covar=tensor([0.0515, 0.0132, 0.0056, 0.0398, 0.0097, 0.0091, 0.0087, 0.0473], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0081, 0.0081, 0.0132, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:14:28,525 INFO [train.py:904] (3/8) Epoch 19, batch 3700, loss[loss=0.185, simple_loss=0.259, pruned_loss=0.05551, over 16491.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2532, pruned_loss=0.04438, over 3295304.25 frames. ], batch size: 146, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:14:48,795 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186415.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:14:57,163 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 3750, loss[loss=0.1645, simple_loss=0.2416, pruned_loss=0.04364, over 16561.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2538, pruned_loss=0.04567, over 3277658.02 frames. ], batch size: 68, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:15:46,684 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186455.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:15:47,759 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186463.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:00,546 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8466, 4.0646, 3.0692, 2.4951, 2.7811, 2.6950, 4.1602, 3.7055], device='cuda:3'), covar=tensor([0.2720, 0.0571, 0.1740, 0.2726, 0.2562, 0.1835, 0.0526, 0.1196], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0267, 0.0301, 0.0306, 0.0298, 0.0253, 0.0290, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:16:15,879 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186475.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:18,159 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4180, 4.4867, 4.7929, 4.7648, 4.8235, 4.4900, 4.4905, 4.3293], device='cuda:3'), covar=tensor([0.0424, 0.0652, 0.0414, 0.0424, 0.0522, 0.0443, 0.0904, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0448, 0.0433, 0.0406, 0.0478, 0.0457, 0.0552, 0.0365], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 21:16:23,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4929, 4.5262, 4.6939, 4.5288, 4.5610, 5.1382, 4.6583, 4.3590], device='cuda:3'), covar=tensor([0.1657, 0.2007, 0.2120, 0.2022, 0.2591, 0.1068, 0.1566, 0.2594], device='cuda:3'), in_proj_covar=tensor([0.0404, 0.0587, 0.0648, 0.0491, 0.0655, 0.0685, 0.0508, 0.0657], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:16:26,550 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:27,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1196, 5.1604, 4.9923, 4.6245, 4.6531, 5.0637, 4.9445, 4.7959], device='cuda:3'), covar=tensor([0.0611, 0.0442, 0.0302, 0.0317, 0.1008, 0.0389, 0.0319, 0.0611], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0429, 0.0352, 0.0343, 0.0362, 0.0398, 0.0240, 0.0417], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:16:53,853 INFO [train.py:904] (3/8) Epoch 19, batch 3800, loss[loss=0.1791, simple_loss=0.2637, pruned_loss=0.04725, over 16599.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2552, pruned_loss=0.04687, over 3286412.04 frames. ], batch size: 62, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:16:55,938 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186503.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:17:11,596 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:15,468 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:43,665 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186536.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:47,976 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 3850, loss[loss=0.2044, simple_loss=0.286, pruned_loss=0.06134, over 12322.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2552, pruned_loss=0.04734, over 3273922.05 frames. ], batch size: 246, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:18:22,160 INFO [optim.py:368] (3/8) 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 21:18:28,820 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:40,297 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 21:18:45,030 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:56,181 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186586.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:16,900 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 3900, loss[loss=0.168, simple_loss=0.2419, pruned_loss=0.04703, over 16793.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2547, pruned_loss=0.04792, over 3267009.12 frames. ], batch size: 102, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:19:27,541 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-30 21:19:37,487 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:39,323 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9637, 4.9091, 4.8772, 4.5604, 4.5127, 4.9282, 4.7682, 4.6561], device='cuda:3'), covar=tensor([0.0619, 0.0544, 0.0310, 0.0308, 0.0945, 0.0468, 0.0408, 0.0610], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0425, 0.0349, 0.0340, 0.0360, 0.0396, 0.0239, 0.0414], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:19:55,048 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:20,753 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:28,330 INFO [train.py:904] (3/8) Epoch 19, batch 3950, loss[loss=0.1663, simple_loss=0.2405, pruned_loss=0.04606, over 16856.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2547, pruned_loss=0.0489, over 3263090.64 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 4.0 2023-04-30 21:20:42,973 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.601e+02 2.399e+02 2.822e+02 3.355e+02 8.994e+02, threshold=5.644e+02, percent-clipped=5.0 2023-04-30 21:21:11,756 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4213, 2.3353, 2.4047, 4.1832, 2.2999, 2.7165, 2.4547, 2.5401], device='cuda:3'), covar=tensor([0.1262, 0.3543, 0.2694, 0.0555, 0.3660, 0.2264, 0.3384, 0.2944], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0441, 0.0363, 0.0328, 0.0434, 0.0509, 0.0410, 0.0516], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:21:20,570 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3821, 4.3879, 4.3550, 3.8068, 4.3700, 1.7272, 4.1337, 3.9397], device='cuda:3'), covar=tensor([0.0116, 0.0104, 0.0172, 0.0298, 0.0093, 0.2778, 0.0147, 0.0251], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0150, 0.0196, 0.0179, 0.0173, 0.0206, 0.0188, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:21:39,185 INFO [train.py:904] (3/8) Epoch 19, batch 4000, loss[loss=0.1913, simple_loss=0.2584, pruned_loss=0.06207, over 16883.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2545, pruned_loss=0.04891, over 3268637.90 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:21:46,151 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6661, 4.6511, 4.5844, 3.9738, 4.6231, 1.9700, 4.3978, 4.2463], device='cuda:3'), covar=tensor([0.0119, 0.0115, 0.0178, 0.0321, 0.0094, 0.2594, 0.0152, 0.0220], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0150, 0.0196, 0.0179, 0.0173, 0.0205, 0.0188, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:21:49,038 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186709.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:22:49,925 INFO [train.py:904] (3/8) Epoch 19, batch 4050, loss[loss=0.177, simple_loss=0.262, pruned_loss=0.04604, over 16457.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2549, pruned_loss=0.04807, over 3270110.21 frames. ], batch size: 68, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:22:57,313 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:06,010 INFO [optim.py:368] (3/8) 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 21:23:16,220 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:26,524 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:41,773 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 21:24:03,614 INFO [train.py:904] (3/8) Epoch 19, batch 4100, loss[loss=0.1792, simple_loss=0.2693, pruned_loss=0.04456, over 16669.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2562, pruned_loss=0.04724, over 3263962.97 frames. ], batch size: 89, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:24:07,124 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:21,796 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:42,364 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4998, 4.6513, 4.3560, 3.0125, 3.9746, 4.5135, 3.9062, 2.4912], device='cuda:3'), covar=tensor([0.0519, 0.0020, 0.0045, 0.0365, 0.0081, 0.0069, 0.0080, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0080, 0.0080, 0.0132, 0.0095, 0.0105, 0.0092, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:24:46,677 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:59,853 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:17,893 INFO [train.py:904] (3/8) Epoch 19, batch 4150, loss[loss=0.1798, simple_loss=0.2808, pruned_loss=0.03943, over 16849.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2632, pruned_loss=0.04931, over 3244702.10 frames. ], batch size: 96, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:25:34,163 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:34,923 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.347e+02 2.016e+02 2.433e+02 2.890e+02 5.187e+02, threshold=4.866e+02, percent-clipped=4.0 2023-04-30 21:25:49,713 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186873.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:23,124 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:31,958 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186901.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:26:32,669 INFO [train.py:904] (3/8) Epoch 19, batch 4200, loss[loss=0.2099, simple_loss=0.31, pruned_loss=0.0549, over 17109.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2703, pruned_loss=0.05136, over 3191125.61 frames. ], batch size: 49, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:26:47,046 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:53,367 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:06,336 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:34,741 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:49,658 INFO [train.py:904] (3/8) Epoch 19, batch 4250, loss[loss=0.1699, simple_loss=0.2668, pruned_loss=0.03651, over 16481.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2739, pruned_loss=0.05127, over 3186512.94 frames. ], batch size: 146, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:27:50,565 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 21:28:05,619 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.275e+02 2.568e+02 3.131e+02 7.157e+02, threshold=5.135e+02, percent-clipped=7.0 2023-04-30 21:28:06,528 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:28:19,197 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 4300, loss[loss=0.2131, simple_loss=0.2966, pruned_loss=0.06475, over 17025.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2755, pruned_loss=0.05066, over 3183281.93 frames. ], batch size: 55, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:29:37,705 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187025.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:56,583 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 4350, loss[loss=0.1922, simple_loss=0.283, pruned_loss=0.05068, over 16833.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2786, pruned_loss=0.05149, over 3181687.59 frames. ], batch size: 83, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:30:32,929 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:53,909 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:07,463 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187086.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:31:26,244 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187099.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:26,352 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5591, 3.6808, 2.7082, 2.2627, 2.4345, 2.3487, 4.0219, 3.3722], device='cuda:3'), covar=tensor([0.2931, 0.0690, 0.1890, 0.2424, 0.2504, 0.1982, 0.0408, 0.1116], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0266, 0.0301, 0.0305, 0.0298, 0.0251, 0.0290, 0.0330], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:31:31,215 INFO [train.py:904] (3/8) Epoch 19, batch 4400, loss[loss=0.2151, simple_loss=0.3023, pruned_loss=0.06397, over 16472.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2809, pruned_loss=0.05277, over 3174407.10 frames. ], batch size: 35, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:05,653 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:13,526 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 4450, loss[loss=0.2166, simple_loss=0.3065, pruned_loss=0.06336, over 16839.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2841, pruned_loss=0.05402, over 3180602.75 frames. ], batch size: 102, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:47,781 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0102, 4.8597, 5.0514, 5.1957, 5.3235, 4.7389, 5.3647, 5.3662], device='cuda:3'), covar=tensor([0.1533, 0.1074, 0.1315, 0.0565, 0.0435, 0.0760, 0.0485, 0.0506], device='cuda:3'), in_proj_covar=tensor([0.0620, 0.0767, 0.0900, 0.0784, 0.0585, 0.0612, 0.0630, 0.0733], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:33:00,621 INFO [optim.py:368] (3/8) 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 21:33:15,037 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187173.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:24,884 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:35,476 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9944, 5.0384, 4.8890, 4.5816, 4.5667, 4.9508, 4.6659, 4.6278], device='cuda:3'), covar=tensor([0.0446, 0.0243, 0.0194, 0.0195, 0.0734, 0.0269, 0.0339, 0.0515], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0404, 0.0335, 0.0325, 0.0345, 0.0376, 0.0229, 0.0397], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:33:47,936 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187195.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:49,993 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187196.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:33:57,507 INFO [train.py:904] (3/8) Epoch 19, batch 4500, loss[loss=0.2049, simple_loss=0.2923, pruned_loss=0.05877, over 16901.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2845, pruned_loss=0.05484, over 3178494.96 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:34:24,640 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:27,128 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:54,213 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:55,420 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187243.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:08,490 INFO [train.py:904] (3/8) Epoch 19, batch 4550, loss[loss=0.1923, simple_loss=0.2807, pruned_loss=0.05196, over 16567.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2852, pruned_loss=0.0555, over 3195705.90 frames. ], batch size: 68, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:35:24,435 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:36,696 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187271.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:43,219 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9534, 2.1209, 2.1279, 3.6033, 2.0279, 2.4309, 2.2322, 2.2219], device='cuda:3'), covar=tensor([0.1406, 0.3409, 0.2922, 0.0566, 0.4297, 0.2404, 0.3284, 0.3448], device='cuda:3'), in_proj_covar=tensor([0.0395, 0.0438, 0.0358, 0.0323, 0.0432, 0.0505, 0.0407, 0.0511], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:36:01,826 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-30 21:36:05,135 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 4600, loss[loss=0.1821, simple_loss=0.2712, pruned_loss=0.04654, over 17084.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2868, pruned_loss=0.0562, over 3203438.51 frames. ], batch size: 53, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:33,043 INFO [train.py:904] (3/8) Epoch 19, batch 4650, loss[loss=0.2394, simple_loss=0.3088, pruned_loss=0.08501, over 11823.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.286, pruned_loss=0.05589, over 3219198.76 frames. ], batch size: 248, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:49,309 INFO [optim.py:368] (3/8) 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 21:37:52,218 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:01,617 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:14,872 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187381.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:38:33,303 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:43,286 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-30 21:38:44,846 INFO [train.py:904] (3/8) Epoch 19, batch 4700, loss[loss=0.185, simple_loss=0.2718, pruned_loss=0.0491, over 16915.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2829, pruned_loss=0.05469, over 3217877.45 frames. ], batch size: 109, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:38:53,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1721, 4.1848, 4.1100, 3.3012, 4.1171, 1.6867, 3.9146, 3.6592], device='cuda:3'), covar=tensor([0.0187, 0.0176, 0.0167, 0.0417, 0.0134, 0.2883, 0.0164, 0.0318], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0146, 0.0191, 0.0175, 0.0168, 0.0201, 0.0183, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:38:59,525 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 21:39:01,715 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187413.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:29,974 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:34,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8396, 4.4703, 4.0632, 4.8042, 4.9897, 4.6618, 5.0652, 5.0300], device='cuda:3'), covar=tensor([0.1362, 0.1402, 0.3217, 0.1259, 0.1131, 0.1054, 0.1192, 0.1270], device='cuda:3'), in_proj_covar=tensor([0.0607, 0.0750, 0.0883, 0.0767, 0.0574, 0.0599, 0.0618, 0.0718], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:39:55,445 INFO [train.py:904] (3/8) Epoch 19, batch 4750, loss[loss=0.1801, simple_loss=0.2621, pruned_loss=0.04903, over 16642.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2788, pruned_loss=0.05266, over 3213695.21 frames. ], batch size: 57, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:40:11,094 INFO [optim.py:368] (3/8) 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 21:40:57,693 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187496.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:05,669 INFO [train.py:904] (3/8) Epoch 19, batch 4800, loss[loss=0.2074, simple_loss=0.2845, pruned_loss=0.0651, over 12018.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2747, pruned_loss=0.05027, over 3215974.19 frames. ], batch size: 246, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:41:20,834 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:28,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9576, 2.7646, 2.8535, 2.1091, 2.6345, 2.1552, 2.6781, 2.9150], device='cuda:3'), covar=tensor([0.0292, 0.0705, 0.0536, 0.1628, 0.0810, 0.0907, 0.0646, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0150, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:41:30,317 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:37,770 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187523.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:40,265 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2296, 5.2010, 5.1432, 4.3110, 5.1392, 1.8847, 4.8774, 5.0010], device='cuda:3'), covar=tensor([0.0077, 0.0096, 0.0137, 0.0528, 0.0104, 0.2612, 0.0130, 0.0199], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0145, 0.0190, 0.0174, 0.0167, 0.0200, 0.0181, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:42:08,454 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187544.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:08,942 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 21:42:19,532 INFO [train.py:904] (3/8) Epoch 19, batch 4850, loss[loss=0.1896, simple_loss=0.291, pruned_loss=0.04407, over 16236.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2755, pruned_loss=0.04973, over 3202736.77 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:42:24,854 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1828, 2.9963, 3.0954, 1.8174, 3.2780, 3.3156, 2.7650, 2.5173], device='cuda:3'), covar=tensor([0.0795, 0.0236, 0.0189, 0.1101, 0.0075, 0.0159, 0.0399, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0107, 0.0096, 0.0137, 0.0077, 0.0122, 0.0126, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 21:42:36,204 INFO [optim.py:368] (3/8) 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 21:42:41,898 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:50,831 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:01,996 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:07,605 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:18,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7238, 2.7517, 2.6255, 1.8998, 2.5306, 2.6680, 2.5917, 1.7932], device='cuda:3'), covar=tensor([0.0473, 0.0081, 0.0070, 0.0363, 0.0127, 0.0119, 0.0123, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0078, 0.0079, 0.0131, 0.0094, 0.0104, 0.0091, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:43:32,170 INFO [train.py:904] (3/8) Epoch 19, batch 4900, loss[loss=0.167, simple_loss=0.263, pruned_loss=0.03552, over 16742.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2752, pruned_loss=0.04866, over 3182187.24 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:43:51,071 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:44:25,017 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0314, 3.1285, 1.9921, 3.2957, 2.3747, 3.3134, 2.0478, 2.5367], device='cuda:3'), covar=tensor([0.0320, 0.0340, 0.1432, 0.0159, 0.0836, 0.0542, 0.1550, 0.0756], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0173, 0.0190, 0.0154, 0.0172, 0.0211, 0.0196, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:44:43,107 INFO [train.py:904] (3/8) Epoch 19, batch 4950, loss[loss=0.1919, simple_loss=0.2826, pruned_loss=0.05062, over 16858.00 frames. ], tot_loss[loss=0.186, simple_loss=0.275, pruned_loss=0.04846, over 3174598.42 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:44:58,341 INFO [optim.py:368] (3/8) 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 21:45:24,843 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187681.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:45:39,672 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2705, 4.3060, 4.5940, 4.5627, 4.5580, 4.2697, 4.2570, 4.1904], device='cuda:3'), covar=tensor([0.0270, 0.0489, 0.0336, 0.0359, 0.0526, 0.0370, 0.0876, 0.0444], device='cuda:3'), in_proj_covar=tensor([0.0385, 0.0425, 0.0414, 0.0386, 0.0458, 0.0435, 0.0522, 0.0343], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 21:45:43,933 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:45:55,107 INFO [train.py:904] (3/8) Epoch 19, batch 5000, loss[loss=0.1862, simple_loss=0.2789, pruned_loss=0.04679, over 16657.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.277, pruned_loss=0.04895, over 3175399.33 frames. ], batch size: 134, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:46:32,202 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:52,593 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:47:07,092 INFO [train.py:904] (3/8) Epoch 19, batch 5050, loss[loss=0.21, simple_loss=0.2926, pruned_loss=0.06368, over 12438.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2775, pruned_loss=0.04843, over 3190257.03 frames. ], batch size: 247, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:47:21,719 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.150e+02 2.460e+02 2.787e+02 5.516e+02, threshold=4.919e+02, percent-clipped=2.0 2023-04-30 21:47:34,786 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0221, 3.9735, 3.9508, 3.1359, 3.9216, 1.7892, 3.7005, 3.4903], device='cuda:3'), covar=tensor([0.0122, 0.0121, 0.0151, 0.0422, 0.0111, 0.2698, 0.0151, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0144, 0.0188, 0.0174, 0.0166, 0.0199, 0.0180, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:48:17,850 INFO [train.py:904] (3/8) Epoch 19, batch 5100, loss[loss=0.1662, simple_loss=0.2545, pruned_loss=0.03892, over 17108.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2755, pruned_loss=0.0477, over 3190240.92 frames. ], batch size: 49, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:48:41,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4660, 4.7927, 4.3086, 4.6986, 4.3525, 4.2860, 4.2860, 4.8582], device='cuda:3'), covar=tensor([0.1921, 0.1353, 0.2210, 0.1184, 0.1561, 0.1994, 0.2159, 0.1564], device='cuda:3'), in_proj_covar=tensor([0.0642, 0.0793, 0.0652, 0.0589, 0.0500, 0.0505, 0.0658, 0.0611], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:49:04,859 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0934, 2.1447, 2.1515, 3.7754, 2.0583, 2.4993, 2.2290, 2.2967], device='cuda:3'), covar=tensor([0.1292, 0.3506, 0.2838, 0.0480, 0.3903, 0.2442, 0.3552, 0.3099], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0435, 0.0358, 0.0321, 0.0430, 0.0502, 0.0405, 0.0508], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:49:30,704 INFO [train.py:904] (3/8) Epoch 19, batch 5150, loss[loss=0.1606, simple_loss=0.242, pruned_loss=0.03962, over 16213.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2745, pruned_loss=0.04667, over 3176734.39 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:47,460 INFO [optim.py:368] (3/8) 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 21:49:54,398 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187868.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:04,862 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187875.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:11,342 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187879.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:15,486 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-04-30 21:50:41,937 INFO [train.py:904] (3/8) Epoch 19, batch 5200, loss[loss=0.1847, simple_loss=0.2751, pruned_loss=0.04713, over 16933.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2727, pruned_loss=0.04596, over 3181324.10 frames. ], batch size: 116, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:50:43,069 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0337, 4.1626, 2.6743, 4.9032, 3.3395, 4.7686, 2.7435, 3.2576], device='cuda:3'), covar=tensor([0.0228, 0.0288, 0.1453, 0.0136, 0.0785, 0.0387, 0.1406, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0173, 0.0190, 0.0154, 0.0172, 0.0211, 0.0197, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:51:18,723 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2302, 4.3451, 4.5029, 4.2427, 4.3684, 4.8353, 4.3876, 4.0475], device='cuda:3'), covar=tensor([0.1698, 0.1852, 0.1913, 0.2113, 0.2412, 0.1050, 0.1466, 0.2457], device='cuda:3'), in_proj_covar=tensor([0.0394, 0.0564, 0.0618, 0.0473, 0.0629, 0.0655, 0.0484, 0.0633], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 21:51:53,322 INFO [train.py:904] (3/8) Epoch 19, batch 5250, loss[loss=0.1649, simple_loss=0.2492, pruned_loss=0.04026, over 17237.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2702, pruned_loss=0.04577, over 3189612.75 frames. ], batch size: 45, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:52:08,308 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.924e+02 2.321e+02 2.643e+02 4.137e+02, threshold=4.643e+02, percent-clipped=0.0 2023-04-30 21:53:08,600 INFO [train.py:904] (3/8) Epoch 19, batch 5300, loss[loss=0.1637, simple_loss=0.2512, pruned_loss=0.03806, over 16558.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2665, pruned_loss=0.04472, over 3207128.94 frames. ], batch size: 68, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:53:46,806 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:53:52,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5288, 3.6367, 3.3827, 3.1110, 3.0991, 3.5002, 3.2900, 3.2322], device='cuda:3'), covar=tensor([0.0633, 0.0618, 0.0364, 0.0313, 0.0773, 0.0545, 0.1716, 0.0569], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0405, 0.0334, 0.0324, 0.0345, 0.0379, 0.0228, 0.0396], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:54:04,424 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 21:54:21,263 INFO [train.py:904] (3/8) Epoch 19, batch 5350, loss[loss=0.1852, simple_loss=0.2645, pruned_loss=0.05291, over 11907.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2645, pruned_loss=0.04376, over 3210101.72 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:54:24,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4619, 4.4370, 4.3392, 3.5682, 4.3688, 1.6864, 4.1130, 3.9947], device='cuda:3'), covar=tensor([0.0090, 0.0083, 0.0150, 0.0389, 0.0099, 0.2742, 0.0130, 0.0270], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0144, 0.0188, 0.0173, 0.0165, 0.0199, 0.0180, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:54:28,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5976, 2.9906, 3.3092, 1.9927, 2.8572, 2.0546, 3.3046, 3.3628], device='cuda:3'), covar=tensor([0.0214, 0.0764, 0.0547, 0.1914, 0.0788, 0.1021, 0.0538, 0.0677], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0150, 0.0143, 0.0127, 0.0142, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:54:38,012 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 19, batch 5400, loss[loss=0.1921, simple_loss=0.2821, pruned_loss=0.05103, over 16656.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2674, pruned_loss=0.04429, over 3217840.13 frames. ], batch size: 62, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:56:42,992 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-30 21:56:53,452 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9259, 2.2551, 2.3656, 2.7104, 1.9377, 3.2569, 1.7481, 2.7156], device='cuda:3'), covar=tensor([0.1130, 0.0678, 0.0995, 0.0149, 0.0117, 0.0334, 0.1416, 0.0650], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0182, 0.0203, 0.0213, 0.0196, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 21:56:54,045 INFO [train.py:904] (3/8) Epoch 19, batch 5450, loss[loss=0.1773, simple_loss=0.2616, pruned_loss=0.0465, over 12051.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2701, pruned_loss=0.04549, over 3193521.36 frames. ], batch size: 246, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:57:02,852 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3854, 3.3704, 3.4006, 3.4845, 3.5194, 3.2499, 3.4914, 3.5594], device='cuda:3'), covar=tensor([0.1321, 0.0936, 0.1181, 0.0717, 0.0712, 0.2798, 0.1102, 0.0916], device='cuda:3'), in_proj_covar=tensor([0.0617, 0.0759, 0.0894, 0.0780, 0.0581, 0.0609, 0.0624, 0.0728], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:57:11,918 INFO [optim.py:368] (3/8) 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 21:57:19,745 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188168.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:32,279 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188175.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:38,522 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:14,499 INFO [train.py:904] (3/8) Epoch 19, batch 5500, loss[loss=0.2881, simple_loss=0.3538, pruned_loss=0.1112, over 11930.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2772, pruned_loss=0.0499, over 3159692.26 frames. ], batch size: 246, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:58:20,577 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1234, 1.5342, 1.9905, 2.0529, 2.1653, 2.3605, 1.7115, 2.3687], device='cuda:3'), covar=tensor([0.0202, 0.0415, 0.0251, 0.0297, 0.0267, 0.0179, 0.0399, 0.0114], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0180, 0.0189, 0.0149, 0.0192, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:58:38,577 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:52,387 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0099, 2.1509, 2.2994, 3.4993, 2.0638, 2.5082, 2.2739, 2.3016], device='cuda:3'), covar=tensor([0.1190, 0.3034, 0.2351, 0.0530, 0.3634, 0.2006, 0.2869, 0.2980], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0431, 0.0356, 0.0319, 0.0426, 0.0498, 0.0402, 0.0502], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:58:55,242 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:59:03,233 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9017, 4.1611, 3.9749, 4.0066, 3.6894, 3.7935, 3.8185, 4.1306], device='cuda:3'), covar=tensor([0.1036, 0.0891, 0.0977, 0.0796, 0.0756, 0.1560, 0.0958, 0.0976], device='cuda:3'), in_proj_covar=tensor([0.0643, 0.0794, 0.0653, 0.0589, 0.0499, 0.0504, 0.0657, 0.0611], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:59:26,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7219, 1.3420, 1.7290, 1.6314, 1.7454, 1.8896, 1.5886, 1.8224], device='cuda:3'), covar=tensor([0.0232, 0.0319, 0.0186, 0.0236, 0.0232, 0.0140, 0.0350, 0.0120], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0181, 0.0190, 0.0149, 0.0192, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:59:35,425 INFO [train.py:904] (3/8) Epoch 19, batch 5550, loss[loss=0.202, simple_loss=0.2806, pruned_loss=0.06173, over 17011.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2852, pruned_loss=0.05556, over 3146050.47 frames. ], batch size: 55, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:59:52,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8632, 2.8291, 2.1949, 2.6596, 3.3085, 2.8764, 3.4634, 3.4510], device='cuda:3'), covar=tensor([0.0088, 0.0405, 0.0523, 0.0410, 0.0203, 0.0375, 0.0217, 0.0212], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0228, 0.0220, 0.0220, 0.0229, 0.0227, 0.0230, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 21:59:53,551 INFO [optim.py:368] (3/8) 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 21:59:55,775 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5609, 1.6393, 2.2050, 2.3908, 2.4060, 2.7775, 1.7894, 2.7653], device='cuda:3'), covar=tensor([0.0190, 0.0507, 0.0305, 0.0309, 0.0291, 0.0175, 0.0531, 0.0136], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0189, 0.0177, 0.0180, 0.0190, 0.0149, 0.0192, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:00:26,052 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:00:55,100 INFO [train.py:904] (3/8) Epoch 19, batch 5600, loss[loss=0.3291, simple_loss=0.3765, pruned_loss=0.1408, over 11431.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.291, pruned_loss=0.06089, over 3092281.08 frames. ], batch size: 248, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:01:31,845 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5459, 2.0582, 1.3943, 1.8230, 2.4313, 2.1533, 2.4932, 2.6446], device='cuda:3'), covar=tensor([0.0201, 0.0548, 0.0748, 0.0560, 0.0294, 0.0442, 0.0263, 0.0320], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0227, 0.0219, 0.0219, 0.0227, 0.0226, 0.0228, 0.0222], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:02:07,259 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:02:19,467 INFO [train.py:904] (3/8) Epoch 19, batch 5650, loss[loss=0.2057, simple_loss=0.2879, pruned_loss=0.06172, over 16289.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2951, pruned_loss=0.06341, over 3092196.79 frames. ], batch size: 35, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:20,149 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3177, 3.6871, 3.6699, 2.2260, 3.0759, 2.6035, 3.6983, 3.9478], device='cuda:3'), covar=tensor([0.0249, 0.0665, 0.0567, 0.1832, 0.0806, 0.0851, 0.0578, 0.0859], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0160, 0.0166, 0.0151, 0.0143, 0.0127, 0.0142, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:02:36,994 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.386e+02 3.211e+02 3.758e+02 5.159e+02 1.051e+03, threshold=7.516e+02, percent-clipped=3.0 2023-04-30 22:03:36,659 INFO [train.py:904] (3/8) Epoch 19, batch 5700, loss[loss=0.2182, simple_loss=0.3067, pruned_loss=0.0648, over 15280.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.296, pruned_loss=0.06451, over 3092489.26 frames. ], batch size: 190, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:04:55,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 22:04:55,839 INFO [train.py:904] (3/8) Epoch 19, batch 5750, loss[loss=0.1977, simple_loss=0.2852, pruned_loss=0.05508, over 16663.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2991, pruned_loss=0.06627, over 3060634.55 frames. ], batch size: 134, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:05:12,679 INFO [optim.py:368] (3/8) 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 22:05:45,175 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:06:05,653 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6501, 2.4969, 2.3303, 3.6776, 2.4868, 3.8314, 1.4558, 2.7898], device='cuda:3'), covar=tensor([0.1447, 0.0822, 0.1317, 0.0179, 0.0183, 0.0391, 0.1806, 0.0832], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0192, 0.0182, 0.0203, 0.0212, 0.0196, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:06:19,319 INFO [train.py:904] (3/8) Epoch 19, batch 5800, loss[loss=0.1693, simple_loss=0.2667, pruned_loss=0.03596, over 16858.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2978, pruned_loss=0.06412, over 3073728.29 frames. ], batch size: 96, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:24,028 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:07:38,949 INFO [train.py:904] (3/8) Epoch 19, batch 5850, loss[loss=0.2259, simple_loss=0.3084, pruned_loss=0.07176, over 16196.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2958, pruned_loss=0.06302, over 3063055.50 frames. ], batch size: 165, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:57,309 INFO [optim.py:368] (3/8) 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 22:08:38,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0340, 3.0327, 1.7631, 3.2548, 2.2981, 3.3006, 2.0013, 2.4886], device='cuda:3'), covar=tensor([0.0300, 0.0396, 0.1795, 0.0233, 0.0875, 0.0544, 0.1570, 0.0846], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0194, 0.0156, 0.0175, 0.0215, 0.0201, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:08:57,059 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9320, 4.1933, 4.0145, 4.0542, 3.7571, 3.8225, 3.8730, 4.1765], device='cuda:3'), covar=tensor([0.1136, 0.0870, 0.1006, 0.0824, 0.0809, 0.1615, 0.0886, 0.1034], device='cuda:3'), in_proj_covar=tensor([0.0630, 0.0775, 0.0637, 0.0576, 0.0485, 0.0493, 0.0639, 0.0594], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:08:59,601 INFO [train.py:904] (3/8) Epoch 19, batch 5900, loss[loss=0.1849, simple_loss=0.2758, pruned_loss=0.04695, over 16855.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2957, pruned_loss=0.06304, over 3063554.15 frames. ], batch size: 116, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:09:02,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4212, 2.5608, 2.1272, 2.3086, 3.0236, 2.6071, 3.0452, 3.1509], device='cuda:3'), covar=tensor([0.0134, 0.0474, 0.0558, 0.0483, 0.0253, 0.0391, 0.0240, 0.0254], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0226, 0.0219, 0.0219, 0.0227, 0.0226, 0.0228, 0.0221], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:09:33,971 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-04-30 22:10:01,854 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:10:21,849 INFO [train.py:904] (3/8) Epoch 19, batch 5950, loss[loss=0.2178, simple_loss=0.3133, pruned_loss=0.06118, over 16878.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2966, pruned_loss=0.06205, over 3072711.49 frames. ], batch size: 90, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:40,535 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 19, batch 6000, loss[loss=0.1827, simple_loss=0.2753, pruned_loss=0.04504, over 16809.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2955, pruned_loss=0.06154, over 3071414.70 frames. ], batch size: 102, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:11:41,786 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 22:11:52,559 INFO [train.py:938] (3/8) Epoch 19, validation: loss=0.1523, simple_loss=0.2653, pruned_loss=0.01968, over 944034.00 frames. 2023-04-30 22:11:52,559 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 22:13:13,860 INFO [train.py:904] (3/8) Epoch 19, batch 6050, loss[loss=0.1853, simple_loss=0.2881, pruned_loss=0.04124, over 16501.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.294, pruned_loss=0.0613, over 3066278.98 frames. ], batch size: 75, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:13:33,103 INFO [optim.py:368] (3/8) 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 22:13:34,770 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4992, 3.5461, 3.3189, 3.0171, 3.1486, 3.4446, 3.2969, 3.2726], device='cuda:3'), covar=tensor([0.0593, 0.0583, 0.0278, 0.0258, 0.0517, 0.0429, 0.1360, 0.0517], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0406, 0.0333, 0.0323, 0.0344, 0.0377, 0.0229, 0.0397], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:13:39,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1784, 2.1440, 2.6299, 2.9698, 2.8363, 3.4394, 2.3357, 3.4810], device='cuda:3'), covar=tensor([0.0160, 0.0436, 0.0283, 0.0248, 0.0270, 0.0146, 0.0415, 0.0120], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0186, 0.0174, 0.0176, 0.0187, 0.0146, 0.0188, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:14:32,176 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8954, 4.0033, 3.0672, 2.4515, 2.8962, 2.6328, 4.3712, 3.7942], device='cuda:3'), covar=tensor([0.2504, 0.0688, 0.1631, 0.2340, 0.2295, 0.1766, 0.0444, 0.1039], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0265, 0.0298, 0.0304, 0.0292, 0.0249, 0.0288, 0.0327], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 22:14:32,938 INFO [train.py:904] (3/8) Epoch 19, batch 6100, loss[loss=0.1933, simple_loss=0.2758, pruned_loss=0.05536, over 16707.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2937, pruned_loss=0.06033, over 3083755.11 frames. ], batch size: 57, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:15:14,180 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 22:15:33,790 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:15:57,515 INFO [train.py:904] (3/8) Epoch 19, batch 6150, loss[loss=0.1936, simple_loss=0.2835, pruned_loss=0.05184, over 16839.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2924, pruned_loss=0.0602, over 3071815.76 frames. ], batch size: 116, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:16:16,882 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.888e+02 3.402e+02 4.182e+02 1.052e+03, threshold=6.804e+02, percent-clipped=3.0 2023-04-30 22:17:16,331 INFO [train.py:904] (3/8) Epoch 19, batch 6200, loss[loss=0.223, simple_loss=0.3048, pruned_loss=0.07058, over 15293.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2906, pruned_loss=0.05944, over 3091022.33 frames. ], batch size: 190, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:17:19,330 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7092, 1.9254, 2.3000, 2.5543, 2.6127, 2.9709, 1.9435, 2.8884], device='cuda:3'), covar=tensor([0.0186, 0.0451, 0.0300, 0.0274, 0.0269, 0.0154, 0.0475, 0.0131], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0187, 0.0175, 0.0177, 0.0189, 0.0148, 0.0190, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:17:58,821 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:01,770 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188931.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:15,035 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:17,287 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0815, 3.9662, 4.1273, 4.2563, 4.3685, 3.9650, 4.3081, 4.4009], device='cuda:3'), covar=tensor([0.1682, 0.1174, 0.1393, 0.0712, 0.0619, 0.1302, 0.0849, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0610, 0.0749, 0.0880, 0.0769, 0.0575, 0.0603, 0.0618, 0.0721], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:18:34,122 INFO [train.py:904] (3/8) Epoch 19, batch 6250, loss[loss=0.1893, simple_loss=0.2755, pruned_loss=0.05159, over 16371.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05941, over 3092064.17 frames. ], batch size: 146, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:18:52,888 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 2.827e+02 3.410e+02 4.287e+02 1.283e+03, threshold=6.821e+02, percent-clipped=4.0 2023-04-30 22:19:28,070 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:32,247 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:36,057 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:41,652 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9366, 4.9153, 4.7472, 4.0860, 4.8267, 1.9444, 4.5879, 4.4820], device='cuda:3'), covar=tensor([0.0082, 0.0072, 0.0175, 0.0340, 0.0087, 0.2479, 0.0132, 0.0202], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0146, 0.0192, 0.0175, 0.0167, 0.0201, 0.0181, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:19:51,020 INFO [train.py:904] (3/8) Epoch 19, batch 6300, loss[loss=0.2292, simple_loss=0.298, pruned_loss=0.08014, over 11995.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2903, pruned_loss=0.05904, over 3088086.23 frames. ], batch size: 247, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:20:35,803 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189030.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:09,037 INFO [train.py:904] (3/8) Epoch 19, batch 6350, loss[loss=0.197, simple_loss=0.2781, pruned_loss=0.05795, over 16636.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.291, pruned_loss=0.05991, over 3086596.92 frames. ], batch size: 57, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:21:27,081 INFO [optim.py:368] (3/8) 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 22:21:28,304 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:48,816 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6812, 4.6606, 4.5381, 3.8045, 4.5975, 1.8741, 4.3792, 4.2707], device='cuda:3'), covar=tensor([0.0089, 0.0074, 0.0168, 0.0338, 0.0080, 0.2576, 0.0123, 0.0195], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0146, 0.0192, 0.0175, 0.0167, 0.0201, 0.0182, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:21:49,115 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 22:21:51,862 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:10,409 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:26,882 INFO [train.py:904] (3/8) Epoch 19, batch 6400, loss[loss=0.1984, simple_loss=0.282, pruned_loss=0.05745, over 15354.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2911, pruned_loss=0.06034, over 3104604.21 frames. ], batch size: 190, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:22:44,218 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0722, 3.5053, 3.3030, 1.9049, 2.8316, 2.3569, 3.4740, 3.7110], device='cuda:3'), covar=tensor([0.0305, 0.0716, 0.0743, 0.2270, 0.0950, 0.1014, 0.0692, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0144, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:23:02,697 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:21,427 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:25,614 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:43,283 INFO [train.py:904] (3/8) Epoch 19, batch 6450, loss[loss=0.1891, simple_loss=0.2807, pruned_loss=0.04871, over 15349.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2912, pruned_loss=0.05977, over 3097300.34 frames. ], batch size: 191, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:24:01,002 INFO [optim.py:368] (3/8) 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 22:24:27,278 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 22:24:34,446 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:25:00,510 INFO [train.py:904] (3/8) Epoch 19, batch 6500, loss[loss=0.1944, simple_loss=0.2742, pruned_loss=0.05728, over 17244.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2891, pruned_loss=0.05884, over 3110783.77 frames. ], batch size: 45, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:25:33,398 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2220, 2.0836, 2.1275, 3.7786, 2.0648, 2.4868, 2.2296, 2.2235], device='cuda:3'), covar=tensor([0.1309, 0.3764, 0.2914, 0.0602, 0.4375, 0.2593, 0.3548, 0.3529], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0432, 0.0355, 0.0318, 0.0429, 0.0497, 0.0403, 0.0503], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:26:10,385 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 22:26:14,540 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:26:19,680 INFO [train.py:904] (3/8) Epoch 19, batch 6550, loss[loss=0.2544, simple_loss=0.3219, pruned_loss=0.09346, over 11737.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2916, pruned_loss=0.05939, over 3124317.86 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:37,017 INFO [optim.py:368] (3/8) 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 22:27:09,603 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:12,559 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 6600, loss[loss=0.2074, simple_loss=0.2868, pruned_loss=0.06402, over 16412.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2942, pruned_loss=0.0601, over 3135941.04 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:27:45,670 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:44,279 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:45,790 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 22:28:50,951 INFO [train.py:904] (3/8) Epoch 19, batch 6650, loss[loss=0.2278, simple_loss=0.3054, pruned_loss=0.07511, over 16875.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2949, pruned_loss=0.06169, over 3101044.09 frames. ], batch size: 116, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:29:08,220 INFO [optim.py:368] (3/8) 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 22:29:42,442 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189386.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:29:53,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5952, 2.7767, 2.4600, 4.2355, 3.0865, 3.9711, 1.5718, 2.8064], device='cuda:3'), covar=tensor([0.1401, 0.0720, 0.1276, 0.0153, 0.0259, 0.0404, 0.1660, 0.0866], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0172, 0.0194, 0.0183, 0.0205, 0.0214, 0.0197, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:30:06,165 INFO [train.py:904] (3/8) Epoch 19, batch 6700, loss[loss=0.1941, simple_loss=0.2734, pruned_loss=0.05741, over 16151.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2942, pruned_loss=0.06234, over 3085501.45 frames. ], batch size: 35, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:30:15,385 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:33,796 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:56,601 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:31:21,910 INFO [train.py:904] (3/8) Epoch 19, batch 6750, loss[loss=0.2249, simple_loss=0.3016, pruned_loss=0.07408, over 16225.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2933, pruned_loss=0.06255, over 3077052.08 frames. ], batch size: 165, lr: 3.55e-03, grad_scale: 4.0 2023-04-30 22:31:42,235 INFO [optim.py:368] (3/8) 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 22:31:53,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1035, 4.0420, 4.2174, 4.3159, 4.4464, 4.0733, 4.3887, 4.4764], device='cuda:3'), covar=tensor([0.1770, 0.1186, 0.1334, 0.0671, 0.0535, 0.1281, 0.0757, 0.0621], device='cuda:3'), in_proj_covar=tensor([0.0609, 0.0750, 0.0880, 0.0768, 0.0577, 0.0604, 0.0621, 0.0720], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:32:17,161 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9469, 2.7704, 2.8280, 2.1075, 2.6661, 2.1956, 2.7815, 3.0118], device='cuda:3'), covar=tensor([0.0240, 0.0697, 0.0518, 0.1663, 0.0784, 0.0880, 0.0508, 0.0638], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0161, 0.0166, 0.0152, 0.0144, 0.0129, 0.0143, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:32:38,316 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:32:39,012 INFO [train.py:904] (3/8) Epoch 19, batch 6800, loss[loss=0.2393, simple_loss=0.302, pruned_loss=0.08832, over 11262.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2933, pruned_loss=0.06248, over 3084885.19 frames. ], batch size: 249, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:33:58,199 INFO [train.py:904] (3/8) Epoch 19, batch 6850, loss[loss=0.2121, simple_loss=0.3049, pruned_loss=0.05971, over 16230.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2942, pruned_loss=0.0628, over 3077219.64 frames. ], batch size: 165, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:34:13,618 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189562.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:17,024 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.732e+02 3.346e+02 4.467e+02 6.913e+02, threshold=6.693e+02, percent-clipped=4.0 2023-04-30 22:34:47,132 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:50,430 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:12,995 INFO [train.py:904] (3/8) Epoch 19, batch 6900, loss[loss=0.1928, simple_loss=0.2819, pruned_loss=0.05188, over 16970.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.296, pruned_loss=0.06171, over 3092329.28 frames. ], batch size: 109, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:35:17,326 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:00,958 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:04,229 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189635.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:12,212 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189640.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:29,235 INFO [train.py:904] (3/8) Epoch 19, batch 6950, loss[loss=0.2039, simple_loss=0.2938, pruned_loss=0.05697, over 16493.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2982, pruned_loss=0.0643, over 3064490.68 frames. ], batch size: 75, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:36:48,878 INFO [optim.py:368] (3/8) 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 22:37:21,359 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:36,981 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-30 22:37:43,822 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:44,541 INFO [train.py:904] (3/8) Epoch 19, batch 7000, loss[loss=0.1939, simple_loss=0.2871, pruned_loss=0.05035, over 16694.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2979, pruned_loss=0.06386, over 3045928.99 frames. ], batch size: 124, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:37:47,738 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189703.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:49,134 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:13,518 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:25,524 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:34,806 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:36,103 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:57,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5176, 5.5577, 5.3800, 5.0114, 4.9743, 5.4286, 5.3454, 5.0975], device='cuda:3'), covar=tensor([0.0587, 0.0475, 0.0266, 0.0268, 0.0965, 0.0425, 0.0250, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0399, 0.0325, 0.0317, 0.0336, 0.0369, 0.0225, 0.0390], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:39:01,643 INFO [train.py:904] (3/8) Epoch 19, batch 7050, loss[loss=0.2657, simple_loss=0.3236, pruned_loss=0.1039, over 11729.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2984, pruned_loss=0.06251, over 3084108.16 frames. ], batch size: 247, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:39:22,081 INFO [optim.py:368] (3/8) 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 22:39:22,598 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:24,631 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0831, 3.6924, 3.6232, 2.2983, 3.3741, 3.6737, 3.3983, 2.0214], device='cuda:3'), covar=tensor([0.0564, 0.0052, 0.0055, 0.0437, 0.0097, 0.0120, 0.0092, 0.0471], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0079, 0.0079, 0.0133, 0.0094, 0.0106, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 22:39:27,396 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189768.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:50,862 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:00,704 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:18,610 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 22:40:19,979 INFO [train.py:904] (3/8) Epoch 19, batch 7100, loss[loss=0.205, simple_loss=0.2967, pruned_loss=0.0566, over 16702.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2966, pruned_loss=0.06247, over 3063158.68 frames. ], batch size: 134, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:41:08,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6367, 2.5800, 1.9056, 2.6655, 2.1351, 2.7678, 2.1720, 2.3451], device='cuda:3'), covar=tensor([0.0322, 0.0383, 0.1162, 0.0222, 0.0573, 0.0505, 0.1090, 0.0555], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0174, 0.0192, 0.0154, 0.0173, 0.0212, 0.0199, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:41:32,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7595, 3.9258, 3.0729, 2.3780, 2.8425, 2.5611, 4.2769, 3.5908], device='cuda:3'), covar=tensor([0.2694, 0.0652, 0.1682, 0.2369, 0.2344, 0.1886, 0.0426, 0.1142], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0264, 0.0297, 0.0304, 0.0292, 0.0250, 0.0287, 0.0326], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 22:41:37,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6515, 2.4250, 2.0779, 3.4621, 2.1494, 3.6147, 1.4393, 2.5192], device='cuda:3'), covar=tensor([0.1502, 0.0889, 0.1577, 0.0222, 0.0244, 0.0466, 0.1822, 0.1048], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0185, 0.0208, 0.0216, 0.0199, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:41:38,323 INFO [train.py:904] (3/8) Epoch 19, batch 7150, loss[loss=0.222, simple_loss=0.3001, pruned_loss=0.07193, over 16202.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2941, pruned_loss=0.06159, over 3082044.29 frames. ], batch size: 165, lr: 3.54e-03, grad_scale: 4.0 2023-04-30 22:41:45,507 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:42:35,114 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0177, 2.4582, 2.6411, 1.8838, 2.7306, 2.8433, 2.4071, 2.3930], device='cuda:3'), covar=tensor([0.0745, 0.0273, 0.0230, 0.1030, 0.0108, 0.0244, 0.0453, 0.0480], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0107, 0.0097, 0.0139, 0.0078, 0.0122, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:42:51,645 INFO [train.py:904] (3/8) Epoch 19, batch 7200, loss[loss=0.1727, simple_loss=0.2632, pruned_loss=0.04106, over 16621.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2916, pruned_loss=0.05982, over 3075009.55 frames. ], batch size: 57, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:42:55,678 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189904.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:43:10,469 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8151, 1.3908, 1.6592, 1.7439, 1.8209, 1.9531, 1.5886, 1.7814], device='cuda:3'), covar=tensor([0.0204, 0.0350, 0.0202, 0.0257, 0.0249, 0.0156, 0.0383, 0.0115], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0189, 0.0174, 0.0178, 0.0190, 0.0148, 0.0191, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:43:30,253 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-30 22:44:00,400 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-04-30 22:44:12,502 INFO [train.py:904] (3/8) Epoch 19, batch 7250, loss[loss=0.1725, simple_loss=0.2627, pruned_loss=0.04113, over 16877.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2896, pruned_loss=0.05839, over 3074834.89 frames. ], batch size: 102, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:44:12,830 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:44:42,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4502, 4.6799, 4.8300, 4.6266, 4.7234, 5.2577, 4.7577, 4.4863], device='cuda:3'), covar=tensor([0.1464, 0.2062, 0.2341, 0.2117, 0.2500, 0.1045, 0.1837, 0.2897], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0574, 0.0630, 0.0480, 0.0638, 0.0662, 0.0495, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 22:45:19,909 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7300, loss[loss=0.2381, simple_loss=0.3026, pruned_loss=0.08678, over 11314.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2896, pruned_loss=0.05832, over 3086512.74 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:45:33,920 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190003.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:24,019 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4418, 4.4904, 4.2850, 4.0288, 4.0086, 4.4114, 4.1153, 4.1254], device='cuda:3'), covar=tensor([0.0506, 0.0392, 0.0264, 0.0250, 0.0742, 0.0362, 0.0658, 0.0578], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0399, 0.0324, 0.0317, 0.0336, 0.0369, 0.0225, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:46:29,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5310, 3.5315, 2.2477, 4.1040, 2.7464, 4.1210, 2.3258, 2.8049], device='cuda:3'), covar=tensor([0.0266, 0.0394, 0.1631, 0.0128, 0.0788, 0.0378, 0.1577, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0174, 0.0192, 0.0154, 0.0173, 0.0212, 0.0200, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:46:36,515 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1644, 1.5293, 1.9643, 2.1053, 2.1796, 2.4265, 1.6908, 2.2282], device='cuda:3'), covar=tensor([0.0216, 0.0465, 0.0266, 0.0306, 0.0290, 0.0175, 0.0455, 0.0137], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0176, 0.0188, 0.0146, 0.0189, 0.0140], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:46:48,531 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7350, loss[loss=0.2458, simple_loss=0.3111, pruned_loss=0.09024, over 10896.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2908, pruned_loss=0.05968, over 3065714.15 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:47:01,677 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:47:10,704 INFO [optim.py:368] (3/8) 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 22:47:40,424 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7400, loss[loss=0.251, simple_loss=0.3161, pruned_loss=0.093, over 11212.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2914, pruned_loss=0.06016, over 3062381.49 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:48:35,180 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0782, 2.3358, 2.3376, 2.7054, 2.0149, 3.1196, 1.8587, 2.7111], device='cuda:3'), covar=tensor([0.1059, 0.0596, 0.1009, 0.0188, 0.0165, 0.0400, 0.1291, 0.0660], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0184, 0.0207, 0.0216, 0.0199, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 22:49:09,858 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190139.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:49:30,149 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7450, loss[loss=0.2252, simple_loss=0.313, pruned_loss=0.06875, over 16496.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2925, pruned_loss=0.0611, over 3063901.97 frames. ], batch size: 68, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:41,095 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:50:24,142 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0906, 2.1056, 2.1737, 3.7812, 2.0778, 2.4882, 2.1851, 2.2534], device='cuda:3'), covar=tensor([0.1340, 0.3556, 0.2939, 0.0560, 0.4172, 0.2591, 0.3566, 0.3481], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0433, 0.0356, 0.0319, 0.0432, 0.0498, 0.0405, 0.0505], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:50:51,747 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190200.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:50:54,217 INFO [train.py:904] (3/8) Epoch 19, batch 7500, loss[loss=0.2305, simple_loss=0.3016, pruned_loss=0.07972, over 11750.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.293, pruned_loss=0.06064, over 3045754.71 frames. ], batch size: 250, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:50:59,305 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:51:10,879 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:51:16,881 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2450, 4.1302, 4.3075, 4.4526, 4.5921, 4.1877, 4.5270, 4.6093], device='cuda:3'), covar=tensor([0.1989, 0.1283, 0.1631, 0.0772, 0.0547, 0.1192, 0.0739, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0602, 0.0741, 0.0873, 0.0761, 0.0571, 0.0598, 0.0618, 0.0712], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:52:06,453 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 22:52:13,611 INFO [train.py:904] (3/8) Epoch 19, batch 7550, loss[loss=0.1809, simple_loss=0.277, pruned_loss=0.04244, over 16777.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2922, pruned_loss=0.06081, over 3041869.34 frames. ], batch size: 89, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:52:34,941 INFO [optim.py:368] (3/8) 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 22:53:22,544 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7600, loss[loss=0.2114, simple_loss=0.2984, pruned_loss=0.06218, over 16415.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2913, pruned_loss=0.06117, over 3028476.87 frames. ], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:53:48,428 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 22:53:53,570 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190316.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:36,982 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7650, loss[loss=0.2195, simple_loss=0.3035, pruned_loss=0.0678, over 16402.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2916, pruned_loss=0.06134, over 3032878.70 frames. ], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:55:02,235 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:10,291 INFO [optim.py:368] (3/8) 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 22:55:27,527 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190377.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:38,671 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:02,461 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7700, loss[loss=0.2448, simple_loss=0.3216, pruned_loss=0.08404, over 11762.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2913, pruned_loss=0.06154, over 3039757.76 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:56:15,128 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:26,112 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 22:56:50,414 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:01,410 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0288, 5.3328, 5.1132, 5.1006, 4.8296, 4.8122, 4.7831, 5.4232], device='cuda:3'), covar=tensor([0.1226, 0.0828, 0.1001, 0.0904, 0.0887, 0.0876, 0.1182, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0637, 0.0772, 0.0642, 0.0579, 0.0485, 0.0498, 0.0647, 0.0595], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:57:18,818 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2042, 4.2086, 4.0852, 3.2792, 4.1601, 1.7095, 3.9186, 3.6416], device='cuda:3'), covar=tensor([0.0126, 0.0099, 0.0186, 0.0336, 0.0098, 0.2777, 0.0150, 0.0260], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0143, 0.0189, 0.0172, 0.0164, 0.0198, 0.0178, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 22:57:21,407 INFO [train.py:904] (3/8) Epoch 19, batch 7750, loss[loss=0.2028, simple_loss=0.2959, pruned_loss=0.05485, over 16550.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2914, pruned_loss=0.06066, over 3074171.60 frames. ], batch size: 75, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:57:35,047 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.766e+02 2.725e+02 3.462e+02 4.075e+02 8.569e+02, threshold=6.924e+02, percent-clipped=2.0 2023-04-30 22:58:30,599 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190495.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 22:58:39,265 INFO [train.py:904] (3/8) Epoch 19, batch 7800, loss[loss=0.1884, simple_loss=0.2822, pruned_loss=0.04733, over 17255.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.292, pruned_loss=0.06089, over 3084330.69 frames. ], batch size: 52, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:58:47,776 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7850, loss[loss=0.2016, simple_loss=0.2858, pruned_loss=0.05867, over 16331.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2929, pruned_loss=0.06143, over 3058321.85 frames. ], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:00:17,836 INFO [optim.py:368] (3/8) 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 23:00:56,960 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 7900, loss[loss=0.2493, simple_loss=0.3233, pruned_loss=0.08763, over 11580.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.292, pruned_loss=0.06071, over 3064178.77 frames. ], batch size: 247, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:01:23,306 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190609.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:02:32,482 INFO [train.py:904] (3/8) Epoch 19, batch 7950, loss[loss=0.2236, simple_loss=0.3016, pruned_loss=0.07282, over 15308.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.293, pruned_loss=0.06155, over 3060861.46 frames. ], batch size: 191, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:02:32,957 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190652.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:02:54,096 INFO [optim.py:368] (3/8) 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 23:03:00,812 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190670.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:03:04,352 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190672.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:08,603 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 8000, loss[loss=0.2036, simple_loss=0.292, pruned_loss=0.05762, over 16672.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2939, pruned_loss=0.06283, over 3048733.65 frames. ], batch size: 134, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:04:28,374 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 23:04:44,501 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190736.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:07,648 INFO [train.py:904] (3/8) Epoch 19, batch 8050, loss[loss=0.1925, simple_loss=0.2841, pruned_loss=0.0504, over 16812.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2935, pruned_loss=0.06228, over 3048484.44 frames. ], batch size: 102, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:05:12,157 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:06:12,403 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190795.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:06:22,314 INFO [train.py:904] (3/8) Epoch 19, batch 8100, loss[loss=0.1835, simple_loss=0.2713, pruned_loss=0.04783, over 17239.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2921, pruned_loss=0.06089, over 3074191.54 frames. ], batch size: 52, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:06:23,082 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-30 23:06:31,316 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:07:25,607 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190843.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:07:39,349 INFO [train.py:904] (3/8) Epoch 19, batch 8150, loss[loss=0.1985, simple_loss=0.2855, pruned_loss=0.05574, over 16707.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2904, pruned_loss=0.06056, over 3077700.65 frames. ], batch size: 124, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:07:44,592 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:08:01,318 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.839e+02 2.760e+02 3.332e+02 4.142e+02 6.995e+02, threshold=6.664e+02, percent-clipped=2.0 2023-04-30 23:08:40,557 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 23:08:57,395 INFO [train.py:904] (3/8) Epoch 19, batch 8200, loss[loss=0.1945, simple_loss=0.2747, pruned_loss=0.05713, over 16651.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2874, pruned_loss=0.05914, over 3094415.48 frames. ], batch size: 57, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:11,508 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190947.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:10:19,456 INFO [train.py:904] (3/8) Epoch 19, batch 8250, loss[loss=0.1939, simple_loss=0.2882, pruned_loss=0.0498, over 16599.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2863, pruned_loss=0.05692, over 3072048.07 frames. ], batch size: 76, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:30,790 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6098, 4.4227, 4.6220, 4.7923, 4.9805, 4.4228, 4.9838, 4.9860], device='cuda:3'), covar=tensor([0.1944, 0.1365, 0.1851, 0.0864, 0.0608, 0.1052, 0.0522, 0.0671], device='cuda:3'), in_proj_covar=tensor([0.0599, 0.0744, 0.0869, 0.0760, 0.0575, 0.0597, 0.0619, 0.0715], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:10:39,904 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190965.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:10:40,163 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1687, 2.1065, 2.1361, 3.8755, 2.0777, 2.4747, 2.2119, 2.2699], device='cuda:3'), covar=tensor([0.1210, 0.3709, 0.3108, 0.0468, 0.4418, 0.2547, 0.3727, 0.3450], device='cuda:3'), in_proj_covar=tensor([0.0385, 0.0429, 0.0352, 0.0316, 0.0427, 0.0492, 0.0400, 0.0500], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:10:41,154 INFO [optim.py:368] (3/8) 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 23:10:48,369 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 23:10:51,838 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:11:29,772 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6446, 3.8943, 3.9967, 2.8060, 3.5316, 3.9359, 3.6509, 2.5758], device='cuda:3'), covar=tensor([0.0451, 0.0060, 0.0043, 0.0336, 0.0101, 0.0096, 0.0079, 0.0395], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0079, 0.0080, 0.0133, 0.0093, 0.0107, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 23:11:39,003 INFO [train.py:904] (3/8) Epoch 19, batch 8300, loss[loss=0.1669, simple_loss=0.2551, pruned_loss=0.03934, over 12197.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2838, pruned_loss=0.05387, over 3080684.85 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:12:10,266 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191031.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:40,323 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-04-30 23:13:00,823 INFO [train.py:904] (3/8) Epoch 19, batch 8350, loss[loss=0.2041, simple_loss=0.2796, pruned_loss=0.06424, over 11813.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2829, pruned_loss=0.05205, over 3067243.76 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:13:08,061 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191055.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:24,341 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191068.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:52,505 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2932, 4.2839, 4.6674, 4.6401, 4.6438, 4.4031, 4.3462, 4.3747], device='cuda:3'), covar=tensor([0.0388, 0.0769, 0.0453, 0.0477, 0.0510, 0.0440, 0.1100, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0431, 0.0418, 0.0394, 0.0465, 0.0439, 0.0535, 0.0355], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-04-30 23:14:22,694 INFO [train.py:904] (3/8) Epoch 19, batch 8400, loss[loss=0.1745, simple_loss=0.2636, pruned_loss=0.04275, over 16329.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2804, pruned_loss=0.04998, over 3065707.61 frames. ], batch size: 146, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:14:24,464 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:06,903 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:11,019 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 8450, loss[loss=0.1745, simple_loss=0.2643, pruned_loss=0.04237, over 12330.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2788, pruned_loss=0.04872, over 3056346.06 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:16:06,334 INFO [optim.py:368] (3/8) 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 23:16:37,003 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:16:50,105 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:17:04,704 INFO [train.py:904] (3/8) Epoch 19, batch 8500, loss[loss=0.1557, simple_loss=0.2369, pruned_loss=0.03728, over 11913.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2748, pruned_loss=0.04639, over 3041112.29 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:22,837 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191247.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:18:32,481 INFO [train.py:904] (3/8) Epoch 19, batch 8550, loss[loss=0.1819, simple_loss=0.2826, pruned_loss=0.0406, over 16792.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2724, pruned_loss=0.04501, over 3040041.70 frames. ], batch size: 83, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:57,579 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191265.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:19:00,618 INFO [optim.py:368] (3/8) 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 23:20:00,162 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191295.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:20:05,708 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9212, 2.2080, 2.2521, 2.9625, 1.7914, 3.2621, 1.7000, 2.8038], device='cuda:3'), covar=tensor([0.1159, 0.0743, 0.1037, 0.0156, 0.0078, 0.0340, 0.1395, 0.0637], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0179, 0.0201, 0.0209, 0.0193, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 23:20:13,437 INFO [train.py:904] (3/8) Epoch 19, batch 8600, loss[loss=0.1568, simple_loss=0.2528, pruned_loss=0.0304, over 16469.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2725, pruned_loss=0.0439, over 3045349.80 frames. ], batch size: 68, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:20:37,122 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:03,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6370, 3.6814, 2.8431, 2.1596, 2.3050, 2.3191, 3.9217, 3.2309], device='cuda:3'), covar=tensor([0.2785, 0.0628, 0.1764, 0.2999, 0.2809, 0.2178, 0.0401, 0.1383], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0259, 0.0294, 0.0299, 0.0286, 0.0246, 0.0282, 0.0321], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 23:21:11,769 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:52,928 INFO [train.py:904] (3/8) Epoch 19, batch 8650, loss[loss=0.1739, simple_loss=0.2774, pruned_loss=0.03518, over 16510.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.271, pruned_loss=0.04266, over 3042579.87 frames. ], batch size: 147, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:22:25,685 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:22:55,702 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:23:39,021 INFO [train.py:904] (3/8) Epoch 19, batch 8700, loss[loss=0.1597, simple_loss=0.264, pruned_loss=0.02767, over 16532.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2682, pruned_loss=0.04145, over 3031554.57 frames. ], batch size: 68, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:24:20,674 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:23,631 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:25:14,446 INFO [train.py:904] (3/8) Epoch 19, batch 8750, loss[loss=0.1818, simple_loss=0.2749, pruned_loss=0.04437, over 15298.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2678, pruned_loss=0.04067, over 3041382.33 frames. ], batch size: 191, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:25:37,771 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6968, 2.4097, 2.2213, 3.3756, 1.8439, 3.5778, 1.4361, 2.8055], device='cuda:3'), covar=tensor([0.1405, 0.0841, 0.1329, 0.0152, 0.0104, 0.0367, 0.1802, 0.0767], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0168, 0.0189, 0.0177, 0.0200, 0.0207, 0.0193, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-04-30 23:25:58,340 INFO [optim.py:368] (3/8) 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 23:26:40,134 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 8800, loss[loss=0.1626, simple_loss=0.2592, pruned_loss=0.03301, over 15387.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2661, pruned_loss=0.03982, over 3037878.80 frames. ], batch size: 191, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:27:34,729 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0542, 2.6144, 2.6191, 1.9241, 2.7470, 2.8715, 2.5209, 2.5358], device='cuda:3'), covar=tensor([0.0672, 0.0242, 0.0245, 0.0966, 0.0099, 0.0225, 0.0418, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0102, 0.0092, 0.0134, 0.0075, 0.0116, 0.0121, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 23:27:52,113 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 23:28:04,445 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 23:28:29,019 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 8850, loss[loss=0.1609, simple_loss=0.2548, pruned_loss=0.03345, over 12494.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2689, pruned_loss=0.03934, over 3025365.63 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:29:00,244 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7794, 4.3704, 4.3755, 3.0508, 3.7510, 4.3973, 3.9681, 2.6125], device='cuda:3'), covar=tensor([0.0451, 0.0031, 0.0031, 0.0339, 0.0087, 0.0069, 0.0063, 0.0395], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0133, 0.0094, 0.0105, 0.0091, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 23:29:28,627 INFO [optim.py:368] (3/8) 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 23:29:49,697 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2171, 4.2880, 4.1090, 3.8078, 3.7914, 4.1902, 3.9022, 3.9196], device='cuda:3'), covar=tensor([0.0644, 0.0722, 0.0397, 0.0347, 0.0818, 0.0614, 0.0820, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0269, 0.0389, 0.0316, 0.0308, 0.0324, 0.0358, 0.0219, 0.0376], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:30:28,189 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5494, 3.6047, 2.6964, 2.1170, 2.3008, 2.2834, 3.7604, 3.2954], device='cuda:3'), covar=tensor([0.2823, 0.0592, 0.1730, 0.2821, 0.2647, 0.2085, 0.0437, 0.1153], device='cuda:3'), in_proj_covar=tensor([0.0316, 0.0256, 0.0291, 0.0295, 0.0282, 0.0244, 0.0279, 0.0317], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-04-30 23:30:39,776 INFO [train.py:904] (3/8) Epoch 19, batch 8900, loss[loss=0.1691, simple_loss=0.2569, pruned_loss=0.04068, over 12643.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2691, pruned_loss=0.03862, over 3038620.56 frames. ], batch size: 250, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:32:45,267 INFO [train.py:904] (3/8) Epoch 19, batch 8950, loss[loss=0.1739, simple_loss=0.2629, pruned_loss=0.04248, over 12763.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2692, pruned_loss=0.03927, over 3039761.92 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:33:21,157 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.162e+02 2.456e+02 2.877e+02 5.499e+02, threshold=4.911e+02, percent-clipped=2.0 2023-04-30 23:34:34,392 INFO [train.py:904] (3/8) Epoch 19, batch 9000, loss[loss=0.1729, simple_loss=0.2634, pruned_loss=0.04117, over 15434.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2666, pruned_loss=0.03817, over 3063526.61 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:34:34,393 INFO [train.py:929] (3/8) Computing validation loss 2023-04-30 23:34:44,207 INFO [train.py:938] (3/8) Epoch 19, validation: loss=0.1462, simple_loss=0.2506, pruned_loss=0.02087, over 944034.00 frames. 2023-04-30 23:34:44,208 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-04-30 23:35:04,329 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 23:35:25,457 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:31,025 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:32,965 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0140, 3.2405, 3.2271, 2.2054, 3.0376, 3.2620, 3.1111, 1.8959], device='cuda:3'), covar=tensor([0.0514, 0.0051, 0.0050, 0.0365, 0.0092, 0.0068, 0.0078, 0.0480], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0077, 0.0078, 0.0131, 0.0092, 0.0103, 0.0089, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-04-30 23:36:20,082 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-30 23:36:27,807 INFO [train.py:904] (3/8) Epoch 19, batch 9050, loss[loss=0.1868, simple_loss=0.2765, pruned_loss=0.04856, over 16182.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2675, pruned_loss=0.03864, over 3073525.60 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:37:04,257 INFO [optim.py:368] (3/8) 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 23:37:09,313 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:37:39,907 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:38:10,475 INFO [train.py:904] (3/8) Epoch 19, batch 9100, loss[loss=0.1816, simple_loss=0.2762, pruned_loss=0.04357, over 16913.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2671, pruned_loss=0.03917, over 3072934.98 frames. ], batch size: 116, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:39:32,422 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191836.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:43,530 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:40:08,016 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 23:40:08,180 INFO [train.py:904] (3/8) Epoch 19, batch 9150, loss[loss=0.1691, simple_loss=0.2535, pruned_loss=0.04237, over 12044.00 frames. ], tot_loss[loss=0.173, simple_loss=0.268, pruned_loss=0.03902, over 3084019.99 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:40:46,342 INFO [optim.py:368] (3/8) 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 23:41:30,261 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:41:52,732 INFO [train.py:904] (3/8) Epoch 19, batch 9200, loss[loss=0.1736, simple_loss=0.2625, pruned_loss=0.04235, over 16226.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2638, pruned_loss=0.0379, over 3081832.21 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:43:29,322 INFO [train.py:904] (3/8) Epoch 19, batch 9250, loss[loss=0.1581, simple_loss=0.2557, pruned_loss=0.03025, over 15237.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2638, pruned_loss=0.03801, over 3073118.36 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:44:05,909 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 19, batch 9300, loss[loss=0.1563, simple_loss=0.2517, pruned_loss=0.03051, over 16763.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2613, pruned_loss=0.0371, over 3062899.60 frames. ], batch size: 134, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:46:09,693 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 9350, loss[loss=0.1602, simple_loss=0.2467, pruned_loss=0.03687, over 12282.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2615, pruned_loss=0.03718, over 3069034.15 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:47:46,336 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:47,132 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.346e+02 1.981e+02 2.500e+02 3.039e+02 5.486e+02, threshold=4.999e+02, percent-clipped=1.0 2023-04-30 23:48:49,308 INFO [train.py:904] (3/8) Epoch 19, batch 9400, loss[loss=0.1821, simple_loss=0.2856, pruned_loss=0.03932, over 16921.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2609, pruned_loss=0.0368, over 3065607.54 frames. ], batch size: 116, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:48:50,410 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:29,958 INFO [train.py:904] (3/8) Epoch 19, batch 9450, loss[loss=0.1547, simple_loss=0.2458, pruned_loss=0.03183, over 12329.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2618, pruned_loss=0.03685, over 3032799.60 frames. ], batch size: 246, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:50:35,895 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:42,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1047, 3.9069, 4.1683, 2.3221, 4.3653, 4.4410, 3.3249, 3.5908], device='cuda:3'), covar=tensor([0.0530, 0.0181, 0.0164, 0.1021, 0.0056, 0.0091, 0.0316, 0.0301], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0101, 0.0089, 0.0131, 0.0073, 0.0113, 0.0119, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-04-30 23:50:51,010 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192163.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:51:05,240 INFO [optim.py:368] (3/8) 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 23:52:10,473 INFO [train.py:904] (3/8) Epoch 19, batch 9500, loss[loss=0.1639, simple_loss=0.244, pruned_loss=0.04189, over 12551.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2613, pruned_loss=0.03694, over 3038043.57 frames. ], batch size: 246, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:52:19,490 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5123, 3.4759, 3.5027, 2.6474, 3.4433, 1.9169, 3.1256, 2.7755], device='cuda:3'), covar=tensor([0.0169, 0.0155, 0.0196, 0.0401, 0.0136, 0.2922, 0.0184, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0139, 0.0181, 0.0163, 0.0159, 0.0194, 0.0171, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:52:21,825 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4185, 3.3517, 3.4962, 3.5691, 3.6035, 3.3124, 3.5560, 3.6554], device='cuda:3'), covar=tensor([0.1340, 0.1026, 0.1065, 0.0628, 0.0603, 0.2335, 0.0944, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0585, 0.0722, 0.0844, 0.0742, 0.0555, 0.0580, 0.0598, 0.0696], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:52:39,652 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:45,145 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:49,324 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:53:50,616 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8232, 3.7393, 3.9210, 4.0089, 4.0764, 3.6403, 4.0338, 4.1069], device='cuda:3'), covar=tensor([0.1448, 0.1002, 0.1164, 0.0653, 0.0490, 0.1950, 0.0679, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0582, 0.0719, 0.0840, 0.0738, 0.0552, 0.0577, 0.0595, 0.0693], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:53:55,147 INFO [train.py:904] (3/8) Epoch 19, batch 9550, loss[loss=0.1828, simple_loss=0.2813, pruned_loss=0.04222, over 16365.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2607, pruned_loss=0.03699, over 3045252.79 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:54:34,515 INFO [optim.py:368] (3/8) 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 23:54:55,530 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:00,674 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:38,440 INFO [train.py:904] (3/8) Epoch 19, batch 9600, loss[loss=0.1876, simple_loss=0.2912, pruned_loss=0.042, over 16676.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2625, pruned_loss=0.03772, over 3039503.12 frames. ], batch size: 134, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:56:59,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7006, 2.4309, 2.3136, 3.5785, 2.1435, 3.6390, 1.4948, 2.8270], device='cuda:3'), covar=tensor([0.1465, 0.0822, 0.1316, 0.0165, 0.0112, 0.0424, 0.1739, 0.0768], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0167, 0.0189, 0.0175, 0.0196, 0.0207, 0.0193, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-04-30 23:57:27,640 INFO [train.py:904] (3/8) Epoch 19, batch 9650, loss[loss=0.1681, simple_loss=0.2626, pruned_loss=0.03678, over 16524.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2638, pruned_loss=0.03786, over 3021023.04 frames. ], batch size: 62, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:58:09,763 INFO [optim.py:368] (3/8) 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 23:58:35,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9588, 5.0263, 4.8531, 4.4051, 4.4623, 4.9454, 4.8454, 4.5566], device='cuda:3'), covar=tensor([0.0772, 0.0742, 0.0604, 0.0396, 0.1204, 0.0645, 0.0425, 0.1049], device='cuda:3'), in_proj_covar=tensor([0.0266, 0.0381, 0.0310, 0.0301, 0.0317, 0.0354, 0.0214, 0.0370], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-04-30 23:58:47,382 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4095, 3.0940, 2.7126, 2.2148, 2.1710, 2.2497, 3.0431, 2.8896], device='cuda:3'), covar=tensor([0.2474, 0.0656, 0.1615, 0.2757, 0.2554, 0.2182, 0.0468, 0.1362], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0255, 0.0289, 0.0294, 0.0277, 0.0243, 0.0277, 0.0315], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-04-30 23:58:54,555 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:09,434 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:15,173 INFO [train.py:904] (3/8) Epoch 19, batch 9700, loss[loss=0.1756, simple_loss=0.2705, pruned_loss=0.04035, over 16943.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2629, pruned_loss=0.03754, over 3038520.51 frames. ], batch size: 109, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:59:34,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9130, 5.2681, 5.4459, 5.2395, 5.3557, 5.8281, 5.3738, 5.0844], device='cuda:3'), covar=tensor([0.0982, 0.1838, 0.2183, 0.2020, 0.2296, 0.0912, 0.1372, 0.2106], device='cuda:3'), in_proj_covar=tensor([0.0377, 0.0544, 0.0603, 0.0452, 0.0605, 0.0631, 0.0471, 0.0604], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:00:19,364 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192433.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:00:23,804 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3201, 6.0190, 6.2090, 5.9363, 6.0165, 6.4724, 6.0940, 5.8286], device='cuda:3'), covar=tensor([0.0700, 0.1551, 0.1779, 0.1683, 0.2038, 0.0784, 0.1198, 0.1921], device='cuda:3'), in_proj_covar=tensor([0.0375, 0.0542, 0.0601, 0.0451, 0.0602, 0.0630, 0.0469, 0.0602], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:00:57,136 INFO [train.py:904] (3/8) Epoch 19, batch 9750, loss[loss=0.1703, simple_loss=0.2642, pruned_loss=0.03817, over 15517.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2608, pruned_loss=0.03748, over 3056233.33 frames. ], batch size: 192, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:58,980 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:09,804 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:14,059 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:31,654 INFO [optim.py:368] (3/8) 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 00:02:23,621 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:02:37,067 INFO [train.py:904] (3/8) Epoch 19, batch 9800, loss[loss=0.1529, simple_loss=0.2434, pruned_loss=0.0312, over 12559.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2609, pruned_loss=0.03648, over 3070889.31 frames. ], batch size: 248, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:02:55,410 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192511.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:04:23,519 INFO [train.py:904] (3/8) Epoch 19, batch 9850, loss[loss=0.1642, simple_loss=0.2631, pruned_loss=0.03266, over 16980.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2622, pruned_loss=0.0364, over 3058458.74 frames. ], batch size: 109, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:05:00,568 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192575.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:05:15,470 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:05:22,110 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-05-01 00:06:14,629 INFO [train.py:904] (3/8) Epoch 19, batch 9900, loss[loss=0.1788, simple_loss=0.283, pruned_loss=0.03728, over 16267.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2626, pruned_loss=0.03632, over 3060637.68 frames. ], batch size: 165, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:06:46,562 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:07:31,357 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 00:08:13,281 INFO [train.py:904] (3/8) Epoch 19, batch 9950, loss[loss=0.1865, simple_loss=0.274, pruned_loss=0.04953, over 12546.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.265, pruned_loss=0.03682, over 3049849.32 frames. ], batch size: 250, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:08:54,784 INFO [optim.py:368] (3/8) 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 00:09:04,358 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:07,639 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 00:09:14,932 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:32,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5090, 5.8682, 5.6329, 5.6451, 5.2832, 5.3171, 5.1757, 5.9549], device='cuda:3'), covar=tensor([0.1219, 0.0814, 0.0970, 0.0725, 0.0755, 0.0615, 0.1087, 0.0832], device='cuda:3'), in_proj_covar=tensor([0.0615, 0.0757, 0.0615, 0.0561, 0.0473, 0.0482, 0.0629, 0.0579], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:10:14,416 INFO [train.py:904] (3/8) Epoch 19, batch 10000, loss[loss=0.1892, simple_loss=0.2729, pruned_loss=0.05275, over 12812.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2629, pruned_loss=0.03566, over 3081386.09 frames. ], batch size: 250, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:10:50,870 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-05-01 00:11:22,421 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192734.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:25,774 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192735.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:50,832 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 19, batch 10050, loss[loss=0.1894, simple_loss=0.2755, pruned_loss=0.05167, over 11763.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2632, pruned_loss=0.03574, over 3080811.35 frames. ], batch size: 250, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:12:02,950 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192755.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:09,568 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:32,890 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.015e+02 2.519e+02 2.909e+02 8.183e+02, threshold=5.037e+02, percent-clipped=1.0 2023-05-01 00:13:08,090 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:21,312 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192796.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:26,918 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 00:13:30,579 INFO [train.py:904] (3/8) Epoch 19, batch 10100, loss[loss=0.1647, simple_loss=0.2558, pruned_loss=0.03676, over 16375.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2636, pruned_loss=0.03593, over 3088035.44 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:13:38,854 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:42,169 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3990, 3.0317, 2.7233, 2.2779, 2.2050, 2.2535, 2.9759, 2.8207], device='cuda:3'), covar=tensor([0.2361, 0.0655, 0.1426, 0.2624, 0.2501, 0.2016, 0.0383, 0.1358], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0253, 0.0287, 0.0292, 0.0273, 0.0241, 0.0274, 0.0313], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:13:48,332 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192811.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:14:28,774 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192831.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:15:13,690 INFO [train.py:904] (3/8) Epoch 20, batch 0, loss[loss=0.1737, simple_loss=0.2551, pruned_loss=0.04614, over 16820.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2551, pruned_loss=0.04614, over 16820.00 frames. ], batch size: 39, lr: 3.43e-03, grad_scale: 8.0 2023-05-01 00:15:13,691 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 00:15:21,162 INFO [train.py:938] (3/8) Epoch 20, validation: loss=0.146, simple_loss=0.2496, pruned_loss=0.02121, over 944034.00 frames. 2023-05-01 00:15:21,163 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 00:15:32,535 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:36,130 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9863, 3.8108, 4.3221, 2.1284, 4.4786, 4.5045, 3.1561, 3.4332], device='cuda:3'), covar=tensor([0.0651, 0.0214, 0.0165, 0.1149, 0.0055, 0.0122, 0.0442, 0.0381], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0101, 0.0089, 0.0133, 0.0074, 0.0114, 0.0120, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 00:15:49,198 INFO [optim.py:368] (3/8) 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 00:15:53,693 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192875.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:56,056 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:16:17,606 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192892.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:16:31,047 INFO [train.py:904] (3/8) Epoch 20, batch 50, loss[loss=0.1954, simple_loss=0.2677, pruned_loss=0.06162, over 16712.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2758, pruned_loss=0.0527, over 745285.40 frames. ], batch size: 134, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:17:00,246 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:02,460 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:21,285 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2192, 5.1514, 4.9146, 4.5936, 4.9896, 1.8271, 4.7425, 4.7830], device='cuda:3'), covar=tensor([0.0076, 0.0074, 0.0227, 0.0300, 0.0101, 0.2712, 0.0138, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0142, 0.0184, 0.0163, 0.0161, 0.0198, 0.0174, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:17:38,938 INFO [train.py:904] (3/8) Epoch 20, batch 100, loss[loss=0.1779, simple_loss=0.2768, pruned_loss=0.03947, over 17089.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2687, pruned_loss=0.04819, over 1327886.64 frames. ], batch size: 49, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:17:48,071 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5937, 6.0153, 5.7539, 5.7401, 5.3461, 5.3811, 5.4030, 6.1506], device='cuda:3'), covar=tensor([0.1311, 0.0896, 0.1088, 0.0863, 0.0932, 0.0679, 0.1147, 0.0872], device='cuda:3'), in_proj_covar=tensor([0.0627, 0.0774, 0.0627, 0.0573, 0.0484, 0.0494, 0.0645, 0.0590], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:18:07,335 INFO [optim.py:368] (3/8) 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 00:18:07,671 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192972.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:18:34,870 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2241, 4.1386, 4.1320, 3.8299, 3.9289, 4.1746, 3.9070, 3.9619], device='cuda:3'), covar=tensor([0.0572, 0.0695, 0.0289, 0.0284, 0.0630, 0.0438, 0.0700, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0271, 0.0387, 0.0316, 0.0306, 0.0324, 0.0359, 0.0218, 0.0377], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:18:37,577 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 00:18:41,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8907, 3.9207, 2.4087, 4.6185, 3.0988, 4.5383, 2.5631, 3.4139], device='cuda:3'), covar=tensor([0.0334, 0.0470, 0.1698, 0.0245, 0.0878, 0.0560, 0.1544, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0171, 0.0189, 0.0152, 0.0172, 0.0207, 0.0198, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 00:18:48,432 INFO [train.py:904] (3/8) Epoch 20, batch 150, loss[loss=0.1768, simple_loss=0.2578, pruned_loss=0.04792, over 16387.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2671, pruned_loss=0.04729, over 1773739.98 frames. ], batch size: 75, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:19:01,819 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 00:19:27,121 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193029.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:43,985 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:52,847 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:58,179 INFO [train.py:904] (3/8) Epoch 20, batch 200, loss[loss=0.1864, simple_loss=0.2658, pruned_loss=0.05347, over 16831.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2657, pruned_loss=0.04671, over 2112920.49 frames. ], batch size: 96, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:20:03,473 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193055.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:05,133 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-01 00:20:27,491 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.769e+02 2.323e+02 2.688e+02 3.539e+02 1.444e+03, threshold=5.377e+02, percent-clipped=5.0 2023-05-01 00:20:50,730 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:53,768 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193091.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:00,957 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 250, loss[loss=0.1781, simple_loss=0.2582, pruned_loss=0.04906, over 12381.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2642, pruned_loss=0.04645, over 2371889.30 frames. ], batch size: 246, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:21:09,278 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193137.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:22:17,119 INFO [train.py:904] (3/8) Epoch 20, batch 300, loss[loss=0.166, simple_loss=0.2472, pruned_loss=0.04234, over 16876.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2609, pruned_loss=0.04548, over 2581093.93 frames. ], batch size: 90, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:22:46,393 INFO [optim.py:368] (3/8) 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 00:23:05,626 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193187.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:23:10,354 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0687, 3.0150, 1.9398, 3.2511, 2.3491, 3.2792, 2.1248, 2.5630], device='cuda:3'), covar=tensor([0.0330, 0.0449, 0.1554, 0.0314, 0.0840, 0.0598, 0.1536, 0.0715], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0157, 0.0176, 0.0212, 0.0202, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 00:23:28,161 INFO [train.py:904] (3/8) Epoch 20, batch 350, loss[loss=0.1605, simple_loss=0.2391, pruned_loss=0.04098, over 16794.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2579, pruned_loss=0.0438, over 2739139.73 frames. ], batch size: 102, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:23:39,586 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:24:37,906 INFO [train.py:904] (3/8) Epoch 20, batch 400, loss[loss=0.141, simple_loss=0.2327, pruned_loss=0.02469, over 17246.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2561, pruned_loss=0.04295, over 2863899.76 frames. ], batch size: 43, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:25:00,561 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7251, 2.8991, 2.7699, 5.0254, 4.1827, 4.4859, 1.6511, 3.2447], device='cuda:3'), covar=tensor([0.1472, 0.0806, 0.1220, 0.0192, 0.0265, 0.0417, 0.1700, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0171, 0.0192, 0.0181, 0.0199, 0.0212, 0.0197, 0.0189], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 00:25:05,937 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193272.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:06,033 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193272.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:25:06,670 INFO [optim.py:368] (3/8) 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 00:25:20,561 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3406, 2.2188, 2.3465, 4.1502, 2.2712, 2.6405, 2.3311, 2.4167], device='cuda:3'), covar=tensor([0.1346, 0.3746, 0.2998, 0.0584, 0.3987, 0.2595, 0.3943, 0.3143], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0436, 0.0362, 0.0321, 0.0432, 0.0500, 0.0406, 0.0509], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:25:23,431 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193284.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:26,084 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 00:25:46,656 INFO [train.py:904] (3/8) Epoch 20, batch 450, loss[loss=0.1576, simple_loss=0.2407, pruned_loss=0.03724, over 16331.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2551, pruned_loss=0.04188, over 2963084.02 frames. ], batch size: 165, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:26:01,946 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7522, 3.7520, 4.0203, 2.8447, 3.6529, 4.0561, 3.7735, 2.3317], device='cuda:3'), covar=tensor([0.0480, 0.0293, 0.0048, 0.0369, 0.0104, 0.0097, 0.0089, 0.0473], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0081, 0.0080, 0.0134, 0.0095, 0.0106, 0.0092, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:26:13,644 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:48,807 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 500, loss[loss=0.1622, simple_loss=0.2462, pruned_loss=0.03914, over 16939.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2541, pruned_loss=0.04179, over 3047336.59 frames. ], batch size: 41, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:27:26,047 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:51,496 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:07,092 INFO [train.py:904] (3/8) Epoch 20, batch 550, loss[loss=0.2014, simple_loss=0.2727, pruned_loss=0.06505, over 16721.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2533, pruned_loss=0.04188, over 3097504.55 frames. ], batch size: 124, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:28:32,310 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:45,450 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 00:28:46,180 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0160, 2.0832, 2.5337, 2.9031, 2.8605, 3.0165, 2.2104, 3.1250], device='cuda:3'), covar=tensor([0.0192, 0.0445, 0.0309, 0.0266, 0.0294, 0.0251, 0.0460, 0.0159], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0189, 0.0174, 0.0177, 0.0190, 0.0147, 0.0191, 0.0141], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:28:56,784 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:58,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9874, 4.7452, 5.0049, 5.2171, 5.4351, 4.7418, 5.4304, 5.4159], device='cuda:3'), covar=tensor([0.1983, 0.1402, 0.1856, 0.0873, 0.0560, 0.1046, 0.0583, 0.0634], device='cuda:3'), in_proj_covar=tensor([0.0627, 0.0778, 0.0909, 0.0795, 0.0594, 0.0621, 0.0644, 0.0743], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:29:14,799 INFO [train.py:904] (3/8) Epoch 20, batch 600, loss[loss=0.2024, simple_loss=0.2745, pruned_loss=0.06517, over 16434.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.253, pruned_loss=0.0429, over 3142051.19 frames. ], batch size: 146, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:29:18,189 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193454.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:29:38,450 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 00:29:43,211 INFO [optim.py:368] (3/8) 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 00:29:55,271 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193482.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:30:03,376 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193487.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:30:21,022 INFO [train.py:904] (3/8) Epoch 20, batch 650, loss[loss=0.1446, simple_loss=0.2261, pruned_loss=0.03152, over 16922.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2519, pruned_loss=0.0422, over 3185123.46 frames. ], batch size: 96, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:30:39,938 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193515.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:30:47,325 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7758, 5.1402, 5.2762, 5.0338, 5.0362, 5.7080, 5.1951, 4.8662], device='cuda:3'), covar=tensor([0.1334, 0.1992, 0.2383, 0.2108, 0.2895, 0.1060, 0.1937, 0.2530], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0581, 0.0645, 0.0484, 0.0649, 0.0673, 0.0504, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:31:07,553 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 700, loss[loss=0.2209, simple_loss=0.293, pruned_loss=0.07443, over 15568.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2509, pruned_loss=0.04144, over 3217431.79 frames. ], batch size: 191, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:31:48,972 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193567.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:31:57,448 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.109e+02 2.476e+02 3.086e+02 5.855e+02, threshold=4.951e+02, percent-clipped=1.0 2023-05-01 00:32:35,677 INFO [train.py:904] (3/8) Epoch 20, batch 750, loss[loss=0.1711, simple_loss=0.2508, pruned_loss=0.0457, over 16808.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.252, pruned_loss=0.04239, over 3235963.94 frames. ], batch size: 96, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:32:54,634 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:04,539 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0969, 3.2461, 3.3659, 2.0932, 2.8815, 2.3657, 3.6079, 3.5156], device='cuda:3'), covar=tensor([0.0256, 0.0928, 0.0652, 0.1913, 0.0877, 0.0983, 0.0545, 0.0982], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0158, 0.0165, 0.0151, 0.0144, 0.0127, 0.0143, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 00:33:28,498 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:42,946 INFO [train.py:904] (3/8) Epoch 20, batch 800, loss[loss=0.1774, simple_loss=0.2527, pruned_loss=0.05102, over 16679.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2515, pruned_loss=0.04221, over 3258012.46 frames. ], batch size: 134, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:33:52,600 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-05-01 00:34:05,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2539, 3.2048, 3.4489, 2.3918, 3.2095, 3.5377, 3.2882, 1.9573], device='cuda:3'), covar=tensor([0.0523, 0.0171, 0.0061, 0.0405, 0.0117, 0.0114, 0.0105, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0135, 0.0096, 0.0107, 0.0093, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:3') 2023-05-01 00:34:10,661 INFO [optim.py:368] (3/8) 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 00:34:16,605 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:45,115 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:52,172 INFO [train.py:904] (3/8) Epoch 20, batch 850, loss[loss=0.1456, simple_loss=0.237, pruned_loss=0.02706, over 17231.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2516, pruned_loss=0.04159, over 3280135.71 frames. ], batch size: 45, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:34:57,820 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6311, 3.6319, 2.9375, 2.2209, 2.4044, 2.3475, 3.8151, 3.2936], device='cuda:3'), covar=tensor([0.2719, 0.0636, 0.1569, 0.3073, 0.2829, 0.2154, 0.0514, 0.1425], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0264, 0.0299, 0.0302, 0.0289, 0.0251, 0.0286, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:35:02,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6925, 3.6894, 2.9690, 2.2638, 2.4686, 2.3997, 3.8774, 3.3648], device='cuda:3'), covar=tensor([0.2664, 0.0646, 0.1593, 0.2776, 0.2535, 0.2101, 0.0511, 0.1394], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0264, 0.0298, 0.0302, 0.0289, 0.0251, 0.0286, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:35:50,806 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:35:59,338 INFO [train.py:904] (3/8) Epoch 20, batch 900, loss[loss=0.1541, simple_loss=0.2482, pruned_loss=0.02999, over 17195.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2504, pruned_loss=0.04098, over 3294346.02 frames. ], batch size: 44, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:36:28,230 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193777.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:36:43,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2166, 4.9762, 5.2110, 5.4057, 5.5514, 4.8638, 5.5497, 5.5334], device='cuda:3'), covar=tensor([0.1834, 0.1383, 0.1916, 0.0917, 0.0691, 0.0891, 0.0706, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0634, 0.0788, 0.0920, 0.0803, 0.0600, 0.0626, 0.0650, 0.0751], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:37:09,296 INFO [train.py:904] (3/8) Epoch 20, batch 950, loss[loss=0.1572, simple_loss=0.2573, pruned_loss=0.02857, over 17137.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2511, pruned_loss=0.04104, over 3297305.91 frames. ], batch size: 48, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:37:20,558 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193810.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:37:21,698 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3510, 4.2918, 4.2169, 3.8879, 3.9669, 4.2766, 4.0670, 4.0359], device='cuda:3'), covar=tensor([0.0678, 0.0771, 0.0319, 0.0329, 0.0842, 0.0542, 0.0704, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0420, 0.0340, 0.0334, 0.0351, 0.0391, 0.0236, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:38:03,669 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-01 00:38:09,541 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4184, 4.4249, 4.7641, 4.7679, 4.7699, 4.4980, 4.5074, 4.3846], device='cuda:3'), covar=tensor([0.0409, 0.0836, 0.0420, 0.0385, 0.0551, 0.0447, 0.0984, 0.0626], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0439, 0.0425, 0.0398, 0.0475, 0.0447, 0.0539, 0.0362], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 00:38:17,860 INFO [train.py:904] (3/8) Epoch 20, batch 1000, loss[loss=0.1589, simple_loss=0.2469, pruned_loss=0.03544, over 16677.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2501, pruned_loss=0.04117, over 3291089.44 frames. ], batch size: 57, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:38:39,294 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193867.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:38:45,940 INFO [optim.py:368] (3/8) 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 00:39:00,613 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193884.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:39:24,779 INFO [train.py:904] (3/8) Epoch 20, batch 1050, loss[loss=0.1689, simple_loss=0.2533, pruned_loss=0.04227, over 16754.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2497, pruned_loss=0.04122, over 3303881.89 frames. ], batch size: 83, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:39:43,670 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:18,119 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:25,169 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:35,458 INFO [train.py:904] (3/8) Epoch 20, batch 1100, loss[loss=0.1709, simple_loss=0.2409, pruned_loss=0.05046, over 16772.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2495, pruned_loss=0.0409, over 3309464.44 frames. ], batch size: 124, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:41:01,676 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:04,218 INFO [optim.py:368] (3/8) 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 00:41:21,653 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 00:41:25,069 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:46,949 INFO [train.py:904] (3/8) Epoch 20, batch 1150, loss[loss=0.1775, simple_loss=0.2536, pruned_loss=0.05073, over 12222.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2485, pruned_loss=0.03997, over 3288660.19 frames. ], batch size: 246, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:42:30,268 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 00:42:56,045 INFO [train.py:904] (3/8) Epoch 20, batch 1200, loss[loss=0.1527, simple_loss=0.2395, pruned_loss=0.03301, over 16826.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2483, pruned_loss=0.04002, over 3297043.59 frames. ], batch size: 42, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:43:25,004 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:43:33,142 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194077.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:43:46,125 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 00:44:06,932 INFO [train.py:904] (3/8) Epoch 20, batch 1250, loss[loss=0.1734, simple_loss=0.2471, pruned_loss=0.04983, over 16640.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2482, pruned_loss=0.04022, over 3288208.99 frames. ], batch size: 134, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:44:17,627 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194110.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:44:26,997 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8039, 4.7827, 4.6834, 4.1658, 4.7452, 1.9268, 4.4916, 4.4794], device='cuda:3'), covar=tensor([0.0142, 0.0100, 0.0182, 0.0330, 0.0109, 0.2551, 0.0140, 0.0194], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0151, 0.0196, 0.0174, 0.0172, 0.0207, 0.0185, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:44:38,137 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:44:49,841 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:45:02,348 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 00:45:15,924 INFO [train.py:904] (3/8) Epoch 20, batch 1300, loss[loss=0.1483, simple_loss=0.2424, pruned_loss=0.02707, over 16843.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2485, pruned_loss=0.04032, over 3290578.90 frames. ], batch size: 42, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:45:26,571 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194158.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:45:46,645 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.297e+02 2.207e+02 2.541e+02 3.000e+02 4.873e+02, threshold=5.083e+02, percent-clipped=0.0 2023-05-01 00:45:59,967 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 00:46:05,941 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8709, 2.0024, 2.5165, 2.8208, 2.6788, 3.3083, 2.1266, 3.2622], device='cuda:3'), covar=tensor([0.0235, 0.0505, 0.0336, 0.0335, 0.0347, 0.0200, 0.0521, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0191, 0.0177, 0.0178, 0.0192, 0.0149, 0.0192, 0.0144], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:46:27,312 INFO [train.py:904] (3/8) Epoch 20, batch 1350, loss[loss=0.1612, simple_loss=0.2516, pruned_loss=0.03545, over 17251.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2488, pruned_loss=0.04048, over 3301769.37 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:47:20,514 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194240.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:47:36,939 INFO [train.py:904] (3/8) Epoch 20, batch 1400, loss[loss=0.1599, simple_loss=0.2537, pruned_loss=0.03307, over 17137.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2488, pruned_loss=0.04044, over 3299175.08 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:48:03,372 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.050e+02 2.370e+02 3.098e+02 5.438e+02, threshold=4.739e+02, percent-clipped=2.0 2023-05-01 00:48:44,297 INFO [train.py:904] (3/8) Epoch 20, batch 1450, loss[loss=0.1596, simple_loss=0.2401, pruned_loss=0.0396, over 16544.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.248, pruned_loss=0.04031, over 3297800.71 frames. ], batch size: 75, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:06,334 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194318.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:08,663 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:54,002 INFO [train.py:904] (3/8) Epoch 20, batch 1500, loss[loss=0.1853, simple_loss=0.2572, pruned_loss=0.05674, over 16844.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2481, pruned_loss=0.04094, over 3307761.07 frames. ], batch size: 116, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:55,523 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:07,989 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9694, 3.8006, 4.2238, 2.1961, 4.5069, 4.6097, 3.1739, 3.5840], device='cuda:3'), covar=tensor([0.0668, 0.0296, 0.0270, 0.1101, 0.0084, 0.0161, 0.0447, 0.0373], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0108, 0.0097, 0.0139, 0.0079, 0.0124, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 00:50:07,999 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194362.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:50:24,392 INFO [optim.py:368] (3/8) 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 00:50:31,271 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 1550, loss[loss=0.1492, simple_loss=0.2429, pruned_loss=0.02775, over 17220.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2486, pruned_loss=0.04171, over 3302061.73 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:51:20,886 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:34,144 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194423.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:51:40,589 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:44,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2671, 4.6580, 4.6411, 3.3165, 3.9130, 4.5969, 4.0488, 3.0988], device='cuda:3'), covar=tensor([0.0403, 0.0054, 0.0043, 0.0330, 0.0110, 0.0085, 0.0083, 0.0348], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0134, 0.0096, 0.0108, 0.0093, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:3') 2023-05-01 00:52:13,192 INFO [train.py:904] (3/8) Epoch 20, batch 1600, loss[loss=0.1763, simple_loss=0.2515, pruned_loss=0.05055, over 16790.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2512, pruned_loss=0.0426, over 3306242.64 frames. ], batch size: 124, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:52:43,858 INFO [optim.py:368] (3/8) 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 00:53:22,112 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-05-01 00:53:22,707 INFO [train.py:904] (3/8) Epoch 20, batch 1650, loss[loss=0.1825, simple_loss=0.2632, pruned_loss=0.05092, over 16387.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2529, pruned_loss=0.04276, over 3310027.06 frames. ], batch size: 146, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:53:29,516 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-01 00:53:56,090 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7981, 4.5913, 4.8621, 5.0371, 5.2371, 4.5563, 5.2171, 5.2299], device='cuda:3'), covar=tensor([0.1947, 0.1372, 0.1718, 0.0840, 0.0591, 0.1098, 0.0603, 0.0615], device='cuda:3'), in_proj_covar=tensor([0.0648, 0.0802, 0.0940, 0.0823, 0.0612, 0.0642, 0.0663, 0.0767], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:54:16,573 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194540.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:54:33,459 INFO [train.py:904] (3/8) Epoch 20, batch 1700, loss[loss=0.153, simple_loss=0.2448, pruned_loss=0.03062, over 17204.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2547, pruned_loss=0.0431, over 3303922.72 frames. ], batch size: 44, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:05,700 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.259e+02 2.640e+02 3.354e+02 1.280e+03, threshold=5.281e+02, percent-clipped=2.0 2023-05-01 00:55:25,665 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 1750, loss[loss=0.1509, simple_loss=0.2439, pruned_loss=0.02895, over 17204.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2554, pruned_loss=0.04287, over 3306720.08 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:47,385 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5949, 4.5803, 4.4967, 3.9614, 4.5681, 1.7695, 4.3043, 4.2360], device='cuda:3'), covar=tensor([0.0125, 0.0113, 0.0195, 0.0370, 0.0109, 0.2900, 0.0167, 0.0218], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0152, 0.0197, 0.0177, 0.0174, 0.0208, 0.0187, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:55:56,482 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6852, 2.4433, 2.3980, 4.6517, 2.3076, 2.9179, 2.4502, 2.6809], device='cuda:3'), covar=tensor([0.1122, 0.3705, 0.3092, 0.0411, 0.4159, 0.2378, 0.3573, 0.3479], device='cuda:3'), in_proj_covar=tensor([0.0397, 0.0439, 0.0364, 0.0326, 0.0434, 0.0505, 0.0409, 0.0515], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:55:56,592 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 00:56:05,243 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9795, 4.4630, 3.2577, 2.3246, 2.7301, 2.6891, 4.7996, 3.6846], device='cuda:3'), covar=tensor([0.2674, 0.0521, 0.1685, 0.2886, 0.2792, 0.1969, 0.0351, 0.1395], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0266, 0.0300, 0.0304, 0.0291, 0.0252, 0.0287, 0.0330], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 00:56:27,125 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194634.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:56:33,596 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0527, 4.0611, 3.9665, 3.6637, 3.7558, 4.0419, 3.7288, 3.8582], device='cuda:3'), covar=tensor([0.0601, 0.0644, 0.0276, 0.0265, 0.0637, 0.0441, 0.0925, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0429, 0.0347, 0.0341, 0.0358, 0.0399, 0.0239, 0.0415], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:56:51,750 INFO [train.py:904] (3/8) Epoch 20, batch 1800, loss[loss=0.1504, simple_loss=0.2425, pruned_loss=0.02916, over 16840.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2563, pruned_loss=0.04311, over 3308034.31 frames. ], batch size: 42, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:57:12,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7837, 2.7402, 2.3488, 2.5436, 3.0219, 2.8510, 3.3286, 3.3230], device='cuda:3'), covar=tensor([0.0132, 0.0415, 0.0529, 0.0479, 0.0304, 0.0382, 0.0273, 0.0267], device='cuda:3'), in_proj_covar=tensor([0.0207, 0.0238, 0.0227, 0.0230, 0.0240, 0.0238, 0.0238, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 00:57:22,476 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 20, batch 1850, loss[loss=0.1536, simple_loss=0.243, pruned_loss=0.03214, over 17229.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2568, pruned_loss=0.04263, over 3320636.38 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:58:08,651 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:21,442 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194718.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:58:33,777 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:59:06,818 INFO [train.py:904] (3/8) Epoch 20, batch 1900, loss[loss=0.1635, simple_loss=0.25, pruned_loss=0.03848, over 16802.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2566, pruned_loss=0.04222, over 3310094.13 frames. ], batch size: 102, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:59:33,815 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 00:59:38,382 INFO [optim.py:368] (3/8) 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 00:59:41,141 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:59:58,684 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8150, 2.8123, 2.4674, 2.6153, 3.0491, 2.9181, 3.4557, 3.3609], device='cuda:3'), covar=tensor([0.0131, 0.0401, 0.0493, 0.0459, 0.0277, 0.0350, 0.0230, 0.0257], device='cuda:3'), in_proj_covar=tensor([0.0206, 0.0237, 0.0226, 0.0229, 0.0239, 0.0237, 0.0237, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:00:16,323 INFO [train.py:904] (3/8) Epoch 20, batch 1950, loss[loss=0.1777, simple_loss=0.2559, pruned_loss=0.04978, over 16721.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2565, pruned_loss=0.04146, over 3299617.14 frames. ], batch size: 134, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 01:01:11,635 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3452, 5.2840, 5.0923, 4.6017, 5.1929, 2.1341, 4.8943, 5.0829], device='cuda:3'), covar=tensor([0.0084, 0.0092, 0.0208, 0.0389, 0.0103, 0.2479, 0.0149, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0152, 0.0197, 0.0177, 0.0174, 0.0208, 0.0188, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:01:17,082 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:01:23,588 INFO [train.py:904] (3/8) Epoch 20, batch 2000, loss[loss=0.174, simple_loss=0.2477, pruned_loss=0.05016, over 16804.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2553, pruned_loss=0.04143, over 3304841.11 frames. ], batch size: 83, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:01:51,822 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0842, 5.1451, 5.5795, 5.5469, 5.5561, 5.2029, 5.1323, 4.9792], device='cuda:3'), covar=tensor([0.0326, 0.0501, 0.0335, 0.0390, 0.0423, 0.0337, 0.0925, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0452, 0.0436, 0.0410, 0.0487, 0.0463, 0.0554, 0.0370], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 01:01:54,985 INFO [optim.py:368] (3/8) 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 01:02:10,626 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8479, 1.9846, 2.4698, 2.8653, 2.6240, 3.3885, 2.2470, 3.3554], device='cuda:3'), covar=tensor([0.0253, 0.0523, 0.0350, 0.0344, 0.0379, 0.0215, 0.0469, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0194, 0.0179, 0.0182, 0.0195, 0.0153, 0.0195, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:02:32,415 INFO [train.py:904] (3/8) Epoch 20, batch 2050, loss[loss=0.1434, simple_loss=0.2342, pruned_loss=0.02629, over 17224.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2549, pruned_loss=0.04141, over 3307082.81 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:02:41,398 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194908.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:03:41,679 INFO [train.py:904] (3/8) Epoch 20, batch 2100, loss[loss=0.2085, simple_loss=0.2814, pruned_loss=0.06775, over 16693.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2563, pruned_loss=0.04171, over 3304433.47 frames. ], batch size: 134, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:04:06,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0658, 4.0605, 3.9547, 3.6573, 3.7722, 4.0671, 3.7166, 3.8708], device='cuda:3'), covar=tensor([0.0646, 0.0666, 0.0279, 0.0268, 0.0561, 0.0431, 0.0986, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0431, 0.0348, 0.0342, 0.0360, 0.0401, 0.0240, 0.0418], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:04:12,916 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194974.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:13,686 INFO [optim.py:368] (3/8) 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 01:04:24,311 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5159, 5.5116, 5.3920, 4.9418, 5.0137, 5.4730, 5.4259, 5.1283], device='cuda:3'), covar=tensor([0.0660, 0.0543, 0.0266, 0.0335, 0.0999, 0.0400, 0.0276, 0.0635], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0430, 0.0348, 0.0342, 0.0360, 0.0401, 0.0240, 0.0418], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:04:34,696 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:50,816 INFO [train.py:904] (3/8) Epoch 20, batch 2150, loss[loss=0.1453, simple_loss=0.2332, pruned_loss=0.02869, over 17009.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2575, pruned_loss=0.04242, over 3311601.48 frames. ], batch size: 41, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:05:01,411 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:03,735 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:05,911 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 01:05:13,321 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195018.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:05:18,550 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:49,769 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195044.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:06:01,906 INFO [train.py:904] (3/8) Epoch 20, batch 2200, loss[loss=0.1524, simple_loss=0.2509, pruned_loss=0.02694, over 17239.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2584, pruned_loss=0.04247, over 3320958.04 frames. ], batch size: 45, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:06:09,187 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:17,081 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 01:06:21,183 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195066.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:06:28,226 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:30,162 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195072.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:06:33,763 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.410e+02 2.761e+02 3.300e+02 5.006e+02, threshold=5.521e+02, percent-clipped=1.0 2023-05-01 01:07:10,547 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 01:07:10,874 INFO [train.py:904] (3/8) Epoch 20, batch 2250, loss[loss=0.1997, simple_loss=0.2932, pruned_loss=0.05312, over 16713.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.259, pruned_loss=0.04305, over 3321264.99 frames. ], batch size: 62, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:07:15,416 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195105.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:07:54,065 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:08:13,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2024, 5.7366, 5.8928, 5.6123, 5.6139, 6.2411, 5.7038, 5.4122], device='cuda:3'), covar=tensor([0.0906, 0.2080, 0.2319, 0.1992, 0.2670, 0.0947, 0.1666, 0.2490], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0610, 0.0667, 0.0506, 0.0671, 0.0698, 0.0518, 0.0675], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:08:13,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0019, 4.7056, 4.9833, 5.1694, 5.4271, 4.6435, 5.4058, 5.3804], device='cuda:3'), covar=tensor([0.2013, 0.1609, 0.2050, 0.0956, 0.0585, 0.1041, 0.0660, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0820, 0.0956, 0.0839, 0.0624, 0.0658, 0.0672, 0.0778], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:08:21,143 INFO [train.py:904] (3/8) Epoch 20, batch 2300, loss[loss=0.1862, simple_loss=0.2682, pruned_loss=0.05214, over 16458.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2597, pruned_loss=0.04362, over 3317824.66 frames. ], batch size: 68, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:08:51,383 INFO [optim.py:368] (3/8) 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 01:09:29,600 INFO [train.py:904] (3/8) Epoch 20, batch 2350, loss[loss=0.1843, simple_loss=0.2682, pruned_loss=0.05017, over 16869.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2591, pruned_loss=0.04334, over 3325254.99 frames. ], batch size: 96, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:09:31,094 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:10:36,693 INFO [train.py:904] (3/8) Epoch 20, batch 2400, loss[loss=0.1645, simple_loss=0.2674, pruned_loss=0.0308, over 17257.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2595, pruned_loss=0.04316, over 3324202.13 frames. ], batch size: 52, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:11:07,017 INFO [optim.py:368] (3/8) 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 01:11:29,084 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 2450, loss[loss=0.1854, simple_loss=0.265, pruned_loss=0.05289, over 16939.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2605, pruned_loss=0.04335, over 3321716.95 frames. ], batch size: 116, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:35,686 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:12:54,078 INFO [train.py:904] (3/8) Epoch 20, batch 2500, loss[loss=0.1786, simple_loss=0.2575, pruned_loss=0.04984, over 16855.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2609, pruned_loss=0.0438, over 3316391.00 frames. ], batch size: 96, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:54,931 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 01:13:07,192 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 01:13:15,892 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195367.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:13:26,924 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.185e+02 2.562e+02 3.023e+02 6.708e+02, threshold=5.124e+02, percent-clipped=4.0 2023-05-01 01:13:50,872 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:01,553 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195400.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:14:04,079 INFO [train.py:904] (3/8) Epoch 20, batch 2550, loss[loss=0.1597, simple_loss=0.2385, pruned_loss=0.04046, over 16751.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.262, pruned_loss=0.04429, over 3310962.53 frames. ], batch size: 39, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:14:38,111 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:11,782 INFO [train.py:904] (3/8) Epoch 20, batch 2600, loss[loss=0.1743, simple_loss=0.2573, pruned_loss=0.04569, over 16770.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2615, pruned_loss=0.04373, over 3321421.71 frames. ], batch size: 83, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:15:13,381 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:40,236 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6670, 6.0364, 5.7736, 5.8875, 5.4829, 5.4067, 5.4516, 6.1631], device='cuda:3'), covar=tensor([0.1537, 0.0994, 0.1141, 0.0916, 0.0983, 0.0693, 0.1154, 0.0939], device='cuda:3'), in_proj_covar=tensor([0.0661, 0.0817, 0.0665, 0.0610, 0.0511, 0.0515, 0.0681, 0.0624], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:15:42,986 INFO [optim.py:368] (3/8) 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 01:16:01,527 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3787, 5.3311, 5.2299, 4.7780, 4.9066, 5.3110, 5.2182, 4.9125], device='cuda:3'), covar=tensor([0.0587, 0.0457, 0.0259, 0.0317, 0.0998, 0.0383, 0.0301, 0.0703], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0432, 0.0349, 0.0344, 0.0361, 0.0400, 0.0241, 0.0417], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:16:20,745 INFO [train.py:904] (3/8) Epoch 20, batch 2650, loss[loss=0.1734, simple_loss=0.2558, pruned_loss=0.04554, over 15673.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2619, pruned_loss=0.04326, over 3320874.37 frames. ], batch size: 191, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:16:22,192 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:16:24,502 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3067, 1.5486, 2.0424, 2.1196, 2.3238, 2.3506, 1.7681, 2.3501], device='cuda:3'), covar=tensor([0.0200, 0.0474, 0.0277, 0.0316, 0.0295, 0.0263, 0.0511, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0194, 0.0179, 0.0183, 0.0195, 0.0153, 0.0196, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:16:48,641 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4475, 4.4181, 4.8172, 4.8110, 4.8111, 4.5129, 4.5104, 4.3647], device='cuda:3'), covar=tensor([0.0354, 0.0626, 0.0393, 0.0414, 0.0535, 0.0463, 0.0905, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0448, 0.0432, 0.0406, 0.0480, 0.0457, 0.0548, 0.0367], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 01:17:03,643 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 01:17:20,731 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 01:17:28,727 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:29,628 INFO [train.py:904] (3/8) Epoch 20, batch 2700, loss[loss=0.1619, simple_loss=0.2521, pruned_loss=0.03584, over 17191.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2619, pruned_loss=0.04264, over 3320666.25 frames. ], batch size: 44, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:00,661 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.029e+02 2.571e+02 3.039e+02 8.808e+02, threshold=5.142e+02, percent-clipped=4.0 2023-05-01 01:18:31,029 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-01 01:18:39,448 INFO [train.py:904] (3/8) Epoch 20, batch 2750, loss[loss=0.1648, simple_loss=0.259, pruned_loss=0.03528, over 17112.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2616, pruned_loss=0.04217, over 3327408.25 frames. ], batch size: 48, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:47,413 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 01:19:40,426 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0099, 5.3709, 5.1238, 5.1321, 4.8787, 4.8208, 4.8265, 5.4693], device='cuda:3'), covar=tensor([0.1268, 0.0953, 0.1074, 0.0884, 0.0811, 0.1019, 0.1183, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0667, 0.0822, 0.0670, 0.0615, 0.0515, 0.0519, 0.0686, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:19:47,491 INFO [train.py:904] (3/8) Epoch 20, batch 2800, loss[loss=0.1475, simple_loss=0.2421, pruned_loss=0.02643, over 17197.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2616, pruned_loss=0.04242, over 3318295.09 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:20:07,202 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 2.150e+02 2.557e+02 3.074e+02 6.713e+02, threshold=5.114e+02, percent-clipped=2.0 2023-05-01 01:20:39,418 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6833, 3.8205, 2.4730, 4.2011, 2.9340, 4.1469, 2.4918, 3.0424], device='cuda:3'), covar=tensor([0.0287, 0.0359, 0.1459, 0.0335, 0.0788, 0.0624, 0.1362, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0179, 0.0195, 0.0165, 0.0178, 0.0219, 0.0203, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:20:52,400 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195700.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:20:54,305 INFO [train.py:904] (3/8) Epoch 20, batch 2850, loss[loss=0.17, simple_loss=0.2619, pruned_loss=0.03901, over 16446.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2615, pruned_loss=0.04249, over 3312634.44 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:21:13,265 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195715.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:28,568 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 01:21:29,691 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:56,478 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195748.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:21:56,485 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 2900, loss[loss=0.2018, simple_loss=0.2789, pruned_loss=0.06239, over 16477.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2612, pruned_loss=0.04285, over 3321162.49 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:22:33,088 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:22:44,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0079, 4.3843, 4.4292, 3.2036, 3.7173, 4.4446, 3.9623, 2.7440], device='cuda:3'), covar=tensor([0.0464, 0.0078, 0.0044, 0.0350, 0.0156, 0.0099, 0.0096, 0.0425], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0083, 0.0082, 0.0134, 0.0097, 0.0109, 0.0094, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:3') 2023-05-01 01:23:10,989 INFO [train.py:904] (3/8) Epoch 20, batch 2950, loss[loss=0.1947, simple_loss=0.2914, pruned_loss=0.04899, over 16730.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2598, pruned_loss=0.04256, over 3328326.41 frames. ], batch size: 57, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,942 INFO [train.py:904] (3/8) Epoch 20, batch 3000, loss[loss=0.2034, simple_loss=0.2742, pruned_loss=0.06629, over 15469.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2598, pruned_loss=0.04377, over 3316091.53 frames. ], batch size: 191, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,942 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 01:24:25,539 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1086, 3.1859, 3.0554, 5.0172, 3.9910, 4.5460, 1.8871, 3.4257], device='cuda:3'), covar=tensor([0.1223, 0.0680, 0.1021, 0.0154, 0.0158, 0.0307, 0.1498, 0.0685], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0187, 0.0204, 0.0214, 0.0197, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:24:27,133 INFO [train.py:938] (3/8) Epoch 20, validation: loss=0.1354, simple_loss=0.2409, pruned_loss=0.01492, over 944034.00 frames. 2023-05-01 01:24:27,134 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 01:24:58,716 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.268e+02 2.640e+02 3.074e+02 4.713e+02, threshold=5.280e+02, percent-clipped=0.0 2023-05-01 01:25:38,246 INFO [train.py:904] (3/8) Epoch 20, batch 3050, loss[loss=0.1802, simple_loss=0.2749, pruned_loss=0.04278, over 17060.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.259, pruned_loss=0.04307, over 3322740.21 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:26:07,543 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4676, 3.7248, 3.9100, 2.1531, 3.2415, 2.5176, 3.9798, 3.9567], device='cuda:3'), covar=tensor([0.0252, 0.0825, 0.0510, 0.1939, 0.0753, 0.0962, 0.0555, 0.1005], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0164, 0.0166, 0.0151, 0.0144, 0.0129, 0.0144, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:26:46,789 INFO [train.py:904] (3/8) Epoch 20, batch 3100, loss[loss=0.1702, simple_loss=0.2436, pruned_loss=0.04835, over 16486.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2578, pruned_loss=0.04285, over 3323349.95 frames. ], batch size: 146, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:27:16,965 INFO [optim.py:368] (3/8) 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 01:27:43,482 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1291, 4.8753, 5.1585, 5.3368, 5.5659, 4.8136, 5.5218, 5.5348], device='cuda:3'), covar=tensor([0.2023, 0.1438, 0.1773, 0.0823, 0.0537, 0.0796, 0.0550, 0.0585], device='cuda:3'), in_proj_covar=tensor([0.0667, 0.0827, 0.0963, 0.0844, 0.0631, 0.0659, 0.0674, 0.0784], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:27:55,506 INFO [train.py:904] (3/8) Epoch 20, batch 3150, loss[loss=0.1914, simple_loss=0.2769, pruned_loss=0.0529, over 16483.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2574, pruned_loss=0.04283, over 3317405.09 frames. ], batch size: 75, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:28:57,124 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:29:03,218 INFO [train.py:904] (3/8) Epoch 20, batch 3200, loss[loss=0.1322, simple_loss=0.2159, pruned_loss=0.02428, over 16778.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.257, pruned_loss=0.04257, over 3325056.58 frames. ], batch size: 39, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:29:13,628 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-01 01:29:24,646 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0842, 5.4483, 5.1898, 5.2010, 4.9230, 4.9091, 4.8021, 5.5368], device='cuda:3'), covar=tensor([0.1321, 0.0938, 0.1054, 0.0882, 0.0853, 0.1010, 0.1283, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0673, 0.0831, 0.0677, 0.0621, 0.0521, 0.0523, 0.0694, 0.0635], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:29:32,951 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6932, 4.5911, 4.5597, 4.2486, 4.2811, 4.6083, 4.4339, 4.3854], device='cuda:3'), covar=tensor([0.0615, 0.0788, 0.0321, 0.0308, 0.0911, 0.0507, 0.0498, 0.0660], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0436, 0.0354, 0.0349, 0.0365, 0.0406, 0.0244, 0.0423], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:29:35,520 INFO [optim.py:368] (3/8) 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 01:30:04,479 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=196096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:30:11,566 INFO [train.py:904] (3/8) Epoch 20, batch 3250, loss[loss=0.2288, simple_loss=0.3047, pruned_loss=0.07648, over 16390.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2568, pruned_loss=0.04208, over 3332181.63 frames. ], batch size: 145, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:30:52,691 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0209, 4.5298, 2.9915, 2.4498, 2.7103, 2.5143, 4.7253, 3.5810], device='cuda:3'), covar=tensor([0.2732, 0.0541, 0.1932, 0.2938, 0.3001, 0.2254, 0.0412, 0.1473], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0270, 0.0303, 0.0307, 0.0298, 0.0255, 0.0292, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:30:56,208 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0506, 2.1522, 2.7513, 2.9985, 2.8210, 3.4370, 2.5172, 3.4436], device='cuda:3'), covar=tensor([0.0251, 0.0450, 0.0304, 0.0316, 0.0333, 0.0215, 0.0409, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0193, 0.0178, 0.0183, 0.0195, 0.0154, 0.0194, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:31:06,391 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 01:31:19,979 INFO [train.py:904] (3/8) Epoch 20, batch 3300, loss[loss=0.1447, simple_loss=0.2293, pruned_loss=0.03006, over 16987.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.257, pruned_loss=0.04196, over 3334145.26 frames. ], batch size: 41, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:52,355 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.123e+02 2.550e+02 3.106e+02 4.546e+02, threshold=5.100e+02, percent-clipped=0.0 2023-05-01 01:32:28,250 INFO [train.py:904] (3/8) Epoch 20, batch 3350, loss[loss=0.1759, simple_loss=0.2497, pruned_loss=0.051, over 16930.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.258, pruned_loss=0.04219, over 3331671.34 frames. ], batch size: 109, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:32:55,684 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4016, 3.5369, 3.9201, 2.1412, 3.0286, 2.4650, 3.7653, 3.6726], device='cuda:3'), covar=tensor([0.0315, 0.0921, 0.0489, 0.1925, 0.0847, 0.0950, 0.0619, 0.1118], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0144, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:32:58,370 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 01:33:22,486 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9680, 4.0979, 2.5944, 4.6870, 3.1494, 4.6771, 2.7560, 3.3299], device='cuda:3'), covar=tensor([0.0288, 0.0373, 0.1548, 0.0342, 0.0795, 0.0457, 0.1389, 0.0720], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0181, 0.0197, 0.0167, 0.0179, 0.0222, 0.0205, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:33:24,913 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6443, 3.9765, 4.0490, 2.2685, 3.3090, 2.9586, 4.0340, 4.1537], device='cuda:3'), covar=tensor([0.0287, 0.0796, 0.0544, 0.1986, 0.0798, 0.0872, 0.0598, 0.1045], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:33:35,776 INFO [train.py:904] (3/8) Epoch 20, batch 3400, loss[loss=0.1411, simple_loss=0.233, pruned_loss=0.0246, over 17232.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2578, pruned_loss=0.0423, over 3334589.62 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:34:06,839 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 20, batch 3450, loss[loss=0.1666, simple_loss=0.2579, pruned_loss=0.03769, over 16536.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2572, pruned_loss=0.04227, over 3319280.47 frames. ], batch size: 68, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:35:10,228 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:35:50,467 INFO [train.py:904] (3/8) Epoch 20, batch 3500, loss[loss=0.1487, simple_loss=0.2407, pruned_loss=0.02836, over 17035.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2556, pruned_loss=0.04135, over 3327669.24 frames. ], batch size: 55, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:36:03,280 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196360.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:36:23,665 INFO [optim.py:368] (3/8) 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 01:36:35,832 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:01,668 INFO [train.py:904] (3/8) Epoch 20, batch 3550, loss[loss=0.1631, simple_loss=0.2577, pruned_loss=0.03424, over 17073.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.254, pruned_loss=0.04089, over 3329203.19 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:37:28,377 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:59,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2382, 3.5070, 3.5531, 3.5376, 3.5455, 3.4144, 3.4162, 3.4652], device='cuda:3'), covar=tensor([0.0433, 0.0552, 0.0485, 0.0472, 0.0592, 0.0530, 0.0750, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0420, 0.0463, 0.0447, 0.0416, 0.0495, 0.0473, 0.0567, 0.0379], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 01:38:10,214 INFO [train.py:904] (3/8) Epoch 20, batch 3600, loss[loss=0.1482, simple_loss=0.2335, pruned_loss=0.03146, over 16774.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.253, pruned_loss=0.04095, over 3303856.84 frames. ], batch size: 39, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:38:17,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2794, 4.1201, 4.3530, 4.4562, 4.5576, 4.1149, 4.3524, 4.5373], device='cuda:3'), covar=tensor([0.1588, 0.1186, 0.1222, 0.0742, 0.0654, 0.1232, 0.2184, 0.0861], device='cuda:3'), in_proj_covar=tensor([0.0662, 0.0821, 0.0958, 0.0842, 0.0626, 0.0655, 0.0669, 0.0780], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:38:32,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5932, 2.5364, 2.0033, 2.1137, 2.8076, 2.5332, 3.2859, 3.1546], device='cuda:3'), covar=tensor([0.0169, 0.0486, 0.0669, 0.0640, 0.0370, 0.0516, 0.0246, 0.0295], device='cuda:3'), in_proj_covar=tensor([0.0210, 0.0239, 0.0227, 0.0230, 0.0240, 0.0238, 0.0241, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:38:41,979 INFO [optim.py:368] (3/8) 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 01:39:20,690 INFO [train.py:904] (3/8) Epoch 20, batch 3650, loss[loss=0.1449, simple_loss=0.2186, pruned_loss=0.03559, over 16743.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2524, pruned_loss=0.04115, over 3299225.89 frames. ], batch size: 83, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:39:44,938 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7322, 3.9162, 4.0828, 3.0157, 3.7509, 4.1616, 3.8394, 2.4497], device='cuda:3'), covar=tensor([0.0485, 0.0220, 0.0058, 0.0346, 0.0092, 0.0106, 0.0088, 0.0464], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0082, 0.0082, 0.0133, 0.0097, 0.0108, 0.0093, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:39:48,180 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 01:40:01,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6121, 3.4314, 3.8254, 2.0108, 3.9815, 3.9828, 3.1888, 3.0141], device='cuda:3'), covar=tensor([0.0773, 0.0283, 0.0194, 0.1146, 0.0108, 0.0166, 0.0377, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0108, 0.0097, 0.0138, 0.0079, 0.0125, 0.0126, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:40:32,641 INFO [train.py:904] (3/8) Epoch 20, batch 3700, loss[loss=0.1659, simple_loss=0.2485, pruned_loss=0.04168, over 15511.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.251, pruned_loss=0.04258, over 3287817.82 frames. ], batch size: 191, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:40:37,542 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7676, 2.3860, 1.7382, 2.0759, 2.7713, 2.5604, 2.9340, 2.8635], device='cuda:3'), covar=tensor([0.0211, 0.0473, 0.0735, 0.0628, 0.0299, 0.0396, 0.0250, 0.0366], device='cuda:3'), in_proj_covar=tensor([0.0209, 0.0238, 0.0226, 0.0229, 0.0239, 0.0237, 0.0241, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:41:07,135 INFO [optim.py:368] (3/8) 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 01:41:11,763 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:41:16,600 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4182, 3.4332, 2.1805, 3.5803, 2.7251, 3.6111, 2.3794, 2.8621], device='cuda:3'), covar=tensor([0.0249, 0.0401, 0.1454, 0.0299, 0.0695, 0.0711, 0.1219, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0179, 0.0196, 0.0167, 0.0178, 0.0221, 0.0204, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:41:25,975 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 01:41:47,089 INFO [train.py:904] (3/8) Epoch 20, batch 3750, loss[loss=0.1813, simple_loss=0.2639, pruned_loss=0.04938, over 15473.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2515, pruned_loss=0.04401, over 3273602.20 frames. ], batch size: 190, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:42:36,638 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6244, 2.3986, 1.8467, 2.1121, 2.8260, 2.5275, 2.7234, 2.8816], device='cuda:3'), covar=tensor([0.0215, 0.0428, 0.0561, 0.0532, 0.0244, 0.0376, 0.0217, 0.0273], device='cuda:3'), in_proj_covar=tensor([0.0209, 0.0238, 0.0226, 0.0229, 0.0239, 0.0236, 0.0240, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:42:38,968 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 3800, loss[loss=0.1787, simple_loss=0.255, pruned_loss=0.05124, over 16667.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.253, pruned_loss=0.04518, over 3259947.78 frames. ], batch size: 76, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:43:31,147 INFO [optim.py:368] (3/8) 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 01:43:35,347 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 3850, loss[loss=0.1711, simple_loss=0.2469, pruned_loss=0.04765, over 15719.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2534, pruned_loss=0.04599, over 3255448.30 frames. ], batch size: 191, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:44:32,371 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:45:08,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0251, 2.7332, 2.7630, 2.0458, 2.5740, 2.1580, 2.7100, 2.9545], device='cuda:3'), covar=tensor([0.0331, 0.0758, 0.0550, 0.1771, 0.0829, 0.0837, 0.0606, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:45:24,178 INFO [train.py:904] (3/8) Epoch 20, batch 3900, loss[loss=0.1903, simple_loss=0.2583, pruned_loss=0.06118, over 16862.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2524, pruned_loss=0.0465, over 3265322.89 frames. ], batch size: 109, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:45:57,453 INFO [optim.py:368] (3/8) 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 01:46:04,926 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196779.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:46:13,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7451, 4.0602, 3.0086, 2.4396, 2.7471, 2.7229, 4.3925, 3.6911], device='cuda:3'), covar=tensor([0.2589, 0.0532, 0.1615, 0.2227, 0.2373, 0.1669, 0.0365, 0.0941], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0268, 0.0302, 0.0306, 0.0297, 0.0254, 0.0293, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:46:36,464 INFO [train.py:904] (3/8) Epoch 20, batch 3950, loss[loss=0.2035, simple_loss=0.2728, pruned_loss=0.06708, over 16427.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2526, pruned_loss=0.04712, over 3270201.89 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:46:58,101 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5128, 3.4227, 3.7598, 2.0477, 3.8387, 3.8773, 3.0963, 2.9706], device='cuda:3'), covar=tensor([0.0736, 0.0209, 0.0146, 0.1016, 0.0097, 0.0176, 0.0347, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0139, 0.0079, 0.0125, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:47:08,491 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 01:47:10,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3392, 4.1625, 4.4221, 4.5492, 4.6724, 4.2644, 4.4793, 4.6601], device='cuda:3'), covar=tensor([0.1653, 0.1217, 0.1305, 0.0691, 0.0603, 0.1022, 0.1782, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0656, 0.0814, 0.0944, 0.0830, 0.0621, 0.0646, 0.0666, 0.0772], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:47:11,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1036, 3.1689, 3.3115, 2.0371, 2.8648, 2.3067, 3.5797, 3.5408], device='cuda:3'), covar=tensor([0.0223, 0.0856, 0.0578, 0.1863, 0.0816, 0.0919, 0.0466, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0164, 0.0165, 0.0150, 0.0143, 0.0128, 0.0144, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:47:31,563 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196840.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:48,374 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7355, 2.9198, 2.3328, 2.6127, 3.1740, 2.8263, 3.2605, 3.3457], device='cuda:3'), covar=tensor([0.0070, 0.0316, 0.0475, 0.0408, 0.0220, 0.0322, 0.0198, 0.0221], device='cuda:3'), in_proj_covar=tensor([0.0206, 0.0235, 0.0224, 0.0226, 0.0236, 0.0234, 0.0238, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:47:48,960 INFO [train.py:904] (3/8) Epoch 20, batch 4000, loss[loss=0.185, simple_loss=0.2658, pruned_loss=0.0521, over 17003.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2533, pruned_loss=0.04773, over 3278300.98 frames. ], batch size: 55, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:52,321 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:48:21,944 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.289e+02 2.579e+02 3.040e+02 5.791e+02, threshold=5.158e+02, percent-clipped=1.0 2023-05-01 01:48:32,410 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 4050, loss[loss=0.1683, simple_loss=0.2649, pruned_loss=0.03588, over 16837.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2543, pruned_loss=0.04693, over 3277306.39 frames. ], batch size: 102, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:49:18,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3609, 4.4543, 4.6056, 4.3672, 4.3874, 4.9901, 4.4894, 4.1648], device='cuda:3'), covar=tensor([0.1521, 0.1906, 0.2013, 0.2025, 0.2691, 0.1052, 0.1681, 0.2723], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0600, 0.0659, 0.0504, 0.0665, 0.0695, 0.0514, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:49:19,939 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:19,979 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8373, 2.6604, 2.8645, 2.0321, 2.6650, 2.0684, 2.7326, 2.8208], device='cuda:3'), covar=tensor([0.0256, 0.0693, 0.0514, 0.1798, 0.0737, 0.0916, 0.0488, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0164, 0.0166, 0.0151, 0.0143, 0.0128, 0.0144, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 01:49:26,540 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2430, 2.2050, 2.2461, 3.9152, 2.2241, 2.5724, 2.3252, 2.4005], device='cuda:3'), covar=tensor([0.1266, 0.3409, 0.2878, 0.0604, 0.3792, 0.2385, 0.3344, 0.3194], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0444, 0.0366, 0.0329, 0.0435, 0.0514, 0.0414, 0.0522], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:49:37,988 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6731, 3.9016, 4.0692, 4.0201, 4.0431, 3.8371, 3.7106, 3.8732], device='cuda:3'), covar=tensor([0.0487, 0.0657, 0.0481, 0.0600, 0.0622, 0.0553, 0.1186, 0.0568], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0453, 0.0435, 0.0408, 0.0484, 0.0462, 0.0550, 0.0369], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 01:49:47,743 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:58,019 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 4100, loss[loss=0.1862, simple_loss=0.2709, pruned_loss=0.05075, over 16791.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2556, pruned_loss=0.04653, over 3258620.43 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:50:44,593 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 1.746e+02 2.102e+02 2.519e+02 4.672e+02, threshold=4.204e+02, percent-clipped=0.0 2023-05-01 01:50:48,514 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:51:23,807 INFO [train.py:904] (3/8) Epoch 20, batch 4150, loss[loss=0.1902, simple_loss=0.2851, pruned_loss=0.04766, over 16863.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2628, pruned_loss=0.04884, over 3232049.35 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:51:45,576 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197016.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:00,049 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:39,292 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 01:52:40,716 INFO [train.py:904] (3/8) Epoch 20, batch 4200, loss[loss=0.2109, simple_loss=0.3014, pruned_loss=0.06023, over 16741.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2694, pruned_loss=0.05015, over 3214293.61 frames. ], batch size: 124, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:52:48,695 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7293, 4.9779, 5.1630, 4.8524, 4.8681, 5.5148, 4.9755, 4.7293], device='cuda:3'), covar=tensor([0.1046, 0.1889, 0.1736, 0.1948, 0.2452, 0.0817, 0.1441, 0.2286], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0593, 0.0648, 0.0495, 0.0655, 0.0685, 0.0505, 0.0660], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:52:58,762 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:53:12,971 INFO [optim.py:368] (3/8) 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 01:53:39,378 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0684, 3.4338, 3.4224, 2.3975, 3.1959, 3.4493, 3.2290, 2.0735], device='cuda:3'), covar=tensor([0.0521, 0.0060, 0.0071, 0.0353, 0.0096, 0.0120, 0.0094, 0.0417], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0133, 0.0096, 0.0108, 0.0093, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:53:51,290 INFO [train.py:904] (3/8) Epoch 20, batch 4250, loss[loss=0.2048, simple_loss=0.3031, pruned_loss=0.0533, over 16737.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2729, pruned_loss=0.04964, over 3205044.72 frames. ], batch size: 76, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:54:40,508 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:54:50,559 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-01 01:55:04,298 INFO [train.py:904] (3/8) Epoch 20, batch 4300, loss[loss=0.1829, simple_loss=0.2766, pruned_loss=0.04461, over 16385.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2738, pruned_loss=0.04893, over 3188531.16 frames. ], batch size: 146, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:55:32,376 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0432, 5.3546, 4.9091, 5.2393, 4.8520, 4.5414, 4.8139, 5.4173], device='cuda:3'), covar=tensor([0.1927, 0.1248, 0.2165, 0.1183, 0.1539, 0.1586, 0.2141, 0.1498], device='cuda:3'), in_proj_covar=tensor([0.0659, 0.0808, 0.0665, 0.0607, 0.0510, 0.0516, 0.0680, 0.0627], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:55:37,987 INFO [optim.py:368] (3/8) 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 01:56:06,821 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 01:56:18,918 INFO [train.py:904] (3/8) Epoch 20, batch 4350, loss[loss=0.1981, simple_loss=0.2946, pruned_loss=0.05079, over 16738.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2776, pruned_loss=0.05047, over 3175518.48 frames. ], batch size: 83, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:56:29,915 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:05,467 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:09,110 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 4400, loss[loss=0.2028, simple_loss=0.2816, pruned_loss=0.06196, over 11527.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2796, pruned_loss=0.05131, over 3189269.38 frames. ], batch size: 247, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:57:32,552 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6990, 3.9916, 3.0812, 2.3924, 2.7169, 2.6595, 4.3647, 3.5473], device='cuda:3'), covar=tensor([0.2824, 0.0598, 0.1622, 0.2420, 0.2402, 0.1784, 0.0396, 0.1065], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0268, 0.0302, 0.0306, 0.0296, 0.0253, 0.0292, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:57:55,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4474, 5.7842, 5.5344, 5.6032, 5.2593, 5.0337, 5.2240, 5.8997], device='cuda:3'), covar=tensor([0.1105, 0.0721, 0.0987, 0.0672, 0.0735, 0.0716, 0.1023, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0809, 0.0666, 0.0608, 0.0510, 0.0517, 0.0681, 0.0630], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 01:58:04,305 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.097e+02 2.441e+02 2.762e+02 5.334e+02, threshold=4.883e+02, percent-clipped=1.0 2023-05-01 01:58:12,932 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6318, 3.9543, 2.9425, 2.2877, 2.6800, 2.5667, 4.3788, 3.4859], device='cuda:3'), covar=tensor([0.2802, 0.0589, 0.1689, 0.2440, 0.2242, 0.1898, 0.0409, 0.1107], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0268, 0.0303, 0.0306, 0.0296, 0.0253, 0.0292, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 01:58:14,504 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:58:42,798 INFO [train.py:904] (3/8) Epoch 20, batch 4450, loss[loss=0.2038, simple_loss=0.2907, pruned_loss=0.05844, over 16903.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2831, pruned_loss=0.05233, over 3208362.63 frames. ], batch size: 109, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:59:55,983 INFO [train.py:904] (3/8) Epoch 20, batch 4500, loss[loss=0.1864, simple_loss=0.2766, pruned_loss=0.04811, over 16515.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2838, pruned_loss=0.05347, over 3199953.72 frames. ], batch size: 75, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:00:15,280 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0616, 5.2041, 5.4966, 5.4441, 5.5165, 5.1013, 5.1013, 4.7392], device='cuda:3'), covar=tensor([0.0264, 0.0399, 0.0252, 0.0326, 0.0303, 0.0311, 0.0815, 0.0431], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0440, 0.0423, 0.0398, 0.0470, 0.0449, 0.0538, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 02:00:25,482 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 02:00:30,610 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 20, batch 4550, loss[loss=0.2305, simple_loss=0.3136, pruned_loss=0.07366, over 16903.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2851, pruned_loss=0.05475, over 3210905.55 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:01:27,217 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197415.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:01:56,147 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197435.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:02:18,308 INFO [train.py:904] (3/8) Epoch 20, batch 4600, loss[loss=0.2116, simple_loss=0.2985, pruned_loss=0.06236, over 15501.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2858, pruned_loss=0.05489, over 3217000.22 frames. ], batch size: 190, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:02:18,859 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9354, 3.0273, 3.4240, 1.9264, 2.8367, 2.0196, 3.4782, 3.2418], device='cuda:3'), covar=tensor([0.0226, 0.0858, 0.0532, 0.2094, 0.0828, 0.1084, 0.0549, 0.0962], device='cuda:3'), in_proj_covar=tensor([0.0153, 0.0162, 0.0164, 0.0150, 0.0142, 0.0128, 0.0142, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:02:52,206 INFO [optim.py:368] (3/8) 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 02:02:52,804 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197476.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:03:03,003 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:24,547 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0650, 5.0958, 4.9490, 4.5996, 4.6462, 5.0510, 4.7883, 4.6877], device='cuda:3'), covar=tensor([0.0421, 0.0270, 0.0190, 0.0225, 0.0677, 0.0230, 0.0302, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0287, 0.0412, 0.0334, 0.0331, 0.0346, 0.0380, 0.0231, 0.0399], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:03:30,131 INFO [train.py:904] (3/8) Epoch 20, batch 4650, loss[loss=0.1832, simple_loss=0.2731, pruned_loss=0.04664, over 16627.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2844, pruned_loss=0.05463, over 3221925.34 frames. ], batch size: 62, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:03:41,189 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:49,590 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197516.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:20,831 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:32,446 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 4700, loss[loss=0.1988, simple_loss=0.2765, pruned_loss=0.06058, over 11421.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2817, pruned_loss=0.05353, over 3207548.81 frames. ], batch size: 248, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:04:52,429 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:52,977 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 02:05:18,233 INFO [optim.py:368] (3/8) 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 02:05:19,846 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197577.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:05:31,366 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:55,397 INFO [train.py:904] (3/8) Epoch 20, batch 4750, loss[loss=0.1685, simple_loss=0.2586, pruned_loss=0.03922, over 16450.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2781, pruned_loss=0.05186, over 3196197.02 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:06:01,836 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:06:11,329 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-05-01 02:07:01,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7128, 3.6377, 4.1565, 1.8420, 4.3962, 4.3703, 3.0258, 3.1493], device='cuda:3'), covar=tensor([0.0769, 0.0279, 0.0170, 0.1301, 0.0055, 0.0106, 0.0422, 0.0464], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0110, 0.0099, 0.0140, 0.0080, 0.0125, 0.0129, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:07:08,961 INFO [train.py:904] (3/8) Epoch 20, batch 4800, loss[loss=0.1861, simple_loss=0.2795, pruned_loss=0.04633, over 16647.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.274, pruned_loss=0.04971, over 3197412.07 frames. ], batch size: 134, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:07:45,093 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 20, batch 4850, loss[loss=0.2233, simple_loss=0.3004, pruned_loss=0.0731, over 12101.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2747, pruned_loss=0.0484, over 3196407.63 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:08:46,873 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-05-01 02:08:57,109 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4050, 3.3621, 3.4421, 3.5270, 3.5646, 3.2751, 3.5328, 3.6114], device='cuda:3'), covar=tensor([0.1067, 0.0913, 0.0929, 0.0597, 0.0571, 0.2319, 0.0906, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0620, 0.0767, 0.0891, 0.0782, 0.0585, 0.0613, 0.0628, 0.0727], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:09:40,078 INFO [train.py:904] (3/8) Epoch 20, batch 4900, loss[loss=0.1736, simple_loss=0.2643, pruned_loss=0.04145, over 16896.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2742, pruned_loss=0.04735, over 3178738.02 frames. ], batch size: 109, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:09:55,941 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7130, 2.7165, 2.2026, 2.6314, 3.1589, 2.8537, 3.1719, 3.3563], device='cuda:3'), covar=tensor([0.0088, 0.0417, 0.0572, 0.0449, 0.0245, 0.0361, 0.0260, 0.0267], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0229, 0.0222, 0.0222, 0.0232, 0.0231, 0.0232, 0.0228], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:10:08,662 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197771.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:10:15,934 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.929e+02 2.192e+02 2.700e+02 5.551e+02, threshold=4.384e+02, percent-clipped=1.0 2023-05-01 02:10:48,318 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0854, 4.9217, 5.0976, 5.3058, 5.5394, 4.8950, 5.4916, 5.5281], device='cuda:3'), covar=tensor([0.1900, 0.1279, 0.1711, 0.0790, 0.0513, 0.0772, 0.0585, 0.0577], device='cuda:3'), in_proj_covar=tensor([0.0621, 0.0768, 0.0894, 0.0784, 0.0584, 0.0615, 0.0629, 0.0728], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:10:52,823 INFO [train.py:904] (3/8) Epoch 20, batch 4950, loss[loss=0.1765, simple_loss=0.2733, pruned_loss=0.03983, over 16697.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2729, pruned_loss=0.04638, over 3192631.22 frames. ], batch size: 124, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:04,523 INFO [train.py:904] (3/8) Epoch 20, batch 5000, loss[loss=0.1864, simple_loss=0.287, pruned_loss=0.04293, over 16230.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2752, pruned_loss=0.04704, over 3191269.64 frames. ], batch size: 165, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:33,415 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197872.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:12:38,856 INFO [optim.py:368] (3/8) 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 02:13:01,272 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197892.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:13,654 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 5050, loss[loss=0.1607, simple_loss=0.2588, pruned_loss=0.03133, over 16935.00 frames. ], tot_loss[loss=0.185, simple_loss=0.276, pruned_loss=0.04706, over 3193788.92 frames. ], batch size: 96, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:13:35,054 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1319, 4.0373, 4.1907, 4.3418, 4.4965, 4.0737, 4.4259, 4.5106], device='cuda:3'), covar=tensor([0.1730, 0.1104, 0.1400, 0.0699, 0.0456, 0.1228, 0.0634, 0.0580], device='cuda:3'), in_proj_covar=tensor([0.0625, 0.0773, 0.0900, 0.0790, 0.0588, 0.0618, 0.0633, 0.0732], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:13:38,939 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 02:14:25,041 INFO [train.py:904] (3/8) Epoch 20, batch 5100, loss[loss=0.1637, simple_loss=0.2493, pruned_loss=0.039, over 17134.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2737, pruned_loss=0.04617, over 3198351.25 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:27,962 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197953.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:15:00,608 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 1.957e+02 2.274e+02 2.606e+02 4.454e+02, threshold=4.549e+02, percent-clipped=0.0 2023-05-01 02:15:41,431 INFO [train.py:904] (3/8) Epoch 20, batch 5150, loss[loss=0.1841, simple_loss=0.2794, pruned_loss=0.04444, over 16919.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2736, pruned_loss=0.04561, over 3206724.04 frames. ], batch size: 109, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:16:52,361 INFO [train.py:904] (3/8) Epoch 20, batch 5200, loss[loss=0.1812, simple_loss=0.2642, pruned_loss=0.04913, over 16731.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2724, pruned_loss=0.04505, over 3206044.57 frames. ], batch size: 124, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:17:06,550 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 02:17:18,559 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5357, 4.5968, 4.9249, 4.8630, 4.8792, 4.5947, 4.5456, 4.4239], device='cuda:3'), covar=tensor([0.0289, 0.0513, 0.0310, 0.0368, 0.0401, 0.0352, 0.0840, 0.0459], device='cuda:3'), in_proj_covar=tensor([0.0389, 0.0431, 0.0415, 0.0388, 0.0461, 0.0439, 0.0528, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 02:17:19,627 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198071.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:17:27,747 INFO [optim.py:368] (3/8) 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 02:17:34,785 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198081.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:18:03,836 INFO [train.py:904] (3/8) Epoch 20, batch 5250, loss[loss=0.1709, simple_loss=0.2631, pruned_loss=0.03939, over 16826.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.27, pruned_loss=0.04474, over 3216167.20 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:18:28,863 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198119.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:18:48,727 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 02:19:01,585 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 5300, loss[loss=0.1485, simple_loss=0.2391, pruned_loss=0.02898, over 16848.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2664, pruned_loss=0.04328, over 3219221.66 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:19:32,121 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.22 vs. limit=5.0 2023-05-01 02:19:45,037 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198172.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:19:49,806 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 1.825e+02 2.153e+02 2.534e+02 4.591e+02, threshold=4.307e+02, percent-clipped=1.0 2023-05-01 02:20:27,528 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 5350, loss[loss=0.1693, simple_loss=0.2647, pruned_loss=0.03691, over 16685.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.265, pruned_loss=0.04278, over 3218419.54 frames. ], batch size: 62, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:20:54,786 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198220.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:59,137 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6079, 3.8633, 4.0163, 3.9620, 3.9486, 3.7859, 3.4974, 3.7556], device='cuda:3'), covar=tensor([0.0568, 0.0814, 0.0610, 0.0683, 0.0832, 0.0647, 0.1535, 0.0651], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0434, 0.0418, 0.0390, 0.0465, 0.0442, 0.0531, 0.0354], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 02:21:35,322 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198248.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:37,093 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 5400, loss[loss=0.1738, simple_loss=0.2656, pruned_loss=0.04102, over 16581.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2679, pruned_loss=0.04382, over 3201248.61 frames. ], batch size: 62, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:22:15,985 INFO [optim.py:368] (3/8) 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 02:22:57,923 INFO [train.py:904] (3/8) Epoch 20, batch 5450, loss[loss=0.1819, simple_loss=0.272, pruned_loss=0.04592, over 11931.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2712, pruned_loss=0.0457, over 3173238.08 frames. ], batch size: 250, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:23:04,395 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4869, 3.5863, 2.7737, 2.1956, 2.3542, 2.3356, 3.7569, 3.2999], device='cuda:3'), covar=tensor([0.2946, 0.0636, 0.1759, 0.2639, 0.2613, 0.2052, 0.0470, 0.1173], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0268, 0.0302, 0.0306, 0.0294, 0.0252, 0.0292, 0.0331], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:23:12,776 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 02:23:18,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6593, 4.6810, 5.0078, 4.9805, 5.0318, 4.6866, 4.6328, 4.5102], device='cuda:3'), covar=tensor([0.0285, 0.0523, 0.0354, 0.0389, 0.0419, 0.0381, 0.0909, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0432, 0.0416, 0.0388, 0.0463, 0.0441, 0.0529, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 02:23:37,600 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3548, 3.3582, 3.8860, 1.8299, 4.0428, 4.0751, 2.9962, 2.8747], device='cuda:3'), covar=tensor([0.0905, 0.0306, 0.0176, 0.1258, 0.0062, 0.0136, 0.0429, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0110, 0.0098, 0.0140, 0.0080, 0.0124, 0.0129, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:24:01,050 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3506, 3.2474, 3.3030, 3.4773, 3.5004, 3.2658, 3.4610, 3.5626], device='cuda:3'), covar=tensor([0.1333, 0.1195, 0.1426, 0.0806, 0.0822, 0.2222, 0.1103, 0.0927], device='cuda:3'), in_proj_covar=tensor([0.0622, 0.0767, 0.0893, 0.0779, 0.0586, 0.0614, 0.0626, 0.0726], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:24:14,759 INFO [train.py:904] (3/8) Epoch 20, batch 5500, loss[loss=0.1927, simple_loss=0.2883, pruned_loss=0.0485, over 16702.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2779, pruned_loss=0.04981, over 3148671.95 frames. ], batch size: 89, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:24:38,062 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1574, 4.2346, 4.0642, 3.8022, 3.7569, 4.1706, 3.8478, 3.9353], device='cuda:3'), covar=tensor([0.0683, 0.0692, 0.0303, 0.0301, 0.0834, 0.0566, 0.0883, 0.0596], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0421, 0.0339, 0.0334, 0.0350, 0.0390, 0.0233, 0.0405], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:24:51,691 INFO [optim.py:368] (3/8) 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 02:25:34,184 INFO [train.py:904] (3/8) Epoch 20, batch 5550, loss[loss=0.2768, simple_loss=0.3351, pruned_loss=0.1092, over 10934.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2849, pruned_loss=0.05477, over 3121026.78 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:25:51,840 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 02:26:30,782 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198437.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:26:53,769 INFO [train.py:904] (3/8) Epoch 20, batch 5600, loss[loss=0.234, simple_loss=0.3288, pruned_loss=0.0696, over 16202.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2897, pruned_loss=0.05845, over 3102501.78 frames. ], batch size: 165, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:27:23,288 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 02:27:34,780 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.278e+02 3.283e+02 3.655e+02 4.380e+02 7.107e+02, threshold=7.309e+02, percent-clipped=1.0 2023-05-01 02:28:14,140 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198499.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:28:18,605 INFO [train.py:904] (3/8) Epoch 20, batch 5650, loss[loss=0.2571, simple_loss=0.3217, pruned_loss=0.09628, over 11392.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2947, pruned_loss=0.06326, over 3055524.06 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:32,494 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198548.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:29:36,984 INFO [train.py:904] (3/8) Epoch 20, batch 5700, loss[loss=0.262, simple_loss=0.323, pruned_loss=0.1005, over 11041.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2958, pruned_loss=0.06444, over 3060854.58 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:50,275 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:14,598 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.943e+02 3.352e+02 4.045e+02 5.075e+02 1.137e+03, threshold=8.090e+02, percent-clipped=5.0 2023-05-01 02:30:21,730 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3766, 3.2502, 3.5546, 1.9592, 3.7422, 3.7756, 2.8741, 2.8077], device='cuda:3'), covar=tensor([0.0839, 0.0274, 0.0230, 0.1134, 0.0088, 0.0173, 0.0469, 0.0482], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0140, 0.0080, 0.0124, 0.0128, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:30:47,281 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 5750, loss[loss=0.2106, simple_loss=0.2948, pruned_loss=0.06317, over 16690.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2998, pruned_loss=0.06699, over 3028842.68 frames. ], batch size: 134, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:31:02,616 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2651, 3.2743, 2.0532, 3.5979, 2.5129, 3.6091, 2.1032, 2.6400], device='cuda:3'), covar=tensor([0.0319, 0.0410, 0.1649, 0.0199, 0.0791, 0.0595, 0.1659, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:31:08,969 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 02:31:17,266 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:32:08,104 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 02:32:16,953 INFO [train.py:904] (3/8) Epoch 20, batch 5800, loss[loss=0.2064, simple_loss=0.2963, pruned_loss=0.05823, over 17035.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2986, pruned_loss=0.06485, over 3054338.22 frames. ], batch size: 53, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:32:53,794 INFO [optim.py:368] (3/8) 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 02:32:54,355 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:33:34,960 INFO [train.py:904] (3/8) Epoch 20, batch 5850, loss[loss=0.1957, simple_loss=0.282, pruned_loss=0.05475, over 11729.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2968, pruned_loss=0.06358, over 3039821.12 frames. ], batch size: 247, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:34:05,310 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4656, 4.6199, 4.7554, 4.5735, 4.6175, 5.1694, 4.6666, 4.3819], device='cuda:3'), covar=tensor([0.1407, 0.1932, 0.2335, 0.1971, 0.2497, 0.0995, 0.1787, 0.2703], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0581, 0.0635, 0.0481, 0.0641, 0.0670, 0.0498, 0.0649], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:34:27,423 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0753, 4.0911, 4.4268, 4.3995, 4.4004, 4.1269, 4.1290, 4.1084], device='cuda:3'), covar=tensor([0.0354, 0.0622, 0.0411, 0.0431, 0.0463, 0.0464, 0.0887, 0.0545], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0442, 0.0426, 0.0396, 0.0474, 0.0449, 0.0538, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 02:34:30,873 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198737.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:34:53,518 INFO [train.py:904] (3/8) Epoch 20, batch 5900, loss[loss=0.1933, simple_loss=0.2772, pruned_loss=0.05472, over 16871.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2962, pruned_loss=0.06332, over 3051676.14 frames. ], batch size: 96, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:35:24,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8182, 3.7063, 3.8859, 3.9940, 4.0685, 3.6809, 4.0191, 4.1101], device='cuda:3'), covar=tensor([0.1607, 0.1169, 0.1239, 0.0706, 0.0661, 0.1930, 0.0865, 0.0697], device='cuda:3'), in_proj_covar=tensor([0.0619, 0.0763, 0.0886, 0.0775, 0.0584, 0.0611, 0.0626, 0.0721], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:35:26,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7763, 4.1275, 3.1095, 2.2956, 2.8372, 2.5925, 4.3750, 3.6638], device='cuda:3'), covar=tensor([0.2788, 0.0511, 0.1653, 0.2739, 0.2463, 0.1861, 0.0440, 0.1225], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0267, 0.0302, 0.0305, 0.0294, 0.0252, 0.0292, 0.0331], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:35:27,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4106, 3.3979, 1.8983, 3.8309, 2.5292, 3.7753, 2.1677, 2.7724], device='cuda:3'), covar=tensor([0.0300, 0.0411, 0.1849, 0.0324, 0.0930, 0.0641, 0.1550, 0.0801], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0158, 0.0175, 0.0213, 0.0201, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:35:34,248 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:36:14,567 INFO [train.py:904] (3/8) Epoch 20, batch 5950, loss[loss=0.2013, simple_loss=0.289, pruned_loss=0.05678, over 16706.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2962, pruned_loss=0.06147, over 3074156.16 frames. ], batch size: 134, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:02,250 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3956, 4.5158, 4.6062, 4.4557, 4.5228, 5.0346, 4.5053, 4.2699], device='cuda:3'), covar=tensor([0.1536, 0.2042, 0.2455, 0.2156, 0.2705, 0.1108, 0.1879, 0.2651], device='cuda:3'), in_proj_covar=tensor([0.0404, 0.0584, 0.0639, 0.0484, 0.0644, 0.0673, 0.0502, 0.0653], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:37:31,022 INFO [train.py:904] (3/8) Epoch 20, batch 6000, loss[loss=0.1941, simple_loss=0.2924, pruned_loss=0.04793, over 16844.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2947, pruned_loss=0.06062, over 3090521.02 frames. ], batch size: 102, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,022 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 02:37:41,834 INFO [train.py:938] (3/8) Epoch 20, validation: loss=0.1516, simple_loss=0.2644, pruned_loss=0.01942, over 944034.00 frames. 2023-05-01 02:37:41,835 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 02:37:42,288 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:46,534 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198855.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:58,509 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.53 vs. limit=2.0 2023-05-01 02:38:17,162 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 2.764e+02 3.257e+02 3.996e+02 6.001e+02, threshold=6.515e+02, percent-clipped=0.0 2023-05-01 02:38:28,085 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 6050, loss[loss=0.1868, simple_loss=0.289, pruned_loss=0.04231, over 16955.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2937, pruned_loss=0.05989, over 3109400.47 frames. ], batch size: 96, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:39:14,661 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198913.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:36,246 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198927.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:47,165 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8320, 1.4972, 1.6900, 1.7274, 1.8665, 1.9483, 1.6617, 1.8458], device='cuda:3'), covar=tensor([0.0228, 0.0359, 0.0185, 0.0285, 0.0242, 0.0155, 0.0357, 0.0124], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0191, 0.0175, 0.0179, 0.0191, 0.0149, 0.0192, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:40:05,097 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198944.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:17,597 INFO [train.py:904] (3/8) Epoch 20, batch 6100, loss[loss=0.1859, simple_loss=0.2851, pruned_loss=0.04338, over 16855.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2931, pruned_loss=0.0591, over 3109074.38 frames. ], batch size: 90, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:40:46,238 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:53,005 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-01 02:40:53,272 INFO [optim.py:368] (3/8) 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 02:41:12,545 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:41:15,531 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 02:41:21,254 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8073, 2.7333, 2.5792, 1.9230, 2.5959, 2.6723, 2.5564, 1.8897], device='cuda:3'), covar=tensor([0.0441, 0.0072, 0.0080, 0.0328, 0.0116, 0.0120, 0.0109, 0.0388], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0132, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:41:32,342 INFO [train.py:904] (3/8) Epoch 20, batch 6150, loss[loss=0.1854, simple_loss=0.2785, pruned_loss=0.04613, over 17215.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2912, pruned_loss=0.05846, over 3091310.05 frames. ], batch size: 45, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:42:49,580 INFO [train.py:904] (3/8) Epoch 20, batch 6200, loss[loss=0.2172, simple_loss=0.2995, pruned_loss=0.06746, over 15330.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2897, pruned_loss=0.05847, over 3072879.93 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:43:28,024 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.934e+02 3.552e+02 4.446e+02 1.093e+03, threshold=7.104e+02, percent-clipped=2.0 2023-05-01 02:44:06,519 INFO [train.py:904] (3/8) Epoch 20, batch 6250, loss[loss=0.208, simple_loss=0.2977, pruned_loss=0.05913, over 16319.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2891, pruned_loss=0.05785, over 3100326.12 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:21,381 INFO [train.py:904] (3/8) Epoch 20, batch 6300, loss[loss=0.1863, simple_loss=0.2787, pruned_loss=0.04693, over 16339.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2882, pruned_loss=0.05679, over 3108478.21 frames. ], batch size: 68, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:24,401 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1235, 5.5915, 5.7793, 5.4456, 5.5512, 6.1204, 5.5790, 5.3800], device='cuda:3'), covar=tensor([0.0903, 0.1746, 0.2329, 0.2144, 0.2565, 0.1057, 0.1682, 0.2355], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0584, 0.0643, 0.0486, 0.0644, 0.0674, 0.0502, 0.0652], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:45:26,464 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199155.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:45:59,366 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:00,086 INFO [optim.py:368] (3/8) 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 02:46:38,966 INFO [train.py:904] (3/8) Epoch 20, batch 6350, loss[loss=0.2166, simple_loss=0.3017, pruned_loss=0.06574, over 16387.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2896, pruned_loss=0.05881, over 3080627.03 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:46:40,593 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:48,061 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199208.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:00,909 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5566, 3.5871, 3.4006, 3.1199, 3.2061, 3.5201, 3.3424, 3.3521], device='cuda:3'), covar=tensor([0.0621, 0.0677, 0.0390, 0.0319, 0.0579, 0.0532, 0.1190, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0283, 0.0409, 0.0329, 0.0324, 0.0339, 0.0377, 0.0227, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:47:30,846 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199237.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:47:34,235 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 6400, loss[loss=0.1949, simple_loss=0.2799, pruned_loss=0.05493, over 16636.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2903, pruned_loss=0.06041, over 3071364.92 frames. ], batch size: 62, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:48:21,658 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:48:21,981 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 02:48:29,165 INFO [optim.py:368] (3/8) 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 02:48:39,477 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199283.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:49:07,111 INFO [train.py:904] (3/8) Epoch 20, batch 6450, loss[loss=0.1985, simple_loss=0.2843, pruned_loss=0.05639, over 15410.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2901, pruned_loss=0.05956, over 3070058.32 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:49:13,062 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 02:49:24,571 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.10 vs. limit=5.0 2023-05-01 02:49:26,803 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9268, 2.1344, 2.4432, 3.1682, 2.2169, 2.3574, 2.3335, 2.2428], device='cuda:3'), covar=tensor([0.1281, 0.3387, 0.2396, 0.0677, 0.3846, 0.2246, 0.2940, 0.3407], device='cuda:3'), in_proj_covar=tensor([0.0395, 0.0438, 0.0361, 0.0321, 0.0432, 0.0505, 0.0407, 0.0513], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:49:33,333 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:49:54,881 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5519, 1.5219, 2.1646, 2.3936, 2.4653, 2.7011, 1.6313, 2.6693], device='cuda:3'), covar=tensor([0.0171, 0.0555, 0.0253, 0.0285, 0.0265, 0.0172, 0.0600, 0.0132], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0192, 0.0177, 0.0182, 0.0193, 0.0150, 0.0193, 0.0145], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:50:24,669 INFO [train.py:904] (3/8) Epoch 20, batch 6500, loss[loss=0.2201, simple_loss=0.3039, pruned_loss=0.06813, over 16854.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2876, pruned_loss=0.05839, over 3095037.54 frames. ], batch size: 83, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:51:02,009 INFO [optim.py:368] (3/8) 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 02:51:22,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9101, 4.1831, 3.9854, 4.0205, 3.7301, 3.7902, 3.8224, 4.1539], device='cuda:3'), covar=tensor([0.1163, 0.0914, 0.1098, 0.0907, 0.0836, 0.1588, 0.1048, 0.1081], device='cuda:3'), in_proj_covar=tensor([0.0641, 0.0780, 0.0645, 0.0589, 0.0495, 0.0503, 0.0657, 0.0605], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 02:51:28,331 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.4829, 2.6113, 2.3626, 4.1100, 2.6527, 3.9797, 1.4926, 2.8410], device='cuda:3'), covar=tensor([0.1534, 0.0824, 0.1330, 0.0150, 0.0223, 0.0399, 0.1828, 0.0897], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0173, 0.0193, 0.0184, 0.0206, 0.0213, 0.0199, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:51:39,196 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:51:41,756 INFO [train.py:904] (3/8) Epoch 20, batch 6550, loss[loss=0.1995, simple_loss=0.2984, pruned_loss=0.05026, over 15446.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2899, pruned_loss=0.0587, over 3110091.11 frames. ], batch size: 191, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:52:28,140 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 02:52:56,160 INFO [train.py:904] (3/8) Epoch 20, batch 6600, loss[loss=0.201, simple_loss=0.2874, pruned_loss=0.05725, over 16656.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2918, pruned_loss=0.05912, over 3117689.86 frames. ], batch size: 57, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:53:05,819 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:53:10,048 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199461.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:53:33,375 INFO [optim.py:368] (3/8) 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 02:53:59,288 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199494.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:54:11,565 INFO [train.py:904] (3/8) Epoch 20, batch 6650, loss[loss=0.1835, simple_loss=0.2787, pruned_loss=0.0442, over 17199.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2917, pruned_loss=0.05946, over 3130305.83 frames. ], batch size: 46, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:54:20,056 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199508.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:54:37,196 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199519.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:54:56,458 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199532.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:00,501 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:06,214 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199539.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:13,570 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 02:55:25,177 INFO [train.py:904] (3/8) Epoch 20, batch 6700, loss[loss=0.2496, simple_loss=0.3079, pruned_loss=0.09566, over 11676.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2905, pruned_loss=0.05984, over 3125240.25 frames. ], batch size: 249, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:55:30,686 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199555.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:55:32,561 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199556.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:34,433 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 02:56:02,758 INFO [optim.py:368] (3/8) 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 02:56:13,319 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:31,474 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199596.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:56:38,604 INFO [train.py:904] (3/8) Epoch 20, batch 6750, loss[loss=0.2586, simple_loss=0.327, pruned_loss=0.09514, over 11899.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2894, pruned_loss=0.05938, over 3133663.56 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:56:39,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5648, 4.7421, 4.8446, 4.6622, 4.7276, 5.2607, 4.7421, 4.4750], device='cuda:3'), covar=tensor([0.1245, 0.1777, 0.2047, 0.1778, 0.2227, 0.0967, 0.1525, 0.2306], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0579, 0.0638, 0.0481, 0.0636, 0.0669, 0.0498, 0.0647], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 02:57:14,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1860, 2.9970, 3.2840, 1.7495, 3.4770, 3.5300, 2.7549, 2.5829], device='cuda:3'), covar=tensor([0.0860, 0.0277, 0.0218, 0.1225, 0.0088, 0.0189, 0.0455, 0.0525], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0097, 0.0139, 0.0079, 0.0123, 0.0128, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 02:57:20,690 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:57:53,090 INFO [train.py:904] (3/8) Epoch 20, batch 6800, loss[loss=0.2256, simple_loss=0.2943, pruned_loss=0.07841, over 11564.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2894, pruned_loss=0.05923, over 3133078.35 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:58:29,264 INFO [optim.py:368] (3/8) 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 02:58:38,287 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 02:59:05,436 INFO [train.py:904] (3/8) Epoch 20, batch 6850, loss[loss=0.1764, simple_loss=0.2818, pruned_loss=0.03544, over 16690.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2909, pruned_loss=0.05952, over 3138058.94 frames. ], batch size: 57, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 03:00:20,622 INFO [train.py:904] (3/8) Epoch 20, batch 6900, loss[loss=0.2255, simple_loss=0.3124, pruned_loss=0.06932, over 16185.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2937, pruned_loss=0.05916, over 3150464.50 frames. ], batch size: 165, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:00:26,872 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199756.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:00:37,249 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7919, 3.7601, 3.8673, 3.6639, 3.8403, 4.2520, 3.9444, 3.6218], device='cuda:3'), covar=tensor([0.2119, 0.2345, 0.2620, 0.2478, 0.2606, 0.1879, 0.1584, 0.2629], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0577, 0.0636, 0.0479, 0.0636, 0.0666, 0.0495, 0.0644], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:00:59,795 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 20, batch 6950, loss[loss=0.2781, simple_loss=0.3301, pruned_loss=0.1131, over 11284.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2955, pruned_loss=0.06145, over 3106763.09 frames. ], batch size: 246, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:01:41,356 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0018, 3.2045, 3.1699, 2.0705, 2.9829, 3.1856, 3.0282, 1.9515], device='cuda:3'), covar=tensor([0.0545, 0.0061, 0.0067, 0.0431, 0.0113, 0.0123, 0.0106, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0133, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:01:54,387 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199814.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:02:22,394 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:02:47,650 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199850.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:02:49,629 INFO [train.py:904] (3/8) Epoch 20, batch 7000, loss[loss=0.1914, simple_loss=0.2894, pruned_loss=0.04672, over 16709.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2956, pruned_loss=0.06117, over 3093935.70 frames. ], batch size: 134, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:03:29,995 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.829e+02 3.376e+02 4.341e+02 9.860e+02, threshold=6.751e+02, percent-clipped=7.0 2023-05-01 03:03:33,438 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:03:49,880 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199891.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:04:07,200 INFO [train.py:904] (3/8) Epoch 20, batch 7050, loss[loss=0.1925, simple_loss=0.2795, pruned_loss=0.05277, over 16615.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2959, pruned_loss=0.06081, over 3093123.09 frames. ], batch size: 57, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:04:18,601 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:05:25,771 INFO [train.py:904] (3/8) Epoch 20, batch 7100, loss[loss=0.207, simple_loss=0.293, pruned_loss=0.06051, over 16571.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2944, pruned_loss=0.06046, over 3104807.55 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:05:54,792 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199970.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:06:05,964 INFO [optim.py:368] (3/8) 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 03:06:45,333 INFO [train.py:904] (3/8) Epoch 20, batch 7150, loss[loss=0.1921, simple_loss=0.2795, pruned_loss=0.05237, over 16661.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2933, pruned_loss=0.06071, over 3097684.48 frames. ], batch size: 57, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:06:59,715 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 03:07:59,653 INFO [train.py:904] (3/8) Epoch 20, batch 7200, loss[loss=0.1998, simple_loss=0.2929, pruned_loss=0.05335, over 16494.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.291, pruned_loss=0.05907, over 3087574.65 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:08:06,667 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200056.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:08:40,490 INFO [optim.py:368] (3/8) 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 03:09:04,173 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4275, 4.0445, 3.9586, 2.6890, 3.5738, 3.9991, 3.6052, 2.2828], device='cuda:3'), covar=tensor([0.0512, 0.0039, 0.0048, 0.0393, 0.0109, 0.0098, 0.0087, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0132, 0.0095, 0.0107, 0.0092, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:09:19,964 INFO [train.py:904] (3/8) Epoch 20, batch 7250, loss[loss=0.1812, simple_loss=0.2687, pruned_loss=0.04688, over 16720.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2891, pruned_loss=0.05833, over 3076980.20 frames. ], batch size: 83, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:09:23,330 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200104.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:09:37,520 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:10:31,970 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 7300, loss[loss=0.2245, simple_loss=0.292, pruned_loss=0.07845, over 11647.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2887, pruned_loss=0.05863, over 3061050.89 frames. ], batch size: 250, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:10:50,981 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200162.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:11:11,462 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 03:11:14,655 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 03:11:24,459 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5004, 5.4259, 5.3043, 4.9374, 4.9559, 5.3598, 5.3015, 4.9725], device='cuda:3'), covar=tensor([0.0499, 0.0313, 0.0253, 0.0264, 0.0956, 0.0329, 0.0235, 0.0604], device='cuda:3'), in_proj_covar=tensor([0.0279, 0.0403, 0.0324, 0.0320, 0.0334, 0.0373, 0.0224, 0.0388], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:11:37,022 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200191.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:11:47,791 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 7350, loss[loss=0.2107, simple_loss=0.3002, pruned_loss=0.06059, over 16830.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2897, pruned_loss=0.05943, over 3060601.50 frames. ], batch size: 116, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:12:40,122 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4793, 2.5748, 2.1598, 2.2731, 2.9500, 2.5094, 3.0701, 3.1084], device='cuda:3'), covar=tensor([0.0107, 0.0383, 0.0519, 0.0423, 0.0238, 0.0387, 0.0187, 0.0248], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0227, 0.0220, 0.0220, 0.0229, 0.0228, 0.0227, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:12:51,300 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200238.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:12:52,207 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200239.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:13:12,412 INFO [train.py:904] (3/8) Epoch 20, batch 7400, loss[loss=0.2214, simple_loss=0.3149, pruned_loss=0.06397, over 16569.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2909, pruned_loss=0.05968, over 3069896.13 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:13:25,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4251, 3.2317, 3.7083, 1.9219, 3.8005, 3.9007, 2.9980, 2.7606], device='cuda:3'), covar=tensor([0.0768, 0.0274, 0.0148, 0.1165, 0.0082, 0.0149, 0.0395, 0.0516], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0096, 0.0139, 0.0079, 0.0122, 0.0127, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 03:13:33,119 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:13:35,484 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-01 03:13:52,749 INFO [optim.py:368] (3/8) 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 03:13:55,420 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 7450, loss[loss=0.2003, simple_loss=0.2921, pruned_loss=0.05425, over 16239.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2917, pruned_loss=0.06075, over 3057065.94 frames. ], batch size: 165, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:15:04,009 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3807, 3.4512, 1.8871, 3.7818, 2.5610, 3.7407, 2.1127, 2.6671], device='cuda:3'), covar=tensor([0.0295, 0.0379, 0.1764, 0.0261, 0.0828, 0.0598, 0.1628, 0.0823], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0194, 0.0157, 0.0175, 0.0214, 0.0201, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 03:15:36,192 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200340.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:15:54,169 INFO [train.py:904] (3/8) Epoch 20, batch 7500, loss[loss=0.2053, simple_loss=0.2909, pruned_loss=0.0598, over 15262.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2916, pruned_loss=0.06002, over 3053098.08 frames. ], batch size: 191, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:16:34,099 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 20, batch 7550, loss[loss=0.1821, simple_loss=0.275, pruned_loss=0.04462, over 16167.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2905, pruned_loss=0.05981, over 3050086.93 frames. ], batch size: 165, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:18:09,835 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7603, 1.3558, 1.6994, 1.6082, 1.7380, 1.8746, 1.5627, 1.7309], device='cuda:3'), covar=tensor([0.0236, 0.0356, 0.0195, 0.0265, 0.0252, 0.0162, 0.0403, 0.0124], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0188, 0.0172, 0.0177, 0.0189, 0.0148, 0.0190, 0.0142], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:18:26,295 INFO [train.py:904] (3/8) Epoch 20, batch 7600, loss[loss=0.2023, simple_loss=0.2805, pruned_loss=0.06212, over 17264.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2893, pruned_loss=0.05973, over 3053329.54 frames. ], batch size: 52, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:19:06,493 INFO [optim.py:368] (3/8) 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 03:19:44,054 INFO [train.py:904] (3/8) Epoch 20, batch 7650, loss[loss=0.1926, simple_loss=0.2806, pruned_loss=0.05235, over 16442.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2899, pruned_loss=0.0604, over 3052191.75 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:20:33,300 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200533.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 03:21:02,176 INFO [train.py:904] (3/8) Epoch 20, batch 7700, loss[loss=0.1747, simple_loss=0.2676, pruned_loss=0.04091, over 16372.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.289, pruned_loss=0.06035, over 3061764.50 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:21:22,803 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 20, batch 7750, loss[loss=0.2538, simple_loss=0.3127, pruned_loss=0.09748, over 11125.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2898, pruned_loss=0.06012, over 3066959.96 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:22:29,481 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 03:22:38,649 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200613.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:08,923 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 03:23:11,917 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200635.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:23:26,531 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 03:23:28,167 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:36,961 INFO [train.py:904] (3/8) Epoch 20, batch 7800, loss[loss=0.2469, simple_loss=0.3084, pruned_loss=0.09273, over 11409.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2906, pruned_loss=0.06066, over 3067670.04 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:24:18,679 INFO [optim.py:368] (3/8) 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 03:24:19,768 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9192, 4.1866, 4.0116, 4.0593, 3.7142, 3.7626, 3.8475, 4.1903], device='cuda:3'), covar=tensor([0.1216, 0.0928, 0.1114, 0.0867, 0.0894, 0.1912, 0.1066, 0.1029], device='cuda:3'), in_proj_covar=tensor([0.0647, 0.0795, 0.0656, 0.0598, 0.0500, 0.0513, 0.0664, 0.0614], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:24:53,425 INFO [train.py:904] (3/8) Epoch 20, batch 7850, loss[loss=0.2048, simple_loss=0.2966, pruned_loss=0.05651, over 16289.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2918, pruned_loss=0.06074, over 3070308.17 frames. ], batch size: 165, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:25:00,355 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:26:08,719 INFO [train.py:904] (3/8) Epoch 20, batch 7900, loss[loss=0.2417, simple_loss=0.3059, pruned_loss=0.08878, over 11144.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2915, pruned_loss=0.0603, over 3077076.15 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:26:49,222 INFO [optim.py:368] (3/8) 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 03:26:52,643 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5651, 4.8392, 4.6100, 4.6449, 4.4041, 4.3671, 4.3181, 4.9167], device='cuda:3'), covar=tensor([0.1163, 0.0843, 0.1009, 0.0859, 0.0777, 0.1155, 0.1216, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0648, 0.0793, 0.0655, 0.0597, 0.0500, 0.0514, 0.0664, 0.0613], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:26:59,910 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9268, 2.3661, 2.2752, 2.8426, 1.8742, 3.1965, 1.7771, 2.7052], device='cuda:3'), covar=tensor([0.1236, 0.0648, 0.1119, 0.0216, 0.0145, 0.0407, 0.1543, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0174, 0.0194, 0.0186, 0.0208, 0.0214, 0.0200, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 03:27:27,129 INFO [train.py:904] (3/8) Epoch 20, batch 7950, loss[loss=0.2052, simple_loss=0.2874, pruned_loss=0.06148, over 16184.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2921, pruned_loss=0.06129, over 3057259.36 frames. ], batch size: 35, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:28:16,257 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200833.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:28:44,616 INFO [train.py:904] (3/8) Epoch 20, batch 8000, loss[loss=0.2973, simple_loss=0.3489, pruned_loss=0.1228, over 11985.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2924, pruned_loss=0.06166, over 3063016.01 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:29:24,818 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.028e+02 3.046e+02 3.376e+02 4.028e+02 7.694e+02, threshold=6.753e+02, percent-clipped=2.0 2023-05-01 03:29:28,342 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200881.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:29:59,854 INFO [train.py:904] (3/8) Epoch 20, batch 8050, loss[loss=0.2156, simple_loss=0.302, pruned_loss=0.0646, over 15285.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2926, pruned_loss=0.06109, over 3079506.65 frames. ], batch size: 191, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:30:21,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1940, 3.2477, 1.7949, 3.5559, 2.3894, 3.5741, 2.0468, 2.5718], device='cuda:3'), covar=tensor([0.0320, 0.0424, 0.1844, 0.0244, 0.0916, 0.0681, 0.1617, 0.0820], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0155, 0.0174, 0.0213, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 03:30:36,825 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 03:30:50,128 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:30:54,645 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 03:31:15,302 INFO [train.py:904] (3/8) Epoch 20, batch 8100, loss[loss=0.1878, simple_loss=0.2787, pruned_loss=0.04849, over 16899.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.292, pruned_loss=0.06079, over 3078959.71 frames. ], batch size: 109, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:31:57,146 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.634e+02 3.129e+02 3.819e+02 6.896e+02, threshold=6.259e+02, percent-clipped=1.0 2023-05-01 03:32:04,949 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200983.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:31,927 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:32,022 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7662, 4.6400, 4.5339, 3.0055, 3.9586, 4.5341, 3.8944, 2.6424], device='cuda:3'), covar=tensor([0.0494, 0.0029, 0.0035, 0.0356, 0.0090, 0.0087, 0.0084, 0.0395], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0081, 0.0081, 0.0133, 0.0096, 0.0108, 0.0092, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:32:33,316 INFO [train.py:904] (3/8) Epoch 20, batch 8150, loss[loss=0.1834, simple_loss=0.2602, pruned_loss=0.05329, over 17121.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2891, pruned_loss=0.05952, over 3090827.44 frames. ], batch size: 48, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:32:55,796 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9185, 4.9236, 4.7801, 4.1260, 4.8490, 1.7294, 4.5486, 4.4660], device='cuda:3'), covar=tensor([0.0095, 0.0076, 0.0172, 0.0342, 0.0082, 0.2638, 0.0125, 0.0218], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0147, 0.0189, 0.0173, 0.0167, 0.0201, 0.0180, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:33:24,012 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8652, 5.1584, 4.9313, 4.9134, 4.6589, 4.6109, 4.5705, 5.2437], device='cuda:3'), covar=tensor([0.1264, 0.0828, 0.1021, 0.0958, 0.0896, 0.1037, 0.1290, 0.0907], device='cuda:3'), in_proj_covar=tensor([0.0649, 0.0791, 0.0655, 0.0598, 0.0500, 0.0514, 0.0664, 0.0613], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:33:52,113 INFO [train.py:904] (3/8) Epoch 20, batch 8200, loss[loss=0.1842, simple_loss=0.2729, pruned_loss=0.04773, over 17015.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2867, pruned_loss=0.05885, over 3101214.14 frames. ], batch size: 50, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:34:24,230 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 03:35:08,319 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6105, 3.8462, 2.9944, 2.0978, 2.3350, 2.3822, 4.0639, 3.2551], device='cuda:3'), covar=tensor([0.3037, 0.0540, 0.1677, 0.3153, 0.2887, 0.2136, 0.0409, 0.1367], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0269, 0.0304, 0.0310, 0.0297, 0.0256, 0.0293, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:35:15,217 INFO [train.py:904] (3/8) Epoch 20, batch 8250, loss[loss=0.1868, simple_loss=0.2826, pruned_loss=0.04552, over 16659.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2857, pruned_loss=0.05619, over 3090003.68 frames. ], batch size: 134, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:36:05,989 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:36:10,541 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5460, 2.6922, 2.7718, 4.1189, 2.9314, 4.0631, 1.2752, 3.2211], device='cuda:3'), covar=tensor([0.1459, 0.0766, 0.0977, 0.0177, 0.0150, 0.0363, 0.1924, 0.0622], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0172, 0.0193, 0.0184, 0.0206, 0.0211, 0.0199, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 03:36:38,968 INFO [train.py:904] (3/8) Epoch 20, batch 8300, loss[loss=0.1641, simple_loss=0.2519, pruned_loss=0.03812, over 11625.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.283, pruned_loss=0.05314, over 3081215.14 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:37:08,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9015, 5.2435, 5.4403, 5.1887, 5.2426, 5.8126, 5.2811, 4.9734], device='cuda:3'), covar=tensor([0.0921, 0.1861, 0.1843, 0.1756, 0.2558, 0.0860, 0.1540, 0.2378], device='cuda:3'), in_proj_covar=tensor([0.0394, 0.0574, 0.0632, 0.0475, 0.0632, 0.0664, 0.0496, 0.0643], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:37:22,446 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.376e+02 2.857e+02 3.605e+02 7.537e+02, threshold=5.714e+02, percent-clipped=2.0 2023-05-01 03:38:00,752 INFO [train.py:904] (3/8) Epoch 20, batch 8350, loss[loss=0.175, simple_loss=0.2719, pruned_loss=0.03905, over 16894.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2819, pruned_loss=0.05085, over 3082490.63 frames. ], batch size: 109, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:09,513 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6271, 2.3614, 2.2147, 4.4527, 2.2539, 2.7617, 2.4195, 2.4853], device='cuda:3'), covar=tensor([0.1068, 0.3644, 0.3132, 0.0409, 0.4264, 0.2552, 0.3745, 0.3412], device='cuda:3'), in_proj_covar=tensor([0.0387, 0.0433, 0.0356, 0.0316, 0.0425, 0.0496, 0.0402, 0.0506], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:39:23,230 INFO [train.py:904] (3/8) Epoch 20, batch 8400, loss[loss=0.1708, simple_loss=0.2691, pruned_loss=0.03627, over 16400.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.279, pruned_loss=0.0489, over 3068631.31 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:26,525 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:00,771 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:08,502 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.430e+02 2.319e+02 2.651e+02 3.336e+02 5.393e+02, threshold=5.301e+02, percent-clipped=0.0 2023-05-01 03:40:44,263 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201301.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:45,024 INFO [train.py:904] (3/8) Epoch 20, batch 8450, loss[loss=0.185, simple_loss=0.2805, pruned_loss=0.0447, over 15265.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2776, pruned_loss=0.04739, over 3065284.29 frames. ], batch size: 190, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:41:05,384 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:41:39,423 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:02,465 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:07,350 INFO [train.py:904] (3/8) Epoch 20, batch 8500, loss[loss=0.1827, simple_loss=0.2644, pruned_loss=0.05048, over 11900.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2738, pruned_loss=0.04535, over 3045599.57 frames. ], batch size: 246, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:42:50,125 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.179e+02 2.765e+02 3.393e+02 6.239e+02, threshold=5.531e+02, percent-clipped=4.0 2023-05-01 03:43:31,052 INFO [train.py:904] (3/8) Epoch 20, batch 8550, loss[loss=0.1919, simple_loss=0.2826, pruned_loss=0.05062, over 15400.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2715, pruned_loss=0.04416, over 3039292.73 frames. ], batch size: 190, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:44:18,656 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:45:08,666 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 03:45:09,639 INFO [train.py:904] (3/8) Epoch 20, batch 8600, loss[loss=0.1873, simple_loss=0.2846, pruned_loss=0.045, over 16255.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2714, pruned_loss=0.0431, over 3024082.09 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:46:02,940 INFO [optim.py:368] (3/8) 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 03:46:24,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6129, 3.6087, 3.5843, 2.9005, 3.5235, 2.0609, 3.2716, 3.0460], device='cuda:3'), covar=tensor([0.0129, 0.0104, 0.0150, 0.0218, 0.0095, 0.2250, 0.0123, 0.0195], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0146, 0.0187, 0.0170, 0.0165, 0.0200, 0.0178, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:46:48,552 INFO [train.py:904] (3/8) Epoch 20, batch 8650, loss[loss=0.158, simple_loss=0.2469, pruned_loss=0.03459, over 12200.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2696, pruned_loss=0.04181, over 3011577.77 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:48:02,750 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2015, 4.2755, 4.4197, 4.2251, 4.3240, 4.8237, 4.3988, 4.0994], device='cuda:3'), covar=tensor([0.1550, 0.2062, 0.2040, 0.2042, 0.2616, 0.1032, 0.1488, 0.2418], device='cuda:3'), in_proj_covar=tensor([0.0385, 0.0562, 0.0619, 0.0464, 0.0616, 0.0650, 0.0486, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:48:36,193 INFO [train.py:904] (3/8) Epoch 20, batch 8700, loss[loss=0.1713, simple_loss=0.2648, pruned_loss=0.03895, over 16807.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2677, pruned_loss=0.04092, over 3032278.82 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:49:28,969 INFO [optim.py:368] (3/8) 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 03:49:44,534 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 03:50:13,601 INFO [train.py:904] (3/8) Epoch 20, batch 8750, loss[loss=0.1791, simple_loss=0.2786, pruned_loss=0.03976, over 16784.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2682, pruned_loss=0.04081, over 3046927.30 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:50:26,426 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0522, 3.1887, 3.2352, 2.1856, 2.9536, 3.2616, 3.0722, 1.9944], device='cuda:3'), covar=tensor([0.0495, 0.0057, 0.0054, 0.0374, 0.0111, 0.0080, 0.0086, 0.0461], device='cuda:3'), in_proj_covar=tensor([0.0132, 0.0079, 0.0078, 0.0129, 0.0094, 0.0105, 0.0090, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:50:31,548 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201609.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:50:56,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8983, 2.0879, 2.3476, 3.1840, 2.1405, 2.2596, 2.2608, 2.1599], device='cuda:3'), covar=tensor([0.1323, 0.3819, 0.2847, 0.0734, 0.4689, 0.2674, 0.3752, 0.3856], device='cuda:3'), in_proj_covar=tensor([0.0386, 0.0432, 0.0355, 0.0315, 0.0425, 0.0495, 0.0402, 0.0504], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:51:22,016 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201630.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:51,054 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 8800, loss[loss=0.1884, simple_loss=0.2765, pruned_loss=0.05017, over 12635.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2664, pruned_loss=0.03935, over 3066852.47 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:52:24,867 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201661.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:06,514 INFO [optim.py:368] (3/8) 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 03:53:51,629 INFO [train.py:904] (3/8) Epoch 20, batch 8850, loss[loss=0.1744, simple_loss=0.2763, pruned_loss=0.03624, over 15498.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2691, pruned_loss=0.03859, over 3074331.17 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:53:57,870 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:59,726 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:10,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0966, 2.0046, 2.0591, 3.6243, 1.9456, 2.3015, 2.1355, 2.1468], device='cuda:3'), covar=tensor([0.1338, 0.4104, 0.3401, 0.0592, 0.4860, 0.2989, 0.4018, 0.3883], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0434, 0.0358, 0.0316, 0.0428, 0.0498, 0.0404, 0.0507], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 03:54:34,717 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:45,991 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 8900, loss[loss=0.1671, simple_loss=0.2688, pruned_loss=0.03273, over 15398.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2691, pruned_loss=0.03793, over 3072754.23 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:56:08,861 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:27,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:46,608 INFO [optim.py:368] (3/8) 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 03:57:41,707 INFO [train.py:904] (3/8) Epoch 20, batch 8950, loss[loss=0.1541, simple_loss=0.2518, pruned_loss=0.02819, over 16884.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2685, pruned_loss=0.0383, over 3072450.98 frames. ], batch size: 96, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:57:55,557 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1872, 3.4943, 3.5414, 2.4518, 3.2600, 3.5618, 3.3421, 2.1367], device='cuda:3'), covar=tensor([0.0536, 0.0046, 0.0046, 0.0368, 0.0098, 0.0067, 0.0075, 0.0460], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0078, 0.0077, 0.0128, 0.0094, 0.0103, 0.0089, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 03:59:10,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8128, 3.1844, 3.3546, 2.0047, 2.8448, 2.2071, 3.4009, 3.4233], device='cuda:3'), covar=tensor([0.0269, 0.0787, 0.0551, 0.2034, 0.0807, 0.0963, 0.0599, 0.0868], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0155, 0.0160, 0.0148, 0.0140, 0.0126, 0.0139, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 03:59:30,703 INFO [train.py:904] (3/8) Epoch 20, batch 9000, loss[loss=0.1777, simple_loss=0.266, pruned_loss=0.04469, over 16365.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2654, pruned_loss=0.03694, over 3081832.55 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:30,703 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 03:59:41,114 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 04:00:23,565 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0886, 5.1206, 4.9492, 4.4904, 4.6086, 4.9972, 4.9167, 4.6694], device='cuda:3'), covar=tensor([0.0642, 0.0644, 0.0324, 0.0338, 0.1081, 0.0573, 0.0333, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0394, 0.0319, 0.0315, 0.0327, 0.0364, 0.0220, 0.0379], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:00:41,793 INFO [optim.py:368] (3/8) 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 04:01:14,809 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-05-01 04:01:18,109 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4185, 3.0582, 2.6675, 2.2359, 2.1771, 2.2885, 2.9339, 2.8616], device='cuda:3'), covar=tensor([0.2489, 0.0734, 0.1608, 0.2806, 0.2497, 0.2030, 0.0441, 0.1267], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0261, 0.0296, 0.0300, 0.0284, 0.0249, 0.0284, 0.0322], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 04:01:24,590 INFO [train.py:904] (3/8) Epoch 20, batch 9050, loss[loss=0.1631, simple_loss=0.2525, pruned_loss=0.03684, over 16883.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2663, pruned_loss=0.03738, over 3088030.22 frames. ], batch size: 96, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:01:40,823 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:02:21,296 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:03:00,304 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-05-01 04:03:12,543 INFO [train.py:904] (3/8) Epoch 20, batch 9100, loss[loss=0.171, simple_loss=0.2725, pruned_loss=0.0347, over 16175.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2658, pruned_loss=0.03766, over 3074157.52 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:03:22,655 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:14,333 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:23,055 INFO [optim.py:368] (3/8) 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 04:04:42,874 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:11,348 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 9150, loss[loss=0.1621, simple_loss=0.2519, pruned_loss=0.0362, over 11969.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2666, pruned_loss=0.03751, over 3068825.29 frames. ], batch size: 250, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:05:48,712 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202017.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:58,588 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 9200, loss[loss=0.1575, simple_loss=0.2421, pruned_loss=0.03646, over 11977.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2622, pruned_loss=0.03671, over 3071795.71 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:07:19,866 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 04:08:00,917 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 04:08:37,604 INFO [train.py:904] (3/8) Epoch 20, batch 9250, loss[loss=0.1759, simple_loss=0.2773, pruned_loss=0.03722, over 16164.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2617, pruned_loss=0.03682, over 3065768.85 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:09:28,874 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0046, 2.7039, 2.9223, 2.1267, 2.7142, 2.2306, 2.6881, 2.9233], device='cuda:3'), covar=tensor([0.0387, 0.0927, 0.0492, 0.1794, 0.0820, 0.0864, 0.0753, 0.0905], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0154, 0.0159, 0.0146, 0.0138, 0.0125, 0.0138, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 04:10:29,963 INFO [train.py:904] (3/8) Epoch 20, batch 9300, loss[loss=0.1477, simple_loss=0.24, pruned_loss=0.0277, over 16436.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2603, pruned_loss=0.0364, over 3079960.53 frames. ], batch size: 68, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:11:16,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8202, 4.9416, 5.2717, 5.2705, 5.2398, 4.9748, 4.9138, 4.7265], device='cuda:3'), covar=tensor([0.0304, 0.0533, 0.0360, 0.0377, 0.0475, 0.0331, 0.0814, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0385, 0.0421, 0.0412, 0.0381, 0.0456, 0.0432, 0.0514, 0.0347], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 04:11:37,345 INFO [optim.py:368] (3/8) 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 04:12:04,540 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5028, 3.4597, 3.4690, 2.5901, 3.4054, 1.9528, 3.1297, 2.7191], device='cuda:3'), covar=tensor([0.0204, 0.0169, 0.0203, 0.0375, 0.0145, 0.3035, 0.0186, 0.0378], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0144, 0.0184, 0.0166, 0.0164, 0.0199, 0.0175, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:12:16,142 INFO [train.py:904] (3/8) Epoch 20, batch 9350, loss[loss=0.195, simple_loss=0.2952, pruned_loss=0.04737, over 15164.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2601, pruned_loss=0.03641, over 3070875.04 frames. ], batch size: 190, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:12:35,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0786, 2.1466, 2.3164, 3.5425, 2.1563, 2.3794, 2.2927, 2.2509], device='cuda:3'), covar=tensor([0.1297, 0.3960, 0.2966, 0.0587, 0.4349, 0.2663, 0.3731, 0.3672], device='cuda:3'), in_proj_covar=tensor([0.0387, 0.0432, 0.0357, 0.0315, 0.0426, 0.0495, 0.0402, 0.0504], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:13:15,308 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9123, 2.6700, 2.7872, 2.0499, 2.5988, 2.1010, 2.7178, 2.8582], device='cuda:3'), covar=tensor([0.0372, 0.0973, 0.0571, 0.2012, 0.0979, 0.1065, 0.0789, 0.0927], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0154, 0.0159, 0.0147, 0.0139, 0.0125, 0.0139, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 04:13:59,197 INFO [train.py:904] (3/8) Epoch 20, batch 9400, loss[loss=0.177, simple_loss=0.2798, pruned_loss=0.03704, over 16318.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2602, pruned_loss=0.03605, over 3067339.72 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:14:09,642 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3511, 3.7266, 3.7965, 2.5606, 3.4267, 3.8844, 3.6707, 1.8548], device='cuda:3'), covar=tensor([0.0587, 0.0078, 0.0068, 0.0452, 0.0122, 0.0096, 0.0090, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0078, 0.0078, 0.0129, 0.0094, 0.0104, 0.0089, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 04:15:00,972 INFO [optim.py:368] (3/8) 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 04:15:39,424 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 9450, loss[loss=0.1823, simple_loss=0.2816, pruned_loss=0.04149, over 16839.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2625, pruned_loss=0.03649, over 3071699.89 frames. ], batch size: 116, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:16:13,624 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202317.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:16:33,513 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 04:17:10,128 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5896, 4.8754, 4.6871, 4.7035, 4.4342, 4.3663, 4.3005, 4.9203], device='cuda:3'), covar=tensor([0.1047, 0.0821, 0.0840, 0.0754, 0.0698, 0.1278, 0.1124, 0.0804], device='cuda:3'), in_proj_covar=tensor([0.0626, 0.0759, 0.0623, 0.0573, 0.0486, 0.0494, 0.0637, 0.0591], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:17:14,534 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202346.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:17,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202348.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:25,745 INFO [train.py:904] (3/8) Epoch 20, batch 9500, loss[loss=0.1814, simple_loss=0.2824, pruned_loss=0.04021, over 16716.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2619, pruned_loss=0.03636, over 3072698.98 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:17:49,640 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:18:21,447 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 04:18:26,629 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 2.130e+02 2.550e+02 3.364e+02 1.443e+03, threshold=5.099e+02, percent-clipped=7.0 2023-05-01 04:19:13,221 INFO [train.py:904] (3/8) Epoch 20, batch 9550, loss[loss=0.1946, simple_loss=0.2945, pruned_loss=0.04732, over 15292.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2617, pruned_loss=0.03631, over 3094518.19 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:19:32,642 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202410.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:20:37,997 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8074, 2.7402, 2.4082, 2.5499, 3.1800, 2.8181, 3.2671, 3.3908], device='cuda:3'), covar=tensor([0.0149, 0.0445, 0.0561, 0.0538, 0.0314, 0.0484, 0.0285, 0.0297], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0226, 0.0218, 0.0218, 0.0225, 0.0226, 0.0221, 0.0219], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:20:55,236 INFO [train.py:904] (3/8) Epoch 20, batch 9600, loss[loss=0.2004, simple_loss=0.2996, pruned_loss=0.05062, over 15385.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2633, pruned_loss=0.03705, over 3088024.91 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:21:53,893 INFO [optim.py:368] (3/8) 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 04:22:19,978 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2008, 4.0358, 4.2392, 4.3833, 4.5550, 4.0629, 4.5060, 4.5413], device='cuda:3'), covar=tensor([0.1827, 0.1231, 0.1714, 0.0796, 0.0531, 0.1121, 0.0591, 0.0701], device='cuda:3'), in_proj_covar=tensor([0.0592, 0.0727, 0.0847, 0.0745, 0.0564, 0.0583, 0.0601, 0.0696], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:22:41,354 INFO [train.py:904] (3/8) Epoch 20, batch 9650, loss[loss=0.1697, simple_loss=0.2688, pruned_loss=0.03531, over 16972.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2646, pruned_loss=0.03732, over 3080454.85 frames. ], batch size: 55, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:23:52,545 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3540, 3.4784, 3.6837, 3.6711, 3.6732, 3.5174, 3.5299, 3.5459], device='cuda:3'), covar=tensor([0.0413, 0.0797, 0.0498, 0.0460, 0.0520, 0.0516, 0.0755, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0380, 0.0418, 0.0409, 0.0379, 0.0452, 0.0428, 0.0510, 0.0343], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 04:24:31,595 INFO [train.py:904] (3/8) Epoch 20, batch 9700, loss[loss=0.1927, simple_loss=0.2817, pruned_loss=0.05182, over 16626.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2636, pruned_loss=0.03733, over 3071058.58 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:25:37,512 INFO [optim.py:368] (3/8) 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 04:26:15,968 INFO [train.py:904] (3/8) Epoch 20, batch 9750, loss[loss=0.1714, simple_loss=0.2525, pruned_loss=0.04518, over 12263.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2621, pruned_loss=0.03759, over 3043698.66 frames. ], batch size: 250, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:27:47,507 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 20, batch 9800, loss[loss=0.1735, simple_loss=0.2815, pruned_loss=0.03276, over 15295.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2625, pruned_loss=0.0368, over 3047271.88 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:28:57,878 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.120e+02 2.499e+02 2.833e+02 7.051e+02, threshold=4.998e+02, percent-clipped=2.0 2023-05-01 04:29:26,526 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:29:42,375 INFO [train.py:904] (3/8) Epoch 20, batch 9850, loss[loss=0.1717, simple_loss=0.2659, pruned_loss=0.03879, over 16691.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2637, pruned_loss=0.03667, over 3055571.82 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:29:49,794 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 04:30:11,212 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7131, 3.0240, 3.3539, 1.9169, 2.7954, 2.1509, 3.2761, 3.2356], device='cuda:3'), covar=tensor([0.0248, 0.0790, 0.0497, 0.2034, 0.0825, 0.0982, 0.0689, 0.0832], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0153, 0.0159, 0.0147, 0.0138, 0.0124, 0.0139, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 04:31:34,887 INFO [train.py:904] (3/8) Epoch 20, batch 9900, loss[loss=0.175, simple_loss=0.2818, pruned_loss=0.03414, over 16366.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2644, pruned_loss=0.03668, over 3067256.58 frames. ], batch size: 166, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:32:49,188 INFO [optim.py:368] (3/8) 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 04:33:05,855 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 04:33:30,682 INFO [train.py:904] (3/8) Epoch 20, batch 9950, loss[loss=0.178, simple_loss=0.2798, pruned_loss=0.03813, over 16324.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2659, pruned_loss=0.03678, over 3073490.02 frames. ], batch size: 165, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:34:12,249 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8434, 3.6775, 3.7979, 4.0329, 4.1456, 3.7546, 4.1488, 4.1632], device='cuda:3'), covar=tensor([0.2094, 0.1422, 0.2072, 0.1040, 0.0908, 0.1799, 0.0945, 0.1083], device='cuda:3'), in_proj_covar=tensor([0.0594, 0.0727, 0.0848, 0.0749, 0.0565, 0.0584, 0.0603, 0.0698], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:34:41,301 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202830.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:34:56,208 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 04:35:16,326 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9163, 2.1269, 2.3281, 3.1952, 2.1409, 2.3187, 2.2650, 2.2073], device='cuda:3'), covar=tensor([0.1296, 0.3876, 0.2763, 0.0695, 0.3960, 0.2463, 0.3898, 0.3240], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0431, 0.0357, 0.0315, 0.0425, 0.0492, 0.0402, 0.0502], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:35:32,300 INFO [train.py:904] (3/8) Epoch 20, batch 10000, loss[loss=0.1966, simple_loss=0.2959, pruned_loss=0.04861, over 16653.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2645, pruned_loss=0.03646, over 3091761.70 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:35:53,926 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5965, 1.7540, 2.0993, 2.5234, 2.4735, 2.7196, 1.9014, 2.7480], device='cuda:3'), covar=tensor([0.0227, 0.0551, 0.0377, 0.0298, 0.0323, 0.0222, 0.0538, 0.0157], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0184, 0.0170, 0.0174, 0.0186, 0.0143, 0.0187, 0.0139], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:36:36,180 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.233e+02 2.660e+02 3.205e+02 6.054e+02, threshold=5.319e+02, percent-clipped=4.0 2023-05-01 04:36:51,594 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202891.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 04:37:11,277 INFO [train.py:904] (3/8) Epoch 20, batch 10050, loss[loss=0.1874, simple_loss=0.2825, pruned_loss=0.04611, over 16700.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.264, pruned_loss=0.03633, over 3078553.75 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:38:41,706 INFO [train.py:904] (3/8) Epoch 20, batch 10100, loss[loss=0.1581, simple_loss=0.2499, pruned_loss=0.0332, over 15421.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2644, pruned_loss=0.03651, over 3074015.08 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:38:45,026 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4433, 4.4969, 4.3132, 4.0083, 4.0448, 4.4303, 4.1694, 4.1253], device='cuda:3'), covar=tensor([0.0606, 0.0777, 0.0344, 0.0297, 0.0787, 0.0808, 0.0600, 0.0631], device='cuda:3'), in_proj_covar=tensor([0.0270, 0.0383, 0.0311, 0.0306, 0.0318, 0.0354, 0.0214, 0.0370], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-05-01 04:39:01,629 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2514, 3.6017, 3.6289, 2.4974, 3.2202, 3.6282, 3.3803, 2.0686], device='cuda:3'), covar=tensor([0.0507, 0.0050, 0.0048, 0.0359, 0.0125, 0.0088, 0.0089, 0.0476], device='cuda:3'), in_proj_covar=tensor([0.0131, 0.0077, 0.0077, 0.0128, 0.0093, 0.0102, 0.0088, 0.0123], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 04:39:40,382 INFO [optim.py:368] (3/8) 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 04:39:46,239 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-01 04:40:23,119 INFO [train.py:904] (3/8) Epoch 21, batch 0, loss[loss=0.1505, simple_loss=0.2386, pruned_loss=0.03122, over 17234.00 frames. ], tot_loss[loss=0.1505, simple_loss=0.2386, pruned_loss=0.03122, over 17234.00 frames. ], batch size: 44, lr: 3.26e-03, grad_scale: 8.0 2023-05-01 04:40:23,119 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 04:40:30,896 INFO [train.py:938] (3/8) Epoch 21, validation: loss=0.1451, simple_loss=0.2491, pruned_loss=0.02058, over 944034.00 frames. 2023-05-01 04:40:30,897 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 04:41:07,228 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9998, 4.7573, 5.0075, 5.1572, 5.3317, 4.7181, 5.3239, 5.3017], device='cuda:3'), covar=tensor([0.1768, 0.1178, 0.1593, 0.0759, 0.0572, 0.0779, 0.0512, 0.0638], device='cuda:3'), in_proj_covar=tensor([0.0594, 0.0726, 0.0847, 0.0750, 0.0565, 0.0582, 0.0603, 0.0699], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:41:19,619 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:41:36,907 INFO [train.py:904] (3/8) Epoch 21, batch 50, loss[loss=0.198, simple_loss=0.2794, pruned_loss=0.05827, over 16363.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2725, pruned_loss=0.05344, over 740954.57 frames. ], batch size: 165, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:42:19,953 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5588, 4.3649, 4.5917, 4.7146, 4.8720, 4.3433, 4.7474, 4.8281], device='cuda:3'), covar=tensor([0.1777, 0.1444, 0.1633, 0.1061, 0.0830, 0.1217, 0.1372, 0.1379], device='cuda:3'), in_proj_covar=tensor([0.0599, 0.0732, 0.0855, 0.0757, 0.0569, 0.0587, 0.0608, 0.0704], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 04:42:25,285 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 04:42:43,747 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 100, loss[loss=0.184, simple_loss=0.2725, pruned_loss=0.04772, over 16824.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2675, pruned_loss=0.0489, over 1299398.87 frames. ], batch size: 116, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:42:52,144 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 04:43:50,269 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:43:57,338 INFO [train.py:904] (3/8) Epoch 21, batch 150, loss[loss=0.1765, simple_loss=0.269, pruned_loss=0.04204, over 17152.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2661, pruned_loss=0.04741, over 1749594.00 frames. ], batch size: 48, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:44:44,682 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203186.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:45:04,109 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6571, 3.7506, 2.6847, 2.2130, 2.5359, 2.2230, 3.8391, 3.3224], device='cuda:3'), covar=tensor([0.2943, 0.0605, 0.1850, 0.2980, 0.2595, 0.2173, 0.0553, 0.1468], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0263, 0.0298, 0.0303, 0.0285, 0.0252, 0.0287, 0.0327], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 04:45:04,855 INFO [train.py:904] (3/8) Epoch 21, batch 200, loss[loss=0.1979, simple_loss=0.2723, pruned_loss=0.06171, over 16802.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2659, pruned_loss=0.04657, over 2098921.91 frames. ], batch size: 116, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:45:38,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7157, 3.8129, 2.1342, 4.4355, 2.8838, 4.3037, 2.2167, 3.0532], device='cuda:3'), covar=tensor([0.0312, 0.0442, 0.2009, 0.0330, 0.0962, 0.0643, 0.1951, 0.0841], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0173, 0.0192, 0.0156, 0.0174, 0.0209, 0.0202, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 04:46:15,294 INFO [train.py:904] (3/8) Epoch 21, batch 250, loss[loss=0.1773, simple_loss=0.2478, pruned_loss=0.05344, over 16926.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2641, pruned_loss=0.04728, over 2366048.00 frames. ], batch size: 96, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:47:00,828 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 300, loss[loss=0.1502, simple_loss=0.2461, pruned_loss=0.02715, over 16971.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2613, pruned_loss=0.0452, over 2584126.21 frames. ], batch size: 41, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:48:32,611 INFO [train.py:904] (3/8) Epoch 21, batch 350, loss[loss=0.1665, simple_loss=0.2466, pruned_loss=0.04322, over 16789.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2571, pruned_loss=0.04356, over 2749649.84 frames. ], batch size: 102, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:49:15,129 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 04:49:17,934 INFO [optim.py:368] (3/8) 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 04:49:30,094 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 400, loss[loss=0.1633, simple_loss=0.2607, pruned_loss=0.033, over 17057.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2556, pruned_loss=0.04315, over 2880847.82 frames. ], batch size: 53, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:50:31,792 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9332, 4.3423, 4.4153, 3.1814, 3.6410, 4.3070, 3.9124, 2.5425], device='cuda:3'), covar=tensor([0.0454, 0.0067, 0.0041, 0.0331, 0.0136, 0.0095, 0.0083, 0.0463], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0082, 0.0081, 0.0133, 0.0097, 0.0107, 0.0092, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 04:50:37,061 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203441.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:50:50,747 INFO [train.py:904] (3/8) Epoch 21, batch 450, loss[loss=0.1547, simple_loss=0.234, pruned_loss=0.03768, over 16831.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2545, pruned_loss=0.04265, over 2981383.26 frames. ], batch size: 102, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:51:10,652 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6850, 3.4118, 3.8578, 2.1054, 3.9885, 3.9907, 3.1655, 2.9782], device='cuda:3'), covar=tensor([0.0724, 0.0273, 0.0172, 0.1118, 0.0085, 0.0181, 0.0382, 0.0445], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0139, 0.0078, 0.0121, 0.0127, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 04:51:11,801 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:51:22,355 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-05-01 04:51:37,694 INFO [optim.py:368] (3/8) 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 04:51:38,100 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203486.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 04:51:59,126 INFO [train.py:904] (3/8) Epoch 21, batch 500, loss[loss=0.1583, simple_loss=0.2367, pruned_loss=0.03995, over 16664.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2536, pruned_loss=0.04203, over 3059189.80 frames. ], batch size: 134, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:52:36,241 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203534.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:53:08,035 INFO [train.py:904] (3/8) Epoch 21, batch 550, loss[loss=0.1405, simple_loss=0.226, pruned_loss=0.02749, over 17212.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2527, pruned_loss=0.04167, over 3110581.86 frames. ], batch size: 44, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:53:56,912 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 600, loss[loss=0.1734, simple_loss=0.2447, pruned_loss=0.05103, over 16767.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2523, pruned_loss=0.04156, over 3160815.25 frames. ], batch size: 83, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:27,189 INFO [train.py:904] (3/8) Epoch 21, batch 650, loss[loss=0.1791, simple_loss=0.2543, pruned_loss=0.05193, over 16740.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2514, pruned_loss=0.04148, over 3186208.64 frames. ], batch size: 134, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:50,895 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 04:56:01,737 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1936, 4.2426, 4.4051, 4.1788, 4.2887, 4.8164, 4.3985, 4.0953], device='cuda:3'), covar=tensor([0.1876, 0.2295, 0.2494, 0.2437, 0.2995, 0.1297, 0.1745, 0.2585], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0584, 0.0641, 0.0483, 0.0644, 0.0674, 0.0503, 0.0646], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 04:56:14,236 INFO [optim.py:368] (3/8) 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 04:56:22,603 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-01 04:56:24,918 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:56:34,645 INFO [train.py:904] (3/8) Epoch 21, batch 700, loss[loss=0.1551, simple_loss=0.2391, pruned_loss=0.03561, over 16818.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2513, pruned_loss=0.04089, over 3222196.09 frames. ], batch size: 42, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:57:30,218 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203741.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:31,297 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:37,639 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 750, loss[loss=0.1717, simple_loss=0.26, pruned_loss=0.04175, over 16438.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2521, pruned_loss=0.04047, over 3250602.63 frames. ], batch size: 68, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:58:03,795 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.172e+02 2.611e+02 3.083e+02 3.128e+03, threshold=5.221e+02, percent-clipped=7.0 2023-05-01 04:58:36,276 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:54,753 INFO [train.py:904] (3/8) Epoch 21, batch 800, loss[loss=0.1747, simple_loss=0.2527, pruned_loss=0.04839, over 16735.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2518, pruned_loss=0.04038, over 3272899.98 frames. ], batch size: 89, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:59:03,679 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203808.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:24,697 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203823.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:30,070 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:58,186 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 850, loss[loss=0.1586, simple_loss=0.2344, pruned_loss=0.04141, over 16732.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2512, pruned_loss=0.0401, over 3278658.82 frames. ], batch size: 124, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:00:53,309 INFO [optim.py:368] (3/8) 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 05:01:11,269 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203900.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:01:12,967 INFO [train.py:904] (3/8) Epoch 21, batch 900, loss[loss=0.1528, simple_loss=0.2358, pruned_loss=0.03494, over 16713.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.25, pruned_loss=0.03923, over 3293523.84 frames. ], batch size: 89, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:01:21,626 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203908.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:02:21,433 INFO [train.py:904] (3/8) Epoch 21, batch 950, loss[loss=0.1638, simple_loss=0.2374, pruned_loss=0.04512, over 16155.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2504, pruned_loss=0.0392, over 3306991.10 frames. ], batch size: 164, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:02:35,108 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203961.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:03:10,185 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 1000, loss[loss=0.1778, simple_loss=0.271, pruned_loss=0.04231, over 17258.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2493, pruned_loss=0.03936, over 3296623.47 frames. ], batch size: 52, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:03:51,404 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0233, 2.1099, 2.5241, 2.9172, 2.7748, 3.1946, 2.2569, 3.2636], device='cuda:3'), covar=tensor([0.0201, 0.0463, 0.0351, 0.0280, 0.0341, 0.0213, 0.0480, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0192, 0.0178, 0.0182, 0.0194, 0.0152, 0.0195, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:04:10,897 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3531, 4.1367, 4.5276, 2.5055, 4.7974, 4.8380, 3.4781, 3.8761], device='cuda:3'), covar=tensor([0.0602, 0.0227, 0.0223, 0.1052, 0.0077, 0.0141, 0.0420, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0108, 0.0096, 0.0138, 0.0079, 0.0122, 0.0127, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 05:04:18,886 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 05:04:43,737 INFO [train.py:904] (3/8) Epoch 21, batch 1050, loss[loss=0.1572, simple_loss=0.2393, pruned_loss=0.03755, over 16798.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2488, pruned_loss=0.03943, over 3300679.66 frames. ], batch size: 102, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:09,320 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-05-01 05:05:10,378 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1723, 4.6000, 4.5819, 3.4776, 3.7979, 4.4669, 4.0490, 2.7304], device='cuda:3'), covar=tensor([0.0409, 0.0048, 0.0038, 0.0285, 0.0134, 0.0094, 0.0078, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:3') 2023-05-01 05:05:19,321 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 05:05:54,729 INFO [train.py:904] (3/8) Epoch 21, batch 1100, loss[loss=0.1493, simple_loss=0.2345, pruned_loss=0.03198, over 16991.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2488, pruned_loss=0.03952, over 3305849.11 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:56,311 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:00,771 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9618, 2.0715, 2.4981, 2.8835, 2.6796, 3.4370, 2.2555, 3.3529], device='cuda:3'), covar=tensor([0.0271, 0.0516, 0.0365, 0.0335, 0.0366, 0.0171, 0.0469, 0.0175], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0192, 0.0179, 0.0183, 0.0194, 0.0152, 0.0195, 0.0147], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:06:24,054 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:25,273 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204123.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:45,634 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204138.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 05:07:05,256 INFO [train.py:904] (3/8) Epoch 21, batch 1150, loss[loss=0.1599, simple_loss=0.2414, pruned_loss=0.03922, over 16958.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2486, pruned_loss=0.03895, over 3312235.01 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:07:20,313 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 05:08:07,156 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6487, 6.0596, 5.7984, 5.8482, 5.4733, 5.5501, 5.4101, 6.1605], device='cuda:3'), covar=tensor([0.1348, 0.0948, 0.1074, 0.0914, 0.0901, 0.0640, 0.1231, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0678, 0.0826, 0.0679, 0.0623, 0.0526, 0.0533, 0.0696, 0.0642], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:08:13,780 INFO [train.py:904] (3/8) Epoch 21, batch 1200, loss[loss=0.1622, simple_loss=0.2596, pruned_loss=0.03242, over 17106.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2477, pruned_loss=0.03892, over 3312435.14 frames. ], batch size: 49, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:08:15,242 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204203.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:08:15,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6909, 4.7996, 4.5805, 4.2971, 3.9227, 4.8332, 4.6840, 4.4206], device='cuda:3'), covar=tensor([0.1078, 0.1159, 0.0608, 0.0512, 0.1852, 0.0633, 0.0591, 0.0897], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0426, 0.0344, 0.0341, 0.0352, 0.0397, 0.0237, 0.0414], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:08:28,559 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9507, 5.2880, 5.4675, 5.1857, 5.3150, 5.8738, 5.3831, 5.1044], device='cuda:3'), covar=tensor([0.1289, 0.2157, 0.2549, 0.2219, 0.2816, 0.1025, 0.1861, 0.2420], device='cuda:3'), in_proj_covar=tensor([0.0404, 0.0594, 0.0654, 0.0491, 0.0656, 0.0683, 0.0515, 0.0659], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:08:44,929 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:09:06,311 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 05:09:24,353 INFO [train.py:904] (3/8) Epoch 21, batch 1250, loss[loss=0.142, simple_loss=0.2349, pruned_loss=0.02459, over 17129.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2476, pruned_loss=0.03963, over 3310082.24 frames. ], batch size: 47, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:09:30,712 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204256.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:10:01,347 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7128, 2.4734, 1.8774, 2.1955, 2.8237, 2.5871, 2.7935, 2.9128], device='cuda:3'), covar=tensor([0.0219, 0.0402, 0.0584, 0.0487, 0.0248, 0.0343, 0.0245, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0208, 0.0238, 0.0228, 0.0228, 0.0237, 0.0236, 0.0238, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:10:04,684 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7946, 1.8969, 2.2978, 2.5123, 2.6773, 2.6118, 1.9649, 2.8249], device='cuda:3'), covar=tensor([0.0186, 0.0486, 0.0367, 0.0299, 0.0277, 0.0296, 0.0502, 0.0166], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0194, 0.0180, 0.0184, 0.0196, 0.0153, 0.0196, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:10:12,294 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.352e+02 2.804e+02 3.486e+02 9.364e+02, threshold=5.607e+02, percent-clipped=5.0 2023-05-01 05:10:21,131 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7710, 2.4344, 1.9102, 2.2045, 2.7902, 2.5445, 2.7864, 2.9186], device='cuda:3'), covar=tensor([0.0257, 0.0460, 0.0622, 0.0488, 0.0285, 0.0398, 0.0247, 0.0314], device='cuda:3'), in_proj_covar=tensor([0.0208, 0.0237, 0.0227, 0.0227, 0.0237, 0.0236, 0.0238, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:10:33,932 INFO [train.py:904] (3/8) Epoch 21, batch 1300, loss[loss=0.1767, simple_loss=0.2499, pruned_loss=0.05177, over 16546.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2473, pruned_loss=0.04008, over 3315079.58 frames. ], batch size: 146, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:17,955 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2716, 3.3574, 3.7396, 2.1124, 3.1418, 2.4693, 3.7282, 3.6074], device='cuda:3'), covar=tensor([0.0263, 0.0923, 0.0545, 0.2070, 0.0822, 0.0966, 0.0575, 0.1069], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0153, 0.0145, 0.0130, 0.0145, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 05:11:42,899 INFO [train.py:904] (3/8) Epoch 21, batch 1350, loss[loss=0.1733, simple_loss=0.2531, pruned_loss=0.0468, over 16677.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2484, pruned_loss=0.0399, over 3318625.92 frames. ], batch size: 134, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:11:48,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2663, 5.0188, 5.3212, 5.4616, 5.6588, 4.8775, 5.5520, 5.6052], device='cuda:3'), covar=tensor([0.1754, 0.1217, 0.1440, 0.0627, 0.0496, 0.0847, 0.0621, 0.0605], device='cuda:3'), in_proj_covar=tensor([0.0650, 0.0795, 0.0928, 0.0818, 0.0611, 0.0635, 0.0657, 0.0759], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:12:17,658 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 05:12:31,606 INFO [optim.py:368] (3/8) 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 05:12:52,517 INFO [train.py:904] (3/8) Epoch 21, batch 1400, loss[loss=0.1714, simple_loss=0.2512, pruned_loss=0.04587, over 16767.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2484, pruned_loss=0.03982, over 3326047.01 frames. ], batch size: 124, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:54,521 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204403.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:20,717 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 21, batch 1450, loss[loss=0.1572, simple_loss=0.2531, pruned_loss=0.03068, over 17235.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2475, pruned_loss=0.03942, over 3316806.39 frames. ], batch size: 52, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:14:26,854 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204470.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:28,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8570, 2.8224, 2.6757, 4.8483, 3.9409, 4.3122, 1.6000, 3.1991], device='cuda:3'), covar=tensor([0.1356, 0.0826, 0.1217, 0.0231, 0.0254, 0.0420, 0.1697, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0189, 0.0205, 0.0216, 0.0202, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 05:14:50,905 INFO [optim.py:368] (3/8) 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 05:14:57,077 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.8105, 6.1899, 5.8652, 6.0023, 5.5123, 5.6390, 5.5326, 6.2752], device='cuda:3'), covar=tensor([0.1373, 0.0968, 0.1122, 0.0900, 0.0945, 0.0577, 0.1100, 0.0862], device='cuda:3'), in_proj_covar=tensor([0.0681, 0.0831, 0.0680, 0.0627, 0.0529, 0.0535, 0.0699, 0.0644], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:15:10,559 INFO [train.py:904] (3/8) Epoch 21, batch 1500, loss[loss=0.1905, simple_loss=0.2711, pruned_loss=0.05495, over 12288.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2475, pruned_loss=0.03978, over 3314446.08 frames. ], batch size: 246, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:15:12,789 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204519.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:17,196 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:18,101 INFO [train.py:904] (3/8) Epoch 21, batch 1550, loss[loss=0.1939, simple_loss=0.2597, pruned_loss=0.06402, over 16660.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2481, pruned_loss=0.0406, over 3317360.58 frames. ], batch size: 89, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:16:23,551 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204556.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:16:26,463 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 05:17:07,000 INFO [optim.py:368] (3/8) 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 05:17:08,259 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 05:17:26,280 INFO [train.py:904] (3/8) Epoch 21, batch 1600, loss[loss=0.1761, simple_loss=0.2743, pruned_loss=0.03895, over 17118.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2511, pruned_loss=0.04182, over 3316641.81 frames. ], batch size: 49, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:17:28,784 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204604.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:18:06,595 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1541, 5.7360, 5.8421, 5.5441, 5.7312, 6.2440, 5.7948, 5.5099], device='cuda:3'), covar=tensor([0.0940, 0.2122, 0.2324, 0.1980, 0.2513, 0.0958, 0.1424, 0.2164], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0591, 0.0651, 0.0487, 0.0651, 0.0680, 0.0513, 0.0654], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:18:24,666 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8361, 2.0071, 2.3565, 2.5733, 2.6761, 2.6693, 1.9775, 2.8552], device='cuda:3'), covar=tensor([0.0173, 0.0436, 0.0322, 0.0277, 0.0285, 0.0287, 0.0500, 0.0174], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0193, 0.0180, 0.0184, 0.0196, 0.0154, 0.0196, 0.0149], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:18:35,818 INFO [train.py:904] (3/8) Epoch 21, batch 1650, loss[loss=0.207, simple_loss=0.2865, pruned_loss=0.06373, over 15429.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2524, pruned_loss=0.04235, over 3310800.10 frames. ], batch size: 190, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:19:25,764 INFO [optim.py:368] (3/8) 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 05:19:38,406 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-05-01 05:19:45,537 INFO [train.py:904] (3/8) Epoch 21, batch 1700, loss[loss=0.1609, simple_loss=0.2441, pruned_loss=0.0388, over 16546.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2545, pruned_loss=0.04238, over 3317154.23 frames. ], batch size: 75, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:20:29,901 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204733.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:20:46,428 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5235, 3.7872, 3.9857, 2.1725, 3.1644, 2.6211, 3.8851, 3.9917], device='cuda:3'), covar=tensor([0.0274, 0.0872, 0.0523, 0.1997, 0.0851, 0.0940, 0.0646, 0.1089], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0144, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 05:20:55,687 INFO [train.py:904] (3/8) Epoch 21, batch 1750, loss[loss=0.1419, simple_loss=0.2277, pruned_loss=0.02806, over 16824.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2555, pruned_loss=0.04212, over 3318488.45 frames. ], batch size: 96, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:21:29,057 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4357, 2.9792, 2.6745, 2.2631, 2.2272, 2.2904, 2.9582, 2.7333], device='cuda:3'), covar=tensor([0.2494, 0.0744, 0.1622, 0.2618, 0.2355, 0.1962, 0.0585, 0.1319], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0270, 0.0303, 0.0309, 0.0295, 0.0256, 0.0293, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:21:37,073 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204781.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:21:46,292 INFO [optim.py:368] (3/8) 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 05:21:48,179 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-05-01 05:22:06,988 INFO [train.py:904] (3/8) Epoch 21, batch 1800, loss[loss=0.1535, simple_loss=0.2338, pruned_loss=0.03658, over 16800.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2569, pruned_loss=0.04242, over 3315932.96 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:22:30,026 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:32,913 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204821.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:23:15,229 INFO [train.py:904] (3/8) Epoch 21, batch 1850, loss[loss=0.1688, simple_loss=0.2445, pruned_loss=0.04651, over 16752.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2586, pruned_loss=0.04301, over 3327334.93 frames. ], batch size: 83, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:23:15,795 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7894, 2.6723, 2.2848, 2.5664, 3.0602, 2.8016, 3.3579, 3.2556], device='cuda:3'), covar=tensor([0.0143, 0.0416, 0.0514, 0.0426, 0.0270, 0.0381, 0.0264, 0.0258], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0240, 0.0229, 0.0230, 0.0240, 0.0239, 0.0241, 0.0236], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:23:16,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3390, 2.2967, 2.4141, 4.1989, 2.2578, 2.7138, 2.3649, 2.5291], device='cuda:3'), covar=tensor([0.1434, 0.3840, 0.2889, 0.0564, 0.4196, 0.2455, 0.3795, 0.3006], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0448, 0.0370, 0.0331, 0.0436, 0.0515, 0.0418, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:23:37,452 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204867.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:23:57,926 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:24:05,419 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3372, 3.5182, 3.6543, 3.6357, 3.6353, 3.4600, 3.4933, 3.5081], device='cuda:3'), covar=tensor([0.0414, 0.0562, 0.0454, 0.0427, 0.0504, 0.0477, 0.0729, 0.0492], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0454, 0.0443, 0.0411, 0.0490, 0.0465, 0.0554, 0.0371], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 05:24:06,221 INFO [optim.py:368] (3/8) 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 05:24:26,172 INFO [train.py:904] (3/8) Epoch 21, batch 1900, loss[loss=0.1647, simple_loss=0.2574, pruned_loss=0.03601, over 17120.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2581, pruned_loss=0.04262, over 3318617.41 frames. ], batch size: 47, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:25:35,963 INFO [train.py:904] (3/8) Epoch 21, batch 1950, loss[loss=0.1663, simple_loss=0.255, pruned_loss=0.0388, over 16549.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2575, pruned_loss=0.04154, over 3328682.36 frames. ], batch size: 75, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:26:26,411 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.195e+02 2.614e+02 3.172e+02 4.787e+02, threshold=5.228e+02, percent-clipped=0.0 2023-05-01 05:26:44,684 INFO [train.py:904] (3/8) Epoch 21, batch 2000, loss[loss=0.2025, simple_loss=0.2857, pruned_loss=0.05959, over 16537.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2572, pruned_loss=0.04171, over 3330781.59 frames. ], batch size: 68, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:27:55,522 INFO [train.py:904] (3/8) Epoch 21, batch 2050, loss[loss=0.1658, simple_loss=0.2484, pruned_loss=0.04166, over 17234.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2575, pruned_loss=0.04221, over 3334329.91 frames. ], batch size: 45, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:28:44,310 INFO [optim.py:368] (3/8) 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 05:28:53,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9888, 4.9448, 4.8320, 4.4350, 4.5385, 4.9671, 4.7839, 4.5830], device='cuda:3'), covar=tensor([0.0655, 0.0572, 0.0307, 0.0363, 0.0882, 0.0446, 0.0444, 0.0657], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0435, 0.0352, 0.0348, 0.0360, 0.0404, 0.0240, 0.0424], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:29:04,151 INFO [train.py:904] (3/8) Epoch 21, batch 2100, loss[loss=0.1457, simple_loss=0.238, pruned_loss=0.02672, over 17239.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2581, pruned_loss=0.04269, over 3338212.75 frames. ], batch size: 46, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:29:14,566 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205109.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:29:35,472 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 05:30:14,923 INFO [train.py:904] (3/8) Epoch 21, batch 2150, loss[loss=0.1725, simple_loss=0.2653, pruned_loss=0.03987, over 16685.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2592, pruned_loss=0.04335, over 3332666.88 frames. ], batch size: 57, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:30:39,979 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205170.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:30:43,403 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 05:30:48,703 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4963, 5.5101, 5.3146, 4.7448, 5.3399, 2.4847, 5.1075, 5.2219], device='cuda:3'), covar=tensor([0.0088, 0.0075, 0.0191, 0.0390, 0.0108, 0.2270, 0.0124, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0156, 0.0198, 0.0179, 0.0176, 0.0209, 0.0189, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:30:50,421 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205177.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:02,193 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4372, 4.3015, 4.4589, 4.6173, 4.7405, 4.2827, 4.5405, 4.7008], device='cuda:3'), covar=tensor([0.1593, 0.1086, 0.1328, 0.0675, 0.0554, 0.1152, 0.2114, 0.0700], device='cuda:3'), in_proj_covar=tensor([0.0656, 0.0805, 0.0944, 0.0830, 0.0617, 0.0647, 0.0668, 0.0772], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:31:04,861 INFO [optim.py:368] (3/8) 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 05:31:07,433 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2021, 5.7792, 5.8986, 5.4840, 5.7012, 6.2100, 5.7315, 5.4745], device='cuda:3'), covar=tensor([0.0927, 0.1830, 0.2225, 0.1855, 0.2364, 0.0915, 0.1327, 0.1973], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0600, 0.0662, 0.0500, 0.0664, 0.0693, 0.0521, 0.0667], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:31:19,688 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205198.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:24,887 INFO [train.py:904] (3/8) Epoch 21, batch 2200, loss[loss=0.152, simple_loss=0.2378, pruned_loss=0.03307, over 16980.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2592, pruned_loss=0.04351, over 3332946.99 frames. ], batch size: 41, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:32:34,214 INFO [train.py:904] (3/8) Epoch 21, batch 2250, loss[loss=0.1641, simple_loss=0.2435, pruned_loss=0.0424, over 16879.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2592, pruned_loss=0.04399, over 3325878.34 frames. ], batch size: 96, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:32:45,040 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:33:23,477 INFO [optim.py:368] (3/8) 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 05:33:35,867 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 05:33:44,279 INFO [train.py:904] (3/8) Epoch 21, batch 2300, loss[loss=0.1624, simple_loss=0.2464, pruned_loss=0.03926, over 15953.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2586, pruned_loss=0.04308, over 3329009.34 frames. ], batch size: 35, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:14,622 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 05:34:53,179 INFO [train.py:904] (3/8) Epoch 21, batch 2350, loss[loss=0.1905, simple_loss=0.2839, pruned_loss=0.04851, over 16681.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2585, pruned_loss=0.04298, over 3329936.31 frames. ], batch size: 62, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:56,110 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 05:35:42,801 INFO [optim.py:368] (3/8) 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 05:35:46,732 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 05:35:52,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0903, 4.8530, 5.1268, 5.2969, 5.5008, 4.8651, 5.4842, 5.4503], device='cuda:3'), covar=tensor([0.1881, 0.1334, 0.1691, 0.0743, 0.0529, 0.0820, 0.0504, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0810, 0.0949, 0.0833, 0.0619, 0.0652, 0.0671, 0.0776], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:36:02,969 INFO [train.py:904] (3/8) Epoch 21, batch 2400, loss[loss=0.1567, simple_loss=0.2474, pruned_loss=0.03306, over 16963.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2592, pruned_loss=0.04322, over 3315558.10 frames. ], batch size: 41, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:36:10,914 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 05:36:50,413 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 05:37:10,842 INFO [train.py:904] (3/8) Epoch 21, batch 2450, loss[loss=0.1595, simple_loss=0.2592, pruned_loss=0.0299, over 17055.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2598, pruned_loss=0.04259, over 3326106.62 frames. ], batch size: 50, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:24,994 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8777, 4.4963, 4.4765, 3.2561, 3.7709, 4.4321, 3.9186, 2.5565], device='cuda:3'), covar=tensor([0.0463, 0.0065, 0.0042, 0.0335, 0.0122, 0.0089, 0.0087, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:37:29,569 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205465.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:37:46,021 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:38:00,843 INFO [optim.py:368] (3/8) 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 05:38:16,908 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8833, 2.0303, 2.5249, 2.8052, 2.6837, 3.1297, 2.2022, 3.1744], device='cuda:3'), covar=tensor([0.0221, 0.0504, 0.0322, 0.0305, 0.0332, 0.0214, 0.0477, 0.0153], device='cuda:3'), in_proj_covar=tensor([0.0189, 0.0195, 0.0180, 0.0186, 0.0198, 0.0155, 0.0198, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:38:21,260 INFO [train.py:904] (3/8) Epoch 21, batch 2500, loss[loss=0.1738, simple_loss=0.2544, pruned_loss=0.04663, over 16765.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2589, pruned_loss=0.04222, over 3332429.90 frames. ], batch size: 124, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:38:40,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6897, 3.9996, 4.1175, 2.9194, 3.6120, 4.1446, 3.7086, 2.5101], device='cuda:3'), covar=tensor([0.0485, 0.0244, 0.0056, 0.0363, 0.0128, 0.0109, 0.0103, 0.0448], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0134, 0.0099, 0.0109, 0.0094, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:38:53,421 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:39:30,352 INFO [train.py:904] (3/8) Epoch 21, batch 2550, loss[loss=0.1535, simple_loss=0.2495, pruned_loss=0.02874, over 17127.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2589, pruned_loss=0.0421, over 3333056.91 frames. ], batch size: 48, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:39:33,575 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:40:17,546 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7836, 4.5728, 4.8418, 4.9888, 5.1782, 4.5851, 5.1737, 5.1467], device='cuda:3'), covar=tensor([0.1826, 0.1256, 0.1575, 0.0780, 0.0559, 0.1057, 0.0684, 0.0598], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0810, 0.0949, 0.0832, 0.0619, 0.0649, 0.0670, 0.0776], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:40:19,332 INFO [optim.py:368] (3/8) 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 05:40:20,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0078, 2.0715, 2.2399, 3.4797, 2.0813, 2.2981, 2.2169, 2.1475], device='cuda:3'), covar=tensor([0.1527, 0.3830, 0.2931, 0.0747, 0.4436, 0.2934, 0.3730, 0.3635], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0448, 0.0370, 0.0330, 0.0436, 0.0516, 0.0417, 0.0525], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:40:38,681 INFO [train.py:904] (3/8) Epoch 21, batch 2600, loss[loss=0.1535, simple_loss=0.2436, pruned_loss=0.0317, over 16833.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2587, pruned_loss=0.04157, over 3330524.44 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:41:49,803 INFO [train.py:904] (3/8) Epoch 21, batch 2650, loss[loss=0.1892, simple_loss=0.2696, pruned_loss=0.05442, over 16743.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2593, pruned_loss=0.04193, over 3325978.18 frames. ], batch size: 124, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:41:54,584 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205655.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:42:40,190 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 2700, loss[loss=0.1792, simple_loss=0.2613, pruned_loss=0.04854, over 16688.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.259, pruned_loss=0.0413, over 3326551.81 frames. ], batch size: 134, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:43:18,674 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:43:59,592 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:44:09,872 INFO [train.py:904] (3/8) Epoch 21, batch 2750, loss[loss=0.1516, simple_loss=0.2353, pruned_loss=0.03393, over 16273.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2585, pruned_loss=0.04097, over 3320298.15 frames. ], batch size: 165, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:44:14,319 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7506, 3.9136, 3.0301, 2.3070, 2.6194, 2.5508, 4.1846, 3.4107], device='cuda:3'), covar=tensor([0.2726, 0.0626, 0.1719, 0.2749, 0.2604, 0.1939, 0.0526, 0.1358], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0309, 0.0297, 0.0257, 0.0294, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:44:29,216 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 21, batch 2800, loss[loss=0.1641, simple_loss=0.2554, pruned_loss=0.03639, over 17080.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.259, pruned_loss=0.04133, over 3326063.80 frames. ], batch size: 53, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:45:22,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9355, 4.8967, 4.7552, 4.2522, 4.8473, 2.0286, 4.5995, 4.6093], device='cuda:3'), covar=tensor([0.0101, 0.0084, 0.0204, 0.0336, 0.0096, 0.2625, 0.0137, 0.0203], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0157, 0.0199, 0.0180, 0.0178, 0.0210, 0.0190, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:45:24,433 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205805.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:35,278 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205813.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:46:29,085 INFO [train.py:904] (3/8) Epoch 21, batch 2850, loss[loss=0.217, simple_loss=0.2963, pruned_loss=0.06889, over 12207.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2589, pruned_loss=0.04126, over 3324134.46 frames. ], batch size: 248, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:46:31,697 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:17,822 INFO [optim.py:368] (3/8) 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 05:47:19,181 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9807, 5.1729, 5.3343, 5.0703, 5.1440, 5.7526, 5.1940, 4.8881], device='cuda:3'), covar=tensor([0.1110, 0.1966, 0.2172, 0.2093, 0.2525, 0.0954, 0.1673, 0.2460], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0611, 0.0669, 0.0505, 0.0669, 0.0699, 0.0526, 0.0673], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:47:19,292 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 2900, loss[loss=0.1715, simple_loss=0.2505, pruned_loss=0.04622, over 16927.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2573, pruned_loss=0.0417, over 3330852.23 frames. ], batch size: 109, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:47:36,635 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205902.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:44,800 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:45,571 INFO [train.py:904] (3/8) Epoch 21, batch 2950, loss[loss=0.1807, simple_loss=0.2558, pruned_loss=0.05277, over 16748.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2568, pruned_loss=0.04188, over 3330128.12 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:12,740 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9342, 2.3011, 2.4247, 3.1752, 2.2795, 2.4451, 2.4732, 2.3720], device='cuda:3'), covar=tensor([0.1231, 0.2775, 0.2192, 0.0662, 0.3426, 0.2020, 0.2634, 0.3033], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0449, 0.0369, 0.0331, 0.0435, 0.0515, 0.0418, 0.0525], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:49:29,011 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4857, 5.4495, 5.3221, 4.8277, 4.9640, 5.3930, 5.3420, 4.9901], device='cuda:3'), covar=tensor([0.0587, 0.0457, 0.0291, 0.0301, 0.1026, 0.0489, 0.0236, 0.0675], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0437, 0.0355, 0.0350, 0.0363, 0.0407, 0.0243, 0.0427], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:49:36,947 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 3000, loss[loss=0.1787, simple_loss=0.2615, pruned_loss=0.04792, over 16482.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.257, pruned_loss=0.04241, over 3332495.45 frames. ], batch size: 68, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:58,057 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 05:50:06,476 INFO [train.py:938] (3/8) Epoch 21, validation: loss=0.1356, simple_loss=0.2408, pruned_loss=0.01521, over 944034.00 frames. 2023-05-01 05:50:06,477 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 05:50:18,435 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:14,524 INFO [train.py:904] (3/8) Epoch 21, batch 3050, loss[loss=0.1548, simple_loss=0.2369, pruned_loss=0.03636, over 15927.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2559, pruned_loss=0.04156, over 3328482.94 frames. ], batch size: 35, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:51:21,082 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206056.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:56,655 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5462, 5.9421, 5.6740, 5.7871, 5.3179, 5.3583, 5.3286, 6.0926], device='cuda:3'), covar=tensor([0.1331, 0.0972, 0.0990, 0.0864, 0.0904, 0.0709, 0.1241, 0.0923], device='cuda:3'), in_proj_covar=tensor([0.0681, 0.0841, 0.0692, 0.0631, 0.0535, 0.0537, 0.0703, 0.0650], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 05:52:01,990 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1172, 4.2379, 4.0602, 3.8619, 3.5905, 4.2915, 3.9967, 3.9839], device='cuda:3'), covar=tensor([0.1101, 0.1013, 0.0452, 0.0417, 0.1375, 0.0605, 0.1042, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0441, 0.0358, 0.0353, 0.0366, 0.0410, 0.0245, 0.0431], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:52:05,512 INFO [optim.py:368] (3/8) 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 05:52:21,874 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:24,078 INFO [train.py:904] (3/8) Epoch 21, batch 3100, loss[loss=0.1834, simple_loss=0.2582, pruned_loss=0.05432, over 16711.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2558, pruned_loss=0.04135, over 3324551.94 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:52:42,951 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 05:52:43,680 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206117.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:53:11,711 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8424, 4.2661, 4.2755, 3.0725, 3.6653, 4.2337, 3.8698, 2.4676], device='cuda:3'), covar=tensor([0.0457, 0.0063, 0.0045, 0.0342, 0.0112, 0.0099, 0.0077, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0083, 0.0082, 0.0133, 0.0098, 0.0108, 0.0093, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:53:30,990 INFO [train.py:904] (3/8) Epoch 21, batch 3150, loss[loss=0.1932, simple_loss=0.2764, pruned_loss=0.05497, over 16471.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2552, pruned_loss=0.04115, over 3331069.16 frames. ], batch size: 146, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:53:40,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9967, 5.1037, 5.4734, 5.4737, 5.5068, 5.1903, 5.1019, 4.8796], device='cuda:3'), covar=tensor([0.0332, 0.0499, 0.0417, 0.0401, 0.0391, 0.0349, 0.0918, 0.0432], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0465, 0.0453, 0.0420, 0.0498, 0.0474, 0.0568, 0.0380], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 05:53:48,279 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1939, 4.0277, 4.5117, 2.2194, 4.7617, 4.7766, 3.4838, 3.6723], device='cuda:3'), covar=tensor([0.0674, 0.0231, 0.0195, 0.1112, 0.0062, 0.0137, 0.0357, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0139, 0.0080, 0.0125, 0.0128, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 05:53:51,772 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9667, 3.7207, 4.2346, 1.8798, 4.5033, 4.5145, 3.2370, 3.3350], device='cuda:3'), covar=tensor([0.0706, 0.0255, 0.0231, 0.1263, 0.0069, 0.0167, 0.0425, 0.0435], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0109, 0.0098, 0.0139, 0.0080, 0.0125, 0.0128, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 05:54:22,972 INFO [optim.py:368] (3/8) 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 05:54:34,985 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206197.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:54:41,559 INFO [train.py:904] (3/8) Epoch 21, batch 3200, loss[loss=0.1664, simple_loss=0.2675, pruned_loss=0.03268, over 17103.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2547, pruned_loss=0.04093, over 3320795.92 frames. ], batch size: 49, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:54:55,298 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8429, 4.4063, 3.1300, 2.3833, 2.7171, 2.6097, 4.7530, 3.6714], device='cuda:3'), covar=tensor([0.2904, 0.0575, 0.1793, 0.2892, 0.3104, 0.2067, 0.0345, 0.1408], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0271, 0.0305, 0.0310, 0.0298, 0.0259, 0.0296, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:55:42,471 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 3250, loss[loss=0.1542, simple_loss=0.2441, pruned_loss=0.03213, over 17231.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2545, pruned_loss=0.04105, over 3314441.47 frames. ], batch size: 44, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:59,435 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:56:30,680 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6355, 4.7658, 4.8964, 4.7273, 4.7581, 5.3182, 4.8559, 4.5479], device='cuda:3'), covar=tensor([0.1453, 0.2035, 0.2297, 0.2131, 0.2659, 0.1043, 0.1560, 0.2546], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0608, 0.0667, 0.0503, 0.0667, 0.0698, 0.0523, 0.0670], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:56:42,634 INFO [optim.py:368] (3/8) 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 05:56:47,591 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 05:57:00,582 INFO [train.py:904] (3/8) Epoch 21, batch 3300, loss[loss=0.1543, simple_loss=0.244, pruned_loss=0.0323, over 16811.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2554, pruned_loss=0.04132, over 3313142.05 frames. ], batch size: 42, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:57:13,939 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206311.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:58:09,885 INFO [train.py:904] (3/8) Epoch 21, batch 3350, loss[loss=0.1872, simple_loss=0.2687, pruned_loss=0.05288, over 16657.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2561, pruned_loss=0.04167, over 3319906.65 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:58:20,156 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:00,090 INFO [optim.py:368] (3/8) 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 05:59:06,194 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 05:59:12,034 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7532, 4.0124, 3.1246, 2.4158, 2.6720, 2.6709, 4.2458, 3.5610], device='cuda:3'), covar=tensor([0.2891, 0.0599, 0.1631, 0.2890, 0.2634, 0.1896, 0.0477, 0.1367], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0272, 0.0306, 0.0311, 0.0299, 0.0260, 0.0296, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 05:59:16,459 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 3400, loss[loss=0.1935, simple_loss=0.2636, pruned_loss=0.06174, over 16897.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.256, pruned_loss=0.04148, over 3324091.66 frames. ], batch size: 116, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:59:32,319 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:21,656 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:26,892 INFO [train.py:904] (3/8) Epoch 21, batch 3450, loss[loss=0.1581, simple_loss=0.2425, pruned_loss=0.03688, over 16984.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2551, pruned_loss=0.04152, over 3309474.41 frames. ], batch size: 41, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:01:17,175 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.080e+02 2.437e+02 2.944e+02 7.361e+02, threshold=4.875e+02, percent-clipped=3.0 2023-05-01 06:01:36,850 INFO [train.py:904] (3/8) Epoch 21, batch 3500, loss[loss=0.1543, simple_loss=0.2439, pruned_loss=0.03239, over 15902.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.254, pruned_loss=0.04132, over 3307458.44 frames. ], batch size: 35, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:23,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1223, 2.1460, 2.3114, 3.8432, 2.2253, 2.4653, 2.2356, 2.2617], device='cuda:3'), covar=tensor([0.1488, 0.3751, 0.2924, 0.0630, 0.3905, 0.2572, 0.3774, 0.3256], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0451, 0.0370, 0.0334, 0.0439, 0.0519, 0.0420, 0.0528], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:02:29,460 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-05-01 06:02:37,909 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:02:44,994 INFO [train.py:904] (3/8) Epoch 21, batch 3550, loss[loss=0.1583, simple_loss=0.2407, pruned_loss=0.03795, over 16752.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2531, pruned_loss=0.0405, over 3303513.06 frames. ], batch size: 83, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:47,116 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:36,211 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1704, 5.7366, 5.7896, 5.5011, 5.6150, 6.1737, 5.7440, 5.4548], device='cuda:3'), covar=tensor([0.0906, 0.1851, 0.2535, 0.2022, 0.2570, 0.0957, 0.1358, 0.2494], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0607, 0.0667, 0.0503, 0.0669, 0.0697, 0.0521, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:03:37,042 INFO [optim.py:368] (3/8) 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 06:03:44,188 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 3600, loss[loss=0.1751, simple_loss=0.2518, pruned_loss=0.04921, over 16925.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2523, pruned_loss=0.04024, over 3302688.23 frames. ], batch size: 109, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:06,975 INFO [train.py:904] (3/8) Epoch 21, batch 3650, loss[loss=0.1757, simple_loss=0.2528, pruned_loss=0.04928, over 11355.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2514, pruned_loss=0.04087, over 3286050.31 frames. ], batch size: 246, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:59,787 INFO [optim.py:368] (3/8) 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 06:06:09,703 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8645, 4.1599, 3.1633, 2.5238, 2.9098, 2.7556, 4.6818, 3.7330], device='cuda:3'), covar=tensor([0.2759, 0.0749, 0.1678, 0.2566, 0.2582, 0.1879, 0.0348, 0.1219], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0272, 0.0306, 0.0312, 0.0300, 0.0260, 0.0296, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:06:16,325 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 06:06:18,969 INFO [train.py:904] (3/8) Epoch 21, batch 3700, loss[loss=0.1835, simple_loss=0.2515, pruned_loss=0.05778, over 16850.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2505, pruned_loss=0.04238, over 3259409.77 frames. ], batch size: 83, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:06:34,571 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:06:45,529 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2738, 4.1564, 4.3727, 4.4477, 4.5567, 4.1282, 4.3578, 4.5476], device='cuda:3'), covar=tensor([0.1519, 0.1065, 0.1244, 0.0700, 0.0574, 0.1267, 0.2250, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0662, 0.0813, 0.0956, 0.0839, 0.0623, 0.0652, 0.0671, 0.0776], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:07:32,563 INFO [train.py:904] (3/8) Epoch 21, batch 3750, loss[loss=0.1603, simple_loss=0.2352, pruned_loss=0.04269, over 16761.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2511, pruned_loss=0.04368, over 3236481.89 frames. ], batch size: 83, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:07:45,684 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206760.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:08:25,702 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2052, 5.2051, 4.9550, 4.3059, 5.1457, 1.9792, 4.8796, 4.6545], device='cuda:3'), covar=tensor([0.0052, 0.0047, 0.0160, 0.0327, 0.0066, 0.2801, 0.0098, 0.0212], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0157, 0.0200, 0.0180, 0.0179, 0.0209, 0.0189, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:08:26,978 INFO [optim.py:368] (3/8) 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 06:08:31,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9044, 2.9245, 2.7514, 4.6075, 3.6591, 4.2423, 1.8728, 3.1229], device='cuda:3'), covar=tensor([0.1307, 0.0730, 0.1130, 0.0175, 0.0279, 0.0347, 0.1495, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0192, 0.0207, 0.0217, 0.0201, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:08:45,460 INFO [train.py:904] (3/8) Epoch 21, batch 3800, loss[loss=0.1898, simple_loss=0.2541, pruned_loss=0.06273, over 16919.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2522, pruned_loss=0.04472, over 3243164.45 frames. ], batch size: 109, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:09:32,811 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206834.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:09:43,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4376, 3.4949, 3.5594, 2.1527, 3.0841, 2.5519, 3.7721, 3.9343], device='cuda:3'), covar=tensor([0.0179, 0.0724, 0.0609, 0.1948, 0.0820, 0.0905, 0.0480, 0.0790], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0165, 0.0167, 0.0153, 0.0145, 0.0130, 0.0144, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:09:58,314 INFO [train.py:904] (3/8) Epoch 21, batch 3850, loss[loss=0.1763, simple_loss=0.2659, pruned_loss=0.04338, over 16588.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2528, pruned_loss=0.04544, over 3259675.36 frames. ], batch size: 62, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:10:00,721 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 06:11:00,909 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206895.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:11:04,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2266, 5.5417, 5.2776, 5.3291, 5.0148, 4.9624, 4.9476, 5.6474], device='cuda:3'), covar=tensor([0.1350, 0.0780, 0.0928, 0.0799, 0.0840, 0.0835, 0.1067, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0676, 0.0828, 0.0682, 0.0627, 0.0528, 0.0531, 0.0696, 0.0643], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:11:09,938 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:11:10,711 INFO [train.py:904] (3/8) Epoch 21, batch 3900, loss[loss=0.1726, simple_loss=0.2508, pruned_loss=0.0472, over 16221.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2529, pruned_loss=0.04631, over 3268313.76 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:12:24,775 INFO [train.py:904] (3/8) Epoch 21, batch 3950, loss[loss=0.1586, simple_loss=0.2316, pruned_loss=0.04286, over 16680.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2526, pruned_loss=0.04684, over 3264252.72 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:16,330 INFO [optim.py:368] (3/8) 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 06:13:35,022 INFO [train.py:904] (3/8) Epoch 21, batch 4000, loss[loss=0.1706, simple_loss=0.2593, pruned_loss=0.04092, over 15592.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2532, pruned_loss=0.04754, over 3266058.71 frames. ], batch size: 190, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:42,719 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6778, 3.7919, 2.3685, 4.2063, 2.8633, 4.2689, 2.3687, 2.9399], device='cuda:3'), covar=tensor([0.0257, 0.0347, 0.1516, 0.0210, 0.0810, 0.0419, 0.1532, 0.0735], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0179, 0.0195, 0.0166, 0.0179, 0.0219, 0.0202, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:13:42,960 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 06:13:53,408 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1025, 2.2896, 2.7891, 3.1380, 3.0439, 3.6615, 2.3361, 3.4869], device='cuda:3'), covar=tensor([0.0218, 0.0489, 0.0322, 0.0309, 0.0303, 0.0119, 0.0507, 0.0109], device='cuda:3'), in_proj_covar=tensor([0.0190, 0.0195, 0.0181, 0.0186, 0.0199, 0.0156, 0.0199, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:14:03,356 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6801, 2.4951, 2.3823, 3.4550, 2.5982, 3.6912, 1.5655, 2.7046], device='cuda:3'), covar=tensor([0.1452, 0.0870, 0.1275, 0.0235, 0.0203, 0.0394, 0.1736, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0175, 0.0193, 0.0192, 0.0206, 0.0216, 0.0201, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:14:12,225 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9334, 4.3509, 2.7096, 2.5034, 2.7553, 2.5296, 4.4725, 3.6096], device='cuda:3'), covar=tensor([0.2821, 0.0627, 0.2321, 0.2506, 0.2732, 0.2102, 0.0555, 0.1142], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0273, 0.0307, 0.0313, 0.0301, 0.0261, 0.0297, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:14:16,890 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2795, 4.1612, 4.3410, 4.4647, 4.5662, 4.1572, 4.3774, 4.5863], device='cuda:3'), covar=tensor([0.1500, 0.1043, 0.1298, 0.0687, 0.0563, 0.1186, 0.1699, 0.0594], device='cuda:3'), in_proj_covar=tensor([0.0656, 0.0810, 0.0948, 0.0830, 0.0617, 0.0646, 0.0666, 0.0771], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:14:36,346 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-05-01 06:14:45,531 INFO [train.py:904] (3/8) Epoch 21, batch 4050, loss[loss=0.1743, simple_loss=0.2558, pruned_loss=0.04636, over 16703.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2539, pruned_loss=0.04691, over 3264047.17 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:14:53,429 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1529, 4.2966, 4.4647, 4.2165, 4.2884, 4.8010, 4.3619, 4.0440], device='cuda:3'), covar=tensor([0.1888, 0.1970, 0.1914, 0.1903, 0.2605, 0.1064, 0.1486, 0.2318], device='cuda:3'), in_proj_covar=tensor([0.0412, 0.0602, 0.0657, 0.0498, 0.0662, 0.0688, 0.0515, 0.0663], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:15:36,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3564, 4.5008, 4.6329, 4.4603, 4.4862, 5.0093, 4.5007, 4.2138], device='cuda:3'), covar=tensor([0.1466, 0.1839, 0.1833, 0.1717, 0.2386, 0.0904, 0.1554, 0.2297], device='cuda:3'), in_proj_covar=tensor([0.0412, 0.0601, 0.0655, 0.0496, 0.0661, 0.0687, 0.0514, 0.0662], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:15:37,537 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.811e+02 2.022e+02 2.362e+02 4.277e+02, threshold=4.045e+02, percent-clipped=0.0 2023-05-01 06:15:55,998 INFO [train.py:904] (3/8) Epoch 21, batch 4100, loss[loss=0.1809, simple_loss=0.2699, pruned_loss=0.046, over 16178.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2548, pruned_loss=0.04599, over 3270575.14 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:17:10,031 INFO [train.py:904] (3/8) Epoch 21, batch 4150, loss[loss=0.267, simple_loss=0.3337, pruned_loss=0.1001, over 11784.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2622, pruned_loss=0.0486, over 3239450.75 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:17:45,646 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:18:04,515 INFO [optim.py:368] (3/8) 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 06:18:06,805 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207190.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 06:18:24,675 INFO [train.py:904] (3/8) Epoch 21, batch 4200, loss[loss=0.2178, simple_loss=0.3131, pruned_loss=0.06128, over 17088.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2687, pruned_loss=0.04985, over 3225615.58 frames. ], batch size: 48, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:18:45,261 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9375, 5.3424, 5.5424, 5.2597, 5.2480, 5.8654, 5.3913, 5.1521], device='cuda:3'), covar=tensor([0.0973, 0.1755, 0.1694, 0.1759, 0.2419, 0.0773, 0.1343, 0.2282], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0594, 0.0648, 0.0493, 0.0656, 0.0683, 0.0509, 0.0658], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:19:18,482 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:19:32,526 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9537, 4.1811, 3.9896, 4.0793, 3.7950, 3.7872, 3.8175, 4.1795], device='cuda:3'), covar=tensor([0.1007, 0.0892, 0.0968, 0.0749, 0.0790, 0.1688, 0.0961, 0.0976], device='cuda:3'), in_proj_covar=tensor([0.0671, 0.0821, 0.0678, 0.0623, 0.0523, 0.0530, 0.0692, 0.0639], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:19:40,047 INFO [train.py:904] (3/8) Epoch 21, batch 4250, loss[loss=0.1915, simple_loss=0.2713, pruned_loss=0.0558, over 12151.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2721, pruned_loss=0.04971, over 3198019.67 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:20:35,820 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 4300, loss[loss=0.1835, simple_loss=0.2897, pruned_loss=0.03864, over 16838.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2733, pruned_loss=0.04915, over 3199160.83 frames. ], batch size: 102, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:22:07,505 INFO [train.py:904] (3/8) Epoch 21, batch 4350, loss[loss=0.1857, simple_loss=0.274, pruned_loss=0.04874, over 16454.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2763, pruned_loss=0.04984, over 3214461.48 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:02,764 INFO [optim.py:368] (3/8) 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 06:23:17,743 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-05-01 06:23:22,045 INFO [train.py:904] (3/8) Epoch 21, batch 4400, loss[loss=0.1852, simple_loss=0.2804, pruned_loss=0.04501, over 16903.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2788, pruned_loss=0.05102, over 3216078.03 frames. ], batch size: 109, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:23:30,557 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5690, 4.4313, 4.6219, 4.7282, 4.9063, 4.4450, 4.9121, 4.9322], device='cuda:3'), covar=tensor([0.1665, 0.1227, 0.1531, 0.0698, 0.0473, 0.1050, 0.0597, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0633, 0.0786, 0.0915, 0.0803, 0.0600, 0.0626, 0.0644, 0.0749], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:24:37,195 INFO [train.py:904] (3/8) Epoch 21, batch 4450, loss[loss=0.2139, simple_loss=0.3016, pruned_loss=0.06313, over 17135.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2818, pruned_loss=0.0522, over 3205210.82 frames. ], batch size: 48, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:25:31,510 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207490.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 06:25:35,161 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8241, 2.5005, 2.3956, 2.9374, 2.3235, 3.5790, 1.6259, 2.7361], device='cuda:3'), covar=tensor([0.1117, 0.0701, 0.1091, 0.0179, 0.0174, 0.0398, 0.1441, 0.0741], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0191, 0.0208, 0.0215, 0.0201, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:25:50,489 INFO [train.py:904] (3/8) Epoch 21, batch 4500, loss[loss=0.196, simple_loss=0.2825, pruned_loss=0.05473, over 15422.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.282, pruned_loss=0.05289, over 3216685.28 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:26:35,634 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207532.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:26:43,761 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 4550, loss[loss=0.2077, simple_loss=0.2951, pruned_loss=0.06014, over 16539.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2833, pruned_loss=0.05402, over 3229565.28 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:27:05,696 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6038, 3.6568, 2.1781, 4.2258, 2.8244, 4.1610, 2.3568, 2.8828], device='cuda:3'), covar=tensor([0.0257, 0.0347, 0.1637, 0.0129, 0.0825, 0.0462, 0.1403, 0.0770], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0177, 0.0194, 0.0163, 0.0177, 0.0216, 0.0200, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:27:31,105 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5362, 4.6615, 4.8285, 4.5615, 4.6548, 5.2396, 4.6557, 4.4219], device='cuda:3'), covar=tensor([0.1218, 0.1738, 0.2063, 0.2054, 0.2655, 0.0882, 0.1537, 0.2479], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0583, 0.0636, 0.0485, 0.0647, 0.0670, 0.0499, 0.0648], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:27:57,547 INFO [optim.py:368] (3/8) 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 06:28:04,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1525, 3.7461, 3.6794, 2.4066, 3.4047, 3.7114, 3.3386, 2.1235], device='cuda:3'), covar=tensor([0.0552, 0.0042, 0.0051, 0.0409, 0.0087, 0.0086, 0.0096, 0.0467], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0083, 0.0083, 0.0133, 0.0097, 0.0108, 0.0093, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:28:16,259 INFO [train.py:904] (3/8) Epoch 21, batch 4600, loss[loss=0.1862, simple_loss=0.2791, pruned_loss=0.04671, over 16663.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2843, pruned_loss=0.05383, over 3227551.90 frames. ], batch size: 134, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:29:29,137 INFO [train.py:904] (3/8) Epoch 21, batch 4650, loss[loss=0.1844, simple_loss=0.268, pruned_loss=0.05044, over 16726.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2838, pruned_loss=0.05433, over 3217141.56 frames. ], batch size: 124, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:29:35,460 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 06:30:17,684 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7446, 3.1511, 2.7950, 5.0682, 3.9872, 4.2618, 1.8015, 3.0611], device='cuda:3'), covar=tensor([0.1356, 0.0736, 0.1128, 0.0158, 0.0352, 0.0351, 0.1646, 0.0824], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0190, 0.0208, 0.0215, 0.0201, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:30:23,480 INFO [optim.py:368] (3/8) 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 06:30:25,154 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7392, 3.7936, 2.3607, 4.4990, 3.0504, 4.3601, 2.4897, 2.9880], device='cuda:3'), covar=tensor([0.0289, 0.0345, 0.1664, 0.0111, 0.0764, 0.0448, 0.1521, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0193, 0.0161, 0.0176, 0.0215, 0.0199, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:30:41,989 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 06:30:42,424 INFO [train.py:904] (3/8) Epoch 21, batch 4700, loss[loss=0.2039, simple_loss=0.2786, pruned_loss=0.0646, over 11465.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2812, pruned_loss=0.0535, over 3210553.95 frames. ], batch size: 248, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:31:32,603 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3700, 3.3823, 1.9211, 3.7964, 2.5832, 3.7819, 2.2437, 2.7222], device='cuda:3'), covar=tensor([0.0330, 0.0408, 0.1932, 0.0134, 0.0873, 0.0484, 0.1577, 0.0823], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0175, 0.0193, 0.0161, 0.0176, 0.0215, 0.0199, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:31:56,747 INFO [train.py:904] (3/8) Epoch 21, batch 4750, loss[loss=0.1564, simple_loss=0.2381, pruned_loss=0.03739, over 17113.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2762, pruned_loss=0.05067, over 3213572.15 frames. ], batch size: 49, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:32:50,130 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.797e+02 2.146e+02 2.444e+02 4.649e+02, threshold=4.292e+02, percent-clipped=1.0 2023-05-01 06:33:11,158 INFO [train.py:904] (3/8) Epoch 21, batch 4800, loss[loss=0.1791, simple_loss=0.2639, pruned_loss=0.04714, over 12157.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2727, pruned_loss=0.04885, over 3202876.64 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:33:24,325 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6501, 2.5733, 1.8075, 2.7169, 2.1468, 2.7732, 2.1391, 2.3568], device='cuda:3'), covar=tensor([0.0310, 0.0329, 0.1391, 0.0194, 0.0629, 0.0436, 0.1224, 0.0605], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0176, 0.0193, 0.0161, 0.0176, 0.0216, 0.0200, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:33:55,472 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:34:24,539 INFO [train.py:904] (3/8) Epoch 21, batch 4850, loss[loss=0.1938, simple_loss=0.2781, pruned_loss=0.05477, over 12162.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2735, pruned_loss=0.04834, over 3182921.44 frames. ], batch size: 247, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:34:26,649 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5796, 4.8727, 4.6610, 4.6971, 4.4349, 4.3906, 4.3139, 4.9333], device='cuda:3'), covar=tensor([0.1207, 0.0770, 0.0873, 0.0753, 0.0716, 0.1109, 0.1061, 0.0744], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0805, 0.0669, 0.0610, 0.0512, 0.0519, 0.0677, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:35:08,160 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 1.907e+02 2.222e+02 2.565e+02 4.310e+02, threshold=4.444e+02, percent-clipped=1.0 2023-05-01 06:35:40,336 INFO [train.py:904] (3/8) Epoch 21, batch 4900, loss[loss=0.1903, simple_loss=0.2829, pruned_loss=0.04889, over 15405.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2726, pruned_loss=0.04666, over 3189846.16 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:35:49,808 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 06:36:52,780 INFO [train.py:904] (3/8) Epoch 21, batch 4950, loss[loss=0.1822, simple_loss=0.2765, pruned_loss=0.04395, over 16627.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2717, pruned_loss=0.04586, over 3189495.04 frames. ], batch size: 134, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:37:31,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9216, 5.0210, 5.2773, 5.2745, 5.3394, 4.9643, 4.8928, 4.6576], device='cuda:3'), covar=tensor([0.0271, 0.0423, 0.0360, 0.0370, 0.0531, 0.0343, 0.1071, 0.0452], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0444, 0.0431, 0.0399, 0.0473, 0.0451, 0.0538, 0.0362], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 06:37:47,806 INFO [optim.py:368] (3/8) 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 06:38:08,405 INFO [train.py:904] (3/8) Epoch 21, batch 5000, loss[loss=0.1875, simple_loss=0.2808, pruned_loss=0.04706, over 12081.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2737, pruned_loss=0.04596, over 3196094.84 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:38:52,052 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208032.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:39:21,546 INFO [train.py:904] (3/8) Epoch 21, batch 5050, loss[loss=0.1804, simple_loss=0.2723, pruned_loss=0.04426, over 16690.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2743, pruned_loss=0.04596, over 3200916.69 frames. ], batch size: 124, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:40:18,551 INFO [optim.py:368] (3/8) 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 06:40:23,280 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208093.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 06:40:35,283 INFO [train.py:904] (3/8) Epoch 21, batch 5100, loss[loss=0.1684, simple_loss=0.257, pruned_loss=0.03986, over 16723.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2727, pruned_loss=0.04557, over 3184973.31 frames. ], batch size: 76, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:40:48,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3696, 2.9211, 2.5910, 2.2138, 2.1835, 2.2383, 2.8717, 2.8157], device='cuda:3'), covar=tensor([0.2674, 0.0660, 0.1634, 0.2607, 0.2391, 0.2049, 0.0521, 0.1246], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0270, 0.0304, 0.0311, 0.0297, 0.0257, 0.0296, 0.0336], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:41:48,625 INFO [train.py:904] (3/8) Epoch 21, batch 5150, loss[loss=0.1759, simple_loss=0.277, pruned_loss=0.03746, over 16343.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2729, pruned_loss=0.04481, over 3188545.96 frames. ], batch size: 146, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:41:57,848 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5049, 4.6323, 4.8739, 4.8604, 4.8625, 4.5980, 4.5323, 4.4372], device='cuda:3'), covar=tensor([0.0305, 0.0513, 0.0351, 0.0387, 0.0468, 0.0348, 0.0870, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0444, 0.0430, 0.0400, 0.0475, 0.0453, 0.0539, 0.0363], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 06:42:36,088 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5130, 4.6232, 4.7790, 4.5191, 4.6372, 5.1663, 4.6450, 4.3347], device='cuda:3'), covar=tensor([0.1263, 0.1863, 0.2013, 0.2122, 0.2595, 0.0874, 0.1643, 0.2548], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0574, 0.0626, 0.0481, 0.0639, 0.0661, 0.0493, 0.0642], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:42:43,676 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 1.920e+02 2.264e+02 2.595e+02 3.716e+02, threshold=4.527e+02, percent-clipped=0.0 2023-05-01 06:43:01,061 INFO [train.py:904] (3/8) Epoch 21, batch 5200, loss[loss=0.1442, simple_loss=0.2386, pruned_loss=0.0249, over 16873.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2716, pruned_loss=0.04441, over 3204456.09 frames. ], batch size: 96, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:00,499 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4684, 4.0465, 4.0197, 2.7726, 3.5787, 4.0432, 3.5621, 2.2138], device='cuda:3'), covar=tensor([0.0511, 0.0049, 0.0043, 0.0339, 0.0093, 0.0111, 0.0103, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0082, 0.0082, 0.0132, 0.0097, 0.0107, 0.0093, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:44:11,872 INFO [train.py:904] (3/8) Epoch 21, batch 5250, loss[loss=0.1958, simple_loss=0.2779, pruned_loss=0.05687, over 12422.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2694, pruned_loss=0.04446, over 3199349.44 frames. ], batch size: 248, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:31,836 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:44:45,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0133, 2.7349, 2.7965, 2.0785, 2.6079, 2.1473, 2.7342, 2.9011], device='cuda:3'), covar=tensor([0.0302, 0.0687, 0.0611, 0.1744, 0.0829, 0.0906, 0.0584, 0.0663], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0163, 0.0167, 0.0152, 0.0145, 0.0130, 0.0144, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:45:07,260 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 5300, loss[loss=0.1529, simple_loss=0.2319, pruned_loss=0.03697, over 17201.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2658, pruned_loss=0.0435, over 3212139.09 frames. ], batch size: 44, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:45:36,058 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2279, 2.2557, 2.2616, 3.9052, 2.1393, 2.6502, 2.3266, 2.4726], device='cuda:3'), covar=tensor([0.1404, 0.3653, 0.3008, 0.0539, 0.4031, 0.2444, 0.3701, 0.3129], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0445, 0.0365, 0.0327, 0.0434, 0.0514, 0.0415, 0.0520], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:46:00,417 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:46:00,738 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 06:46:05,885 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5744, 3.6551, 3.4376, 3.1049, 3.2626, 3.5128, 3.3650, 3.3536], device='cuda:3'), covar=tensor([0.0576, 0.0553, 0.0294, 0.0259, 0.0558, 0.0464, 0.1296, 0.0440], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0416, 0.0337, 0.0332, 0.0346, 0.0388, 0.0230, 0.0404], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:46:27,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0280, 5.0864, 4.9047, 4.5120, 4.5310, 4.9535, 4.9141, 4.6445], device='cuda:3'), covar=tensor([0.0704, 0.0515, 0.0305, 0.0296, 0.1113, 0.0581, 0.0287, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0416, 0.0338, 0.0333, 0.0347, 0.0389, 0.0231, 0.0405], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:46:37,966 INFO [train.py:904] (3/8) Epoch 21, batch 5350, loss[loss=0.1831, simple_loss=0.2638, pruned_loss=0.0512, over 16599.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2641, pruned_loss=0.04261, over 3208382.42 frames. ], batch size: 57, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:46:57,171 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5847, 2.5685, 2.4579, 3.8823, 2.6692, 3.8787, 1.5024, 2.8144], device='cuda:3'), covar=tensor([0.1369, 0.0812, 0.1208, 0.0157, 0.0181, 0.0345, 0.1691, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0190, 0.0208, 0.0214, 0.0201, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:47:32,375 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208388.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 06:47:34,960 INFO [optim.py:368] (3/8) 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 06:47:53,657 INFO [train.py:904] (3/8) Epoch 21, batch 5400, loss[loss=0.1723, simple_loss=0.2643, pruned_loss=0.04013, over 16442.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2668, pruned_loss=0.04322, over 3201895.46 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:49:10,973 INFO [train.py:904] (3/8) Epoch 21, batch 5450, loss[loss=0.2665, simple_loss=0.3411, pruned_loss=0.09595, over 15308.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.269, pruned_loss=0.04443, over 3197856.18 frames. ], batch size: 190, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:50:04,997 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0347, 3.0836, 1.8396, 3.3114, 2.3840, 3.3212, 2.0952, 2.5640], device='cuda:3'), covar=tensor([0.0304, 0.0411, 0.1660, 0.0208, 0.0808, 0.0677, 0.1428, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0192, 0.0160, 0.0175, 0.0214, 0.0200, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 06:50:09,143 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 5500, loss[loss=0.2458, simple_loss=0.3162, pruned_loss=0.08767, over 11745.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2757, pruned_loss=0.04854, over 3153396.62 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:51:09,375 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7186, 3.5793, 3.7669, 3.5011, 3.8259, 4.1943, 3.8516, 3.5554], device='cuda:3'), covar=tensor([0.2325, 0.2685, 0.2756, 0.2862, 0.2810, 0.2227, 0.1802, 0.2827], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0578, 0.0630, 0.0483, 0.0642, 0.0666, 0.0496, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 06:51:47,645 INFO [train.py:904] (3/8) Epoch 21, batch 5550, loss[loss=0.2189, simple_loss=0.2984, pruned_loss=0.06968, over 16602.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2831, pruned_loss=0.05382, over 3124922.15 frames. ], batch size: 62, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:52:25,666 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 06:52:27,834 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208576.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:52:49,283 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 3.115e+02 3.921e+02 4.748e+02 1.231e+03, threshold=7.841e+02, percent-clipped=12.0 2023-05-01 06:53:07,717 INFO [train.py:904] (3/8) Epoch 21, batch 5600, loss[loss=0.2883, simple_loss=0.3487, pruned_loss=0.1139, over 11163.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2886, pruned_loss=0.05799, over 3106770.84 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:53:39,495 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:06,475 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208637.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:18,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9289, 2.1822, 2.4304, 3.1483, 2.2388, 2.3826, 2.3825, 2.2981], device='cuda:3'), covar=tensor([0.1226, 0.2868, 0.2125, 0.0694, 0.3627, 0.2139, 0.2773, 0.2764], device='cuda:3'), in_proj_covar=tensor([0.0397, 0.0441, 0.0361, 0.0323, 0.0429, 0.0507, 0.0410, 0.0514], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 06:54:29,857 INFO [train.py:904] (3/8) Epoch 21, batch 5650, loss[loss=0.2019, simple_loss=0.2911, pruned_loss=0.05642, over 16455.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2932, pruned_loss=0.06122, over 3107946.98 frames. ], batch size: 68, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:55:28,093 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:55:30,678 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 5700, loss[loss=0.1804, simple_loss=0.2764, pruned_loss=0.04222, over 16515.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.294, pruned_loss=0.06212, over 3102587.08 frames. ], batch size: 75, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:56:12,143 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 06:56:45,711 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:57:10,976 INFO [train.py:904] (3/8) Epoch 21, batch 5750, loss[loss=0.2463, simple_loss=0.3109, pruned_loss=0.09083, over 11154.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2969, pruned_loss=0.06391, over 3078441.93 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:57:52,435 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 06:58:13,749 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 06:58:13,920 INFO [optim.py:368] (3/8) 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 06:58:33,852 INFO [train.py:904] (3/8) Epoch 21, batch 5800, loss[loss=0.2244, simple_loss=0.2973, pruned_loss=0.07572, over 11957.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2972, pruned_loss=0.06357, over 3071909.29 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:58:38,587 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 06:59:52,595 INFO [train.py:904] (3/8) Epoch 21, batch 5850, loss[loss=0.1939, simple_loss=0.2762, pruned_loss=0.05577, over 16971.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2948, pruned_loss=0.06176, over 3080675.62 frames. ], batch size: 41, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:00:53,525 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.740e+02 3.068e+02 3.692e+02 6.527e+02, threshold=6.136e+02, percent-clipped=0.0 2023-05-01 07:01:12,710 INFO [train.py:904] (3/8) Epoch 21, batch 5900, loss[loss=0.2485, simple_loss=0.3137, pruned_loss=0.09164, over 11563.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2952, pruned_loss=0.06178, over 3103855.81 frames. ], batch size: 247, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:01:48,500 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:04,489 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208932.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:36,229 INFO [train.py:904] (3/8) Epoch 21, batch 5950, loss[loss=0.1869, simple_loss=0.2865, pruned_loss=0.04365, over 16491.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.295, pruned_loss=0.05989, over 3132637.75 frames. ], batch size: 68, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:02,962 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:03:36,545 INFO [optim.py:368] (3/8) 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 07:03:44,574 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-05-01 07:03:54,096 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4781, 5.4703, 5.2126, 4.5581, 5.4070, 2.0205, 5.0743, 5.0404], device='cuda:3'), covar=tensor([0.0060, 0.0055, 0.0168, 0.0389, 0.0069, 0.2517, 0.0114, 0.0186], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0150, 0.0192, 0.0173, 0.0170, 0.0201, 0.0181, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:03:56,989 INFO [train.py:904] (3/8) Epoch 21, batch 6000, loss[loss=0.2101, simple_loss=0.2998, pruned_loss=0.0602, over 16988.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2941, pruned_loss=0.05992, over 3114149.62 frames. ], batch size: 55, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:56,989 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 07:04:08,271 INFO [train.py:938] (3/8) Epoch 21, validation: loss=0.1512, simple_loss=0.2639, pruned_loss=0.01924, over 944034.00 frames. 2023-05-01 07:04:08,272 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 07:04:52,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-05-01 07:05:23,557 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-01 07:05:25,681 INFO [train.py:904] (3/8) Epoch 21, batch 6050, loss[loss=0.2098, simple_loss=0.3026, pruned_loss=0.05851, over 16629.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2923, pruned_loss=0.05895, over 3122592.70 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:06:26,994 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 6100, loss[loss=0.1882, simple_loss=0.2789, pruned_loss=0.04876, over 16909.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2922, pruned_loss=0.05868, over 3110940.26 frames. ], batch size: 96, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:07:29,431 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9952, 3.8783, 4.0387, 4.1528, 4.2544, 3.8705, 4.2042, 4.2756], device='cuda:3'), covar=tensor([0.1586, 0.1180, 0.1320, 0.0657, 0.0565, 0.1586, 0.0835, 0.0662], device='cuda:3'), in_proj_covar=tensor([0.0622, 0.0767, 0.0896, 0.0782, 0.0588, 0.0618, 0.0636, 0.0736], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:07:55,595 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7523, 4.1628, 2.9894, 2.3583, 2.7939, 2.6515, 4.4664, 3.5746], device='cuda:3'), covar=tensor([0.2822, 0.0564, 0.1751, 0.2551, 0.2591, 0.1833, 0.0405, 0.1284], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0269, 0.0303, 0.0309, 0.0295, 0.0256, 0.0294, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 07:08:04,715 INFO [train.py:904] (3/8) Epoch 21, batch 6150, loss[loss=0.2557, simple_loss=0.3241, pruned_loss=0.09365, over 11577.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2902, pruned_loss=0.05779, over 3118503.01 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:08:10,663 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9031, 2.7408, 2.8161, 2.1502, 2.6744, 2.1262, 2.7642, 2.9551], device='cuda:3'), covar=tensor([0.0257, 0.0732, 0.0512, 0.1680, 0.0769, 0.0869, 0.0516, 0.0642], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0142, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:09:04,296 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 2.685e+02 3.365e+02 4.144e+02 7.779e+02, threshold=6.731e+02, percent-clipped=2.0 2023-05-01 07:09:23,293 INFO [train.py:904] (3/8) Epoch 21, batch 6200, loss[loss=0.2162, simple_loss=0.3066, pruned_loss=0.06289, over 16900.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2885, pruned_loss=0.05714, over 3113675.76 frames. ], batch size: 109, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:23,828 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:10,032 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209232.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:39,923 INFO [train.py:904] (3/8) Epoch 21, batch 6250, loss[loss=0.2087, simple_loss=0.2845, pruned_loss=0.06645, over 11747.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2887, pruned_loss=0.05729, over 3094679.59 frames. ], batch size: 247, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:10:45,605 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:57,638 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209263.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:22,157 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209280.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:35,896 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.681e+02 3.167e+02 3.896e+02 8.805e+02, threshold=6.333e+02, percent-clipped=2.0 2023-05-01 07:11:49,869 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 6300, loss[loss=0.2121, simple_loss=0.2967, pruned_loss=0.06372, over 16185.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2884, pruned_loss=0.05664, over 3119239.47 frames. ], batch size: 165, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:12:17,938 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:13:13,590 INFO [train.py:904] (3/8) Epoch 21, batch 6350, loss[loss=0.2056, simple_loss=0.2879, pruned_loss=0.06161, over 15421.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2892, pruned_loss=0.05763, over 3114745.02 frames. ], batch size: 190, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:13:24,829 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:14:13,986 INFO [optim.py:368] (3/8) 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 07:14:31,828 INFO [train.py:904] (3/8) Epoch 21, batch 6400, loss[loss=0.198, simple_loss=0.284, pruned_loss=0.05602, over 16733.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2898, pruned_loss=0.0591, over 3100950.55 frames. ], batch size: 76, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:15:35,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7088, 3.6847, 2.3523, 4.2490, 2.8412, 4.2113, 2.4515, 2.9636], device='cuda:3'), covar=tensor([0.0312, 0.0469, 0.1676, 0.0276, 0.0853, 0.0718, 0.1534, 0.0809], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0177, 0.0194, 0.0161, 0.0176, 0.0216, 0.0201, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:15:41,428 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5753, 2.6248, 2.4076, 3.9751, 2.9590, 3.8996, 1.3809, 3.0037], device='cuda:3'), covar=tensor([0.1466, 0.0803, 0.1362, 0.0170, 0.0261, 0.0483, 0.1875, 0.0812], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0174, 0.0194, 0.0189, 0.0206, 0.0215, 0.0201, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:15:48,022 INFO [train.py:904] (3/8) Epoch 21, batch 6450, loss[loss=0.2397, simple_loss=0.3016, pruned_loss=0.08891, over 11677.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2908, pruned_loss=0.05971, over 3071530.76 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:15:53,560 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8869, 3.5777, 4.0689, 2.0517, 4.3294, 4.2698, 3.1530, 3.1929], device='cuda:3'), covar=tensor([0.0690, 0.0265, 0.0191, 0.1125, 0.0063, 0.0184, 0.0370, 0.0434], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0106, 0.0095, 0.0136, 0.0078, 0.0122, 0.0126, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 07:16:10,759 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209466.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:16:11,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9159, 4.9350, 4.7850, 4.5321, 4.4967, 4.8725, 4.7589, 4.5815], device='cuda:3'), covar=tensor([0.0654, 0.0671, 0.0314, 0.0314, 0.0911, 0.0509, 0.0418, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0285, 0.0413, 0.0333, 0.0330, 0.0341, 0.0383, 0.0229, 0.0400], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:16:33,106 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 07:16:52,760 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.025e+02 2.825e+02 3.422e+02 4.079e+02 7.372e+02, threshold=6.844e+02, percent-clipped=3.0 2023-05-01 07:17:08,455 INFO [train.py:904] (3/8) Epoch 21, batch 6500, loss[loss=0.2053, simple_loss=0.2778, pruned_loss=0.0664, over 11580.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.289, pruned_loss=0.05918, over 3075225.79 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:17:20,816 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:17:48,037 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 6550, loss[loss=0.2113, simple_loss=0.3117, pruned_loss=0.05548, over 17091.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2914, pruned_loss=0.05931, over 3098939.70 frames. ], batch size: 53, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:18:40,049 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:18:41,783 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 07:18:57,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7903, 2.2946, 1.9754, 1.9878, 2.6900, 2.3418, 2.5261, 2.7963], device='cuda:3'), covar=tensor([0.0215, 0.0430, 0.0506, 0.0505, 0.0256, 0.0391, 0.0239, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0203, 0.0230, 0.0221, 0.0222, 0.0231, 0.0229, 0.0230, 0.0226], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:19:00,594 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209571.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:22,027 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 07:19:33,742 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 6600, loss[loss=0.222, simple_loss=0.3121, pruned_loss=0.06595, over 16349.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2927, pruned_loss=0.05957, over 3114292.62 frames. ], batch size: 146, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:19:56,035 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:07,717 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 07:21:07,995 INFO [train.py:904] (3/8) Epoch 21, batch 6650, loss[loss=0.1911, simple_loss=0.285, pruned_loss=0.0486, over 16488.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2936, pruned_loss=0.06075, over 3108448.20 frames. ], batch size: 75, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:21:12,196 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209654.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:32,348 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209667.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:04,997 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:11,939 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 6700, loss[loss=0.1858, simple_loss=0.2749, pruned_loss=0.04831, over 16503.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2926, pruned_loss=0.06088, over 3097730.16 frames. ], batch size: 146, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:22:31,724 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 07:23:40,251 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 6750, loss[loss=0.2121, simple_loss=0.2918, pruned_loss=0.06618, over 16634.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2918, pruned_loss=0.06087, over 3099101.57 frames. ], batch size: 62, lr: 3.20e-03, grad_scale: 2.0 2023-05-01 07:24:47,245 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 6800, loss[loss=0.1993, simple_loss=0.2922, pruned_loss=0.05326, over 16895.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2918, pruned_loss=0.06049, over 3119114.96 frames. ], batch size: 109, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:25:33,965 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:19,706 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0500, 4.1659, 2.4874, 4.6288, 3.0722, 4.5592, 2.5973, 3.2021], device='cuda:3'), covar=tensor([0.0218, 0.0313, 0.1715, 0.0287, 0.0806, 0.0648, 0.1614, 0.0862], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:26:20,405 INFO [train.py:904] (3/8) Epoch 21, batch 6850, loss[loss=0.1866, simple_loss=0.2853, pruned_loss=0.04397, over 17118.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2932, pruned_loss=0.06086, over 3123178.21 frames. ], batch size: 49, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:26:29,587 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:30,263 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 07:26:42,230 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:21,615 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 6900, loss[loss=0.1903, simple_loss=0.2815, pruned_loss=0.04951, over 16446.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2943, pruned_loss=0.05892, over 3154105.24 frames. ], batch size: 146, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:27:41,303 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:48,361 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:28:51,094 INFO [train.py:904] (3/8) Epoch 21, batch 6950, loss[loss=0.2299, simple_loss=0.3125, pruned_loss=0.07364, over 16311.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2958, pruned_loss=0.06023, over 3156916.06 frames. ], batch size: 35, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:28:53,937 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:06,371 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.035e+02 3.251e+02 3.766e+02 4.564e+02 9.580e+02, threshold=7.532e+02, percent-clipped=9.0 2023-05-01 07:30:09,077 INFO [train.py:904] (3/8) Epoch 21, batch 7000, loss[loss=0.1946, simple_loss=0.3, pruned_loss=0.04458, over 16761.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2962, pruned_loss=0.05983, over 3156981.28 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:30:09,334 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210002.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:30:49,950 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3523, 2.4312, 2.3751, 4.3448, 2.3697, 2.9097, 2.4561, 2.5908], device='cuda:3'), covar=tensor([0.1330, 0.3393, 0.2811, 0.0446, 0.3770, 0.2261, 0.3439, 0.3084], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0444, 0.0364, 0.0324, 0.0434, 0.0510, 0.0414, 0.0518], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:31:11,217 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 7050, loss[loss=0.2065, simple_loss=0.2972, pruned_loss=0.05793, over 16481.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2971, pruned_loss=0.05997, over 3146062.95 frames. ], batch size: 75, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:31:30,583 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:32:37,546 INFO [train.py:904] (3/8) Epoch 21, batch 7100, loss[loss=0.1949, simple_loss=0.2821, pruned_loss=0.05382, over 16490.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.295, pruned_loss=0.06008, over 3122408.06 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:33:03,290 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:08,725 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 07:33:09,193 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:11,648 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 07:33:55,429 INFO [train.py:904] (3/8) Epoch 21, batch 7150, loss[loss=0.219, simple_loss=0.3013, pruned_loss=0.06837, over 15305.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2936, pruned_loss=0.06028, over 3119209.76 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:34:16,438 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210170.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:47,416 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 07:34:53,944 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 7200, loss[loss=0.1962, simple_loss=0.2828, pruned_loss=0.05479, over 16391.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2908, pruned_loss=0.05834, over 3110656.17 frames. ], batch size: 146, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:35:15,418 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 07:35:26,244 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:36:03,735 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 07:36:14,473 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-05-01 07:36:28,284 INFO [train.py:904] (3/8) Epoch 21, batch 7250, loss[loss=0.1787, simple_loss=0.2646, pruned_loss=0.04641, over 16195.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2886, pruned_loss=0.05706, over 3107507.76 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:36:42,613 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3608, 3.3132, 3.3474, 3.4750, 3.4868, 3.2401, 3.4403, 3.5374], device='cuda:3'), covar=tensor([0.1297, 0.1103, 0.1262, 0.0773, 0.0823, 0.2852, 0.1220, 0.0949], device='cuda:3'), in_proj_covar=tensor([0.0623, 0.0768, 0.0891, 0.0783, 0.0589, 0.0618, 0.0638, 0.0735], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:36:43,759 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210262.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:37:31,772 INFO [optim.py:368] (3/8) 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 07:37:44,653 INFO [train.py:904] (3/8) Epoch 21, batch 7300, loss[loss=0.2591, simple_loss=0.3148, pruned_loss=0.1017, over 11535.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2883, pruned_loss=0.05732, over 3099965.38 frames. ], batch size: 246, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:37:58,789 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:48,805 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210344.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:59,603 INFO [train.py:904] (3/8) Epoch 21, batch 7350, loss[loss=0.186, simple_loss=0.2774, pruned_loss=0.04724, over 17213.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2891, pruned_loss=0.05829, over 3070931.62 frames. ], batch size: 46, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:39:31,677 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7673, 1.8360, 1.6224, 1.4152, 1.9201, 1.5419, 1.6569, 1.8438], device='cuda:3'), covar=tensor([0.0169, 0.0291, 0.0385, 0.0336, 0.0190, 0.0278, 0.0142, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0229, 0.0222, 0.0223, 0.0230, 0.0228, 0.0229, 0.0225], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:39:37,842 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 07:40:00,127 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:02,081 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.648e+02 3.204e+02 3.841e+02 6.410e+02, threshold=6.409e+02, percent-clipped=2.0 2023-05-01 07:40:14,853 INFO [train.py:904] (3/8) Epoch 21, batch 7400, loss[loss=0.2121, simple_loss=0.301, pruned_loss=0.0616, over 16752.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2903, pruned_loss=0.05921, over 3047599.62 frames. ], batch size: 89, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:40:20,257 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-05-01 07:40:32,233 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210413.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:41:02,745 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-01 07:41:32,223 INFO [train.py:904] (3/8) Epoch 21, batch 7450, loss[loss=0.2198, simple_loss=0.3135, pruned_loss=0.06299, over 16796.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2922, pruned_loss=0.06062, over 3037449.44 frames. ], batch size: 124, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:41:57,707 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7238, 3.0914, 3.2711, 2.0012, 2.8112, 2.0954, 3.2723, 3.3222], device='cuda:3'), covar=tensor([0.0305, 0.0794, 0.0581, 0.2121, 0.0884, 0.1118, 0.0711, 0.0904], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0143, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:42:42,641 INFO [optim.py:368] (3/8) 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 07:42:48,834 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6744, 4.6861, 4.5278, 3.7609, 4.5823, 1.7414, 4.3675, 4.2438], device='cuda:3'), covar=tensor([0.0091, 0.0088, 0.0185, 0.0344, 0.0099, 0.2852, 0.0123, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0149, 0.0192, 0.0173, 0.0170, 0.0202, 0.0181, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:42:53,262 INFO [train.py:904] (3/8) Epoch 21, batch 7500, loss[loss=0.2046, simple_loss=0.2912, pruned_loss=0.05896, over 16772.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.292, pruned_loss=0.05983, over 3032948.58 frames. ], batch size: 124, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:43:25,840 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 07:43:47,671 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8495, 2.6668, 2.6652, 1.8942, 2.5911, 2.7009, 2.5495, 1.9046], device='cuda:3'), covar=tensor([0.0458, 0.0091, 0.0089, 0.0390, 0.0128, 0.0134, 0.0129, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0134, 0.0096, 0.0107, 0.0093, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 07:44:09,157 INFO [train.py:904] (3/8) Epoch 21, batch 7550, loss[loss=0.1963, simple_loss=0.2844, pruned_loss=0.05408, over 16642.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2914, pruned_loss=0.06036, over 3030854.78 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:45:11,351 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 7600, loss[loss=0.1882, simple_loss=0.2713, pruned_loss=0.05251, over 17014.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2902, pruned_loss=0.06058, over 3023874.81 frames. ], batch size: 55, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:45:47,420 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-01 07:46:02,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7773, 1.2859, 1.7162, 1.6348, 1.7900, 1.9239, 1.6575, 1.8117], device='cuda:3'), covar=tensor([0.0295, 0.0403, 0.0224, 0.0299, 0.0268, 0.0165, 0.0427, 0.0146], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0191, 0.0176, 0.0182, 0.0194, 0.0150, 0.0194, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:46:08,275 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210632.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:46:37,229 INFO [train.py:904] (3/8) Epoch 21, batch 7650, loss[loss=0.2242, simple_loss=0.3195, pruned_loss=0.06441, over 16596.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2911, pruned_loss=0.06188, over 3011687.25 frames. ], batch size: 62, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:59,511 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210666.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:25,955 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,134 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,757 INFO [optim.py:368] (3/8) 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 07:47:44,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7921, 2.4654, 2.2855, 3.2528, 2.1926, 3.5300, 1.5164, 2.7840], device='cuda:3'), covar=tensor([0.1346, 0.0797, 0.1322, 0.0202, 0.0164, 0.0383, 0.1780, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0196, 0.0190, 0.0208, 0.0216, 0.0202, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:47:55,351 INFO [train.py:904] (3/8) Epoch 21, batch 7700, loss[loss=0.1868, simple_loss=0.2765, pruned_loss=0.04854, over 16644.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2914, pruned_loss=0.06202, over 3013946.37 frames. ], batch size: 76, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:48:12,273 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:34,126 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:58,947 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:12,139 INFO [scaling.py:679] (3/8) 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] (3/8) Epoch 21, batch 7750, loss[loss=0.1758, simple_loss=0.2709, pruned_loss=0.04035, over 17192.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2914, pruned_loss=0.06162, over 3025625.15 frames. ], batch size: 46, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:49:27,324 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210761.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:33,010 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210764.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:36,112 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6708, 4.7361, 4.5723, 4.2666, 4.2541, 4.6597, 4.4216, 4.3655], device='cuda:3'), covar=tensor([0.0625, 0.0589, 0.0266, 0.0291, 0.0852, 0.0541, 0.0458, 0.0634], device='cuda:3'), in_proj_covar=tensor([0.0279, 0.0405, 0.0326, 0.0320, 0.0335, 0.0375, 0.0226, 0.0391], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 07:49:36,623 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 07:50:18,619 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 7800, loss[loss=0.1818, simple_loss=0.2704, pruned_loss=0.04657, over 16931.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2917, pruned_loss=0.0615, over 3042046.56 frames. ], batch size: 109, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:51:07,568 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 7850, loss[loss=0.1933, simple_loss=0.284, pruned_loss=0.05135, over 16590.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.292, pruned_loss=0.06072, over 3065455.06 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:52:54,099 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 7900, loss[loss=0.2415, simple_loss=0.3076, pruned_loss=0.08771, over 11732.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2905, pruned_loss=0.06007, over 3061794.92 frames. ], batch size: 249, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:53:59,037 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 07:54:03,492 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-05-01 07:54:24,386 INFO [train.py:904] (3/8) Epoch 21, batch 7950, loss[loss=0.2087, simple_loss=0.2933, pruned_loss=0.06203, over 16850.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2915, pruned_loss=0.06125, over 3040511.76 frames. ], batch size: 96, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:54:33,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4990, 4.5507, 4.8852, 4.8395, 4.8867, 4.5676, 4.5537, 4.4076], device='cuda:3'), covar=tensor([0.0300, 0.0557, 0.0349, 0.0436, 0.0451, 0.0408, 0.0937, 0.0523], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0443, 0.0431, 0.0401, 0.0478, 0.0453, 0.0539, 0.0361], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 07:55:14,483 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:20,868 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:29,109 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 8000, loss[loss=0.1984, simple_loss=0.2895, pruned_loss=0.05363, over 17137.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2929, pruned_loss=0.06235, over 3039717.94 frames. ], batch size: 47, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:56:12,595 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:35,345 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 21, batch 8050, loss[loss=0.2204, simple_loss=0.304, pruned_loss=0.06846, over 16943.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2922, pruned_loss=0.06115, over 3073859.40 frames. ], batch size: 109, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:57:56,769 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5252, 3.3289, 3.6859, 1.9150, 3.8391, 3.8743, 2.9136, 2.8240], device='cuda:3'), covar=tensor([0.0772, 0.0255, 0.0182, 0.1189, 0.0074, 0.0186, 0.0451, 0.0497], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0107, 0.0096, 0.0137, 0.0079, 0.0123, 0.0128, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 07:57:59,279 INFO [optim.py:368] (3/8) 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 07:58:08,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0334, 3.0643, 1.9445, 3.2458, 2.3755, 3.3266, 2.1535, 2.5825], device='cuda:3'), covar=tensor([0.0316, 0.0383, 0.1528, 0.0232, 0.0795, 0.0591, 0.1384, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0192, 0.0158, 0.0174, 0.0214, 0.0199, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 07:58:10,495 INFO [train.py:904] (3/8) Epoch 21, batch 8100, loss[loss=0.2103, simple_loss=0.2997, pruned_loss=0.06049, over 16924.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2916, pruned_loss=0.06015, over 3091028.32 frames. ], batch size: 109, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:58:38,222 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211120.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:59:05,850 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0819, 3.6502, 3.6678, 2.2206, 3.3351, 3.6602, 3.3469, 2.0822], device='cuda:3'), covar=tensor([0.0598, 0.0059, 0.0061, 0.0484, 0.0109, 0.0121, 0.0104, 0.0484], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0082, 0.0083, 0.0134, 0.0096, 0.0108, 0.0093, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 07:59:22,868 INFO [train.py:904] (3/8) Epoch 21, batch 8150, loss[loss=0.1722, simple_loss=0.2605, pruned_loss=0.04194, over 16854.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2886, pruned_loss=0.05899, over 3079421.59 frames. ], batch size: 96, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:59:36,383 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 08:00:19,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8172, 3.7227, 3.8850, 4.0046, 4.0851, 3.6950, 4.0013, 4.0993], device='cuda:3'), covar=tensor([0.1654, 0.1144, 0.1221, 0.0665, 0.0611, 0.2025, 0.0895, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0616, 0.0759, 0.0879, 0.0772, 0.0581, 0.0611, 0.0631, 0.0728], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:00:27,467 INFO [optim.py:368] (3/8) 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 08:00:34,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5315, 5.5437, 5.3490, 4.6293, 5.4533, 2.0666, 5.1729, 5.1272], device='cuda:3'), covar=tensor([0.0103, 0.0086, 0.0206, 0.0466, 0.0092, 0.2611, 0.0138, 0.0191], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0151, 0.0195, 0.0175, 0.0172, 0.0204, 0.0182, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:00:40,743 INFO [train.py:904] (3/8) Epoch 21, batch 8200, loss[loss=0.2174, simple_loss=0.2861, pruned_loss=0.07435, over 11636.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2865, pruned_loss=0.05918, over 3056233.69 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:01:59,680 INFO [train.py:904] (3/8) Epoch 21, batch 8250, loss[loss=0.1941, simple_loss=0.2938, pruned_loss=0.04715, over 16191.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2859, pruned_loss=0.05689, over 3032298.47 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:02:37,445 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 08:02:56,783 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.346e+02 2.848e+02 3.670e+02 8.296e+02, threshold=5.695e+02, percent-clipped=3.0 2023-05-01 08:03:18,674 INFO [train.py:904] (3/8) Epoch 21, batch 8300, loss[loss=0.167, simple_loss=0.2693, pruned_loss=0.03238, over 16706.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2831, pruned_loss=0.05376, over 3038311.85 frames. ], batch size: 89, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:03:51,699 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211322.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:12,347 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211336.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:15,986 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:18,456 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:36,925 INFO [train.py:904] (3/8) Epoch 21, batch 8350, loss[loss=0.2445, simple_loss=0.3116, pruned_loss=0.08871, over 12097.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2827, pruned_loss=0.05204, over 3040291.00 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:05:05,458 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,343 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,500 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 21, batch 8400, loss[loss=0.1787, simple_loss=0.2694, pruned_loss=0.04403, over 16878.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2798, pruned_loss=0.04982, over 3046284.69 frames. ], batch size: 116, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:06:10,911 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:06:16,954 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-01 08:06:22,462 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:03,733 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 8450, loss[loss=0.1678, simple_loss=0.2651, pruned_loss=0.03523, over 15444.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2774, pruned_loss=0.04783, over 3040872.17 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:07:34,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8704, 2.9825, 2.7025, 4.7467, 3.3821, 4.3848, 1.7534, 3.2714], device='cuda:3'), covar=tensor([0.1297, 0.0722, 0.1083, 0.0179, 0.0165, 0.0336, 0.1577, 0.0655], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0187, 0.0204, 0.0213, 0.0200, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 08:07:36,103 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:44,868 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211473.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:08:18,590 INFO [optim.py:368] (3/8) 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 08:08:27,697 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 08:08:29,988 INFO [train.py:904] (3/8) Epoch 21, batch 8500, loss[loss=0.1642, simple_loss=0.244, pruned_loss=0.04223, over 11879.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2734, pruned_loss=0.04531, over 3037729.55 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:08:47,701 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3491, 4.3272, 4.6686, 4.6616, 4.6541, 4.4044, 4.3636, 4.2958], device='cuda:3'), covar=tensor([0.0328, 0.0701, 0.0448, 0.0440, 0.0490, 0.0459, 0.1013, 0.0546], device='cuda:3'), in_proj_covar=tensor([0.0397, 0.0443, 0.0428, 0.0398, 0.0476, 0.0449, 0.0535, 0.0360], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 08:09:54,385 INFO [train.py:904] (3/8) Epoch 21, batch 8550, loss[loss=0.1661, simple_loss=0.2692, pruned_loss=0.03153, over 16843.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2711, pruned_loss=0.04417, over 3027485.25 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:10:34,206 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9653, 2.7332, 2.8861, 2.1100, 2.6431, 2.1486, 2.6806, 2.9480], device='cuda:3'), covar=tensor([0.0359, 0.0956, 0.0564, 0.1923, 0.0871, 0.1013, 0.0707, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0159, 0.0163, 0.0150, 0.0141, 0.0126, 0.0140, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 08:11:18,418 INFO [optim.py:368] (3/8) 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 08:11:28,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 08:11:32,287 INFO [train.py:904] (3/8) Epoch 21, batch 8600, loss[loss=0.1851, simple_loss=0.2779, pruned_loss=0.04613, over 15306.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2709, pruned_loss=0.04333, over 3004365.69 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:11:41,083 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7215, 2.6048, 2.4676, 3.9970, 2.4226, 3.9490, 1.5198, 2.8706], device='cuda:3'), covar=tensor([0.1377, 0.0765, 0.1204, 0.0159, 0.0122, 0.0324, 0.1660, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0172, 0.0193, 0.0186, 0.0203, 0.0212, 0.0199, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 08:11:43,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9487, 4.2635, 4.0812, 4.0914, 3.7796, 3.8581, 3.8866, 4.2583], device='cuda:3'), covar=tensor([0.1289, 0.1027, 0.1010, 0.1033, 0.0971, 0.1910, 0.1065, 0.1109], device='cuda:3'), in_proj_covar=tensor([0.0646, 0.0779, 0.0651, 0.0596, 0.0496, 0.0510, 0.0657, 0.0616], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:12:00,657 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4643, 2.6522, 2.2171, 2.3896, 2.9175, 2.6741, 3.0212, 3.2113], device='cuda:3'), covar=tensor([0.0140, 0.0395, 0.0523, 0.0451, 0.0290, 0.0351, 0.0231, 0.0256], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0225, 0.0218, 0.0218, 0.0225, 0.0224, 0.0224, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:12:30,105 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4248, 3.0042, 2.7006, 2.2634, 2.1813, 2.3170, 2.9947, 2.8249], device='cuda:3'), covar=tensor([0.2772, 0.0701, 0.1698, 0.3174, 0.2843, 0.2270, 0.0544, 0.1614], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0264, 0.0301, 0.0308, 0.0292, 0.0255, 0.0290, 0.0330], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 08:12:48,731 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:13:09,685 INFO [train.py:904] (3/8) Epoch 21, batch 8650, loss[loss=0.1557, simple_loss=0.2574, pruned_loss=0.02703, over 15425.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2691, pruned_loss=0.04212, over 2992533.90 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:14:30,915 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:14:57,320 INFO [train.py:904] (3/8) Epoch 21, batch 8700, loss[loss=0.1641, simple_loss=0.2572, pruned_loss=0.03556, over 16847.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2666, pruned_loss=0.04095, over 3012328.29 frames. ], batch size: 116, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:15:21,116 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0164, 5.3043, 5.0327, 5.0849, 4.8574, 4.8460, 4.7375, 5.3672], device='cuda:3'), covar=tensor([0.1136, 0.0860, 0.1038, 0.0834, 0.0831, 0.0885, 0.1157, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0647, 0.0781, 0.0653, 0.0598, 0.0498, 0.0510, 0.0658, 0.0617], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:16:03,005 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6238, 4.0902, 3.6361, 3.9652, 3.5986, 3.6750, 3.6666, 4.0554], device='cuda:3'), covar=tensor([0.3067, 0.1739, 0.2956, 0.1685, 0.2069, 0.3289, 0.2501, 0.1963], device='cuda:3'), in_proj_covar=tensor([0.0648, 0.0782, 0.0654, 0.0599, 0.0498, 0.0511, 0.0658, 0.0617], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:16:10,366 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 8750, loss[loss=0.1822, simple_loss=0.2807, pruned_loss=0.04185, over 16729.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2664, pruned_loss=0.04022, over 3029073.17 frames. ], batch size: 124, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:17:13,814 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 21, batch 8800, loss[loss=0.1383, simple_loss=0.2353, pruned_loss=0.02066, over 17122.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2647, pruned_loss=0.03887, over 3043060.85 frames. ], batch size: 49, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:19:10,028 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5099, 2.5646, 2.2835, 4.0347, 2.4263, 3.8595, 1.4403, 2.6786], device='cuda:3'), covar=tensor([0.1724, 0.0895, 0.1438, 0.0217, 0.0156, 0.0410, 0.1972, 0.0924], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0170, 0.0190, 0.0183, 0.0200, 0.0210, 0.0198, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 08:20:07,833 INFO [train.py:904] (3/8) Epoch 21, batch 8850, loss[loss=0.1565, simple_loss=0.2486, pruned_loss=0.03223, over 12114.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2665, pruned_loss=0.03817, over 3025224.17 frames. ], batch size: 246, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:21:38,908 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.129e+02 2.519e+02 2.912e+02 9.426e+02, threshold=5.038e+02, percent-clipped=1.0 2023-05-01 08:21:54,319 INFO [train.py:904] (3/8) Epoch 21, batch 8900, loss[loss=0.1591, simple_loss=0.2572, pruned_loss=0.0305, over 16760.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2661, pruned_loss=0.03716, over 3027994.26 frames. ], batch size: 76, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:23:57,811 INFO [train.py:904] (3/8) Epoch 21, batch 8950, loss[loss=0.1656, simple_loss=0.2584, pruned_loss=0.03646, over 15250.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2661, pruned_loss=0.0374, over 3048432.95 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:24:11,376 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4283, 3.1928, 3.4347, 1.8239, 3.6177, 3.6847, 2.9982, 2.9108], device='cuda:3'), covar=tensor([0.0708, 0.0244, 0.0173, 0.1171, 0.0073, 0.0156, 0.0384, 0.0416], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0104, 0.0092, 0.0134, 0.0076, 0.0118, 0.0124, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 08:25:05,281 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 08:25:26,536 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7725, 3.1092, 3.3819, 1.9488, 2.9674, 2.2162, 3.3303, 3.3017], device='cuda:3'), covar=tensor([0.0262, 0.0886, 0.0522, 0.2107, 0.0796, 0.0972, 0.0633, 0.0986], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0157, 0.0162, 0.0149, 0.0141, 0.0126, 0.0139, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 08:25:29,293 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 9000, loss[loss=0.1757, simple_loss=0.255, pruned_loss=0.04819, over 12200.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2634, pruned_loss=0.03647, over 3043190.21 frames. ], batch size: 250, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:46,899 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 08:25:57,433 INFO [train.py:938] (3/8) Epoch 21, validation: loss=0.1457, simple_loss=0.2498, pruned_loss=0.02077, over 944034.00 frames. 2023-05-01 08:25:57,433 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 08:26:21,718 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6318, 4.7880, 4.9334, 4.7789, 4.7872, 5.3444, 4.9558, 4.6877], device='cuda:3'), covar=tensor([0.1138, 0.2140, 0.2326, 0.2078, 0.2477, 0.1057, 0.1577, 0.2301], device='cuda:3'), in_proj_covar=tensor([0.0383, 0.0559, 0.0614, 0.0466, 0.0617, 0.0648, 0.0488, 0.0627], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 08:27:20,310 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212042.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:27:39,212 INFO [train.py:904] (3/8) Epoch 21, batch 9050, loss[loss=0.1656, simple_loss=0.2571, pruned_loss=0.03704, over 16172.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2644, pruned_loss=0.03734, over 3043132.66 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:28:14,863 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:28:58,995 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212090.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:29:10,423 INFO [optim.py:368] (3/8) 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 08:29:26,249 INFO [train.py:904] (3/8) Epoch 21, batch 9100, loss[loss=0.1664, simple_loss=0.2687, pruned_loss=0.03206, over 16795.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2647, pruned_loss=0.03847, over 3026620.83 frames. ], batch size: 83, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:29:30,865 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212104.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:29:53,443 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:31:22,740 INFO [train.py:904] (3/8) Epoch 21, batch 9150, loss[loss=0.1555, simple_loss=0.2568, pruned_loss=0.02716, over 16883.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2646, pruned_loss=0.03754, over 3010074.60 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:31:49,981 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8779, 2.0411, 2.4131, 2.8349, 2.7276, 3.2176, 2.1346, 3.1633], device='cuda:3'), covar=tensor([0.0209, 0.0493, 0.0351, 0.0300, 0.0311, 0.0165, 0.0496, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0186, 0.0171, 0.0175, 0.0189, 0.0146, 0.0188, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:31:52,334 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212165.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:32:07,706 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7378, 1.3757, 1.6484, 1.6773, 1.8288, 1.8963, 1.6692, 1.7946], device='cuda:3'), covar=tensor([0.0297, 0.0433, 0.0238, 0.0324, 0.0310, 0.0223, 0.0428, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0186, 0.0171, 0.0175, 0.0189, 0.0146, 0.0188, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:32:29,967 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0934, 3.0854, 1.9161, 3.3452, 2.3237, 3.3398, 2.1225, 2.6165], device='cuda:3'), covar=tensor([0.0324, 0.0391, 0.1691, 0.0235, 0.0878, 0.0524, 0.1577, 0.0793], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0169, 0.0188, 0.0153, 0.0171, 0.0207, 0.0195, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-01 08:32:57,277 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.079e+02 2.542e+02 3.144e+02 5.396e+02, threshold=5.084e+02, percent-clipped=1.0 2023-05-01 08:33:09,950 INFO [train.py:904] (3/8) Epoch 21, batch 9200, loss[loss=0.1655, simple_loss=0.2475, pruned_loss=0.04173, over 12196.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2602, pruned_loss=0.03655, over 3003057.05 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:48,818 INFO [train.py:904] (3/8) Epoch 21, batch 9250, loss[loss=0.1587, simple_loss=0.2579, pruned_loss=0.02981, over 16427.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.26, pruned_loss=0.03674, over 3001490.60 frames. ], batch size: 68, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:51,544 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 08:36:39,742 INFO [train.py:904] (3/8) Epoch 21, batch 9300, loss[loss=0.1569, simple_loss=0.2409, pruned_loss=0.03646, over 12236.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2587, pruned_loss=0.03633, over 3009251.16 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:37:07,687 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212314.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:37:28,616 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4643, 2.0203, 1.7395, 1.7024, 2.2786, 1.9983, 1.9591, 2.3694], device='cuda:3'), covar=tensor([0.0187, 0.0411, 0.0483, 0.0472, 0.0263, 0.0374, 0.0187, 0.0267], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0225, 0.0218, 0.0218, 0.0225, 0.0224, 0.0222, 0.0218], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:37:30,382 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 08:38:22,832 INFO [train.py:904] (3/8) Epoch 21, batch 9350, loss[loss=0.1859, simple_loss=0.2769, pruned_loss=0.04742, over 12476.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2597, pruned_loss=0.03654, over 3015453.54 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:38:25,888 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8187, 3.8562, 4.0326, 3.7361, 3.9702, 4.3575, 3.9743, 3.5892], device='cuda:3'), covar=tensor([0.2119, 0.2323, 0.2123, 0.2602, 0.2457, 0.1435, 0.1518, 0.2786], device='cuda:3'), in_proj_covar=tensor([0.0381, 0.0561, 0.0618, 0.0466, 0.0616, 0.0651, 0.0488, 0.0625], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 08:39:47,852 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 9400, loss[loss=0.1597, simple_loss=0.2487, pruned_loss=0.03531, over 12401.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2604, pruned_loss=0.03615, over 3050815.74 frames. ], batch size: 247, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:18,513 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 9450, loss[loss=0.1478, simple_loss=0.2454, pruned_loss=0.02507, over 16507.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.262, pruned_loss=0.03623, over 3060033.63 frames. ], batch size: 68, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:58,731 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212460.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:43:11,246 INFO [optim.py:368] (3/8) 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 08:43:21,181 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:43:25,062 INFO [train.py:904] (3/8) Epoch 21, batch 9500, loss[loss=0.1549, simple_loss=0.257, pruned_loss=0.02635, over 16900.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2614, pruned_loss=0.03579, over 3069768.31 frames. ], batch size: 102, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:44:22,795 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1083, 5.2161, 5.0059, 4.6150, 4.6842, 5.1156, 4.9518, 4.7662], device='cuda:3'), covar=tensor([0.0584, 0.0526, 0.0311, 0.0296, 0.0886, 0.0521, 0.0332, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0274, 0.0396, 0.0321, 0.0316, 0.0327, 0.0367, 0.0222, 0.0383], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-05-01 08:45:12,072 INFO [train.py:904] (3/8) Epoch 21, batch 9550, loss[loss=0.1759, simple_loss=0.2815, pruned_loss=0.03514, over 16204.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2607, pruned_loss=0.03589, over 3075757.09 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:43,202 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5879, 3.5473, 3.5237, 2.8760, 3.4621, 2.0171, 3.2135, 2.8502], device='cuda:3'), covar=tensor([0.0131, 0.0114, 0.0170, 0.0188, 0.0100, 0.2489, 0.0130, 0.0270], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0145, 0.0185, 0.0164, 0.0164, 0.0197, 0.0173, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:46:39,906 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 21, batch 9600, loss[loss=0.1577, simple_loss=0.2436, pruned_loss=0.03586, over 12103.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.263, pruned_loss=0.03714, over 3068886.31 frames. ], batch size: 246, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:47:06,243 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212609.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:48:32,074 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 08:48:38,602 INFO [train.py:904] (3/8) Epoch 21, batch 9650, loss[loss=0.1562, simple_loss=0.2559, pruned_loss=0.02831, over 16443.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2649, pruned_loss=0.03737, over 3070700.52 frames. ], batch size: 68, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:11,921 INFO [optim.py:368] (3/8) 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 08:50:20,717 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 9700, loss[loss=0.1664, simple_loss=0.2521, pruned_loss=0.04034, over 12060.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2635, pruned_loss=0.03672, over 3072951.11 frames. ], batch size: 247, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:33,950 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2332, 4.2091, 4.5284, 4.5040, 4.5283, 4.2675, 4.2576, 4.3030], device='cuda:3'), covar=tensor([0.0356, 0.1083, 0.0617, 0.0629, 0.0616, 0.0645, 0.0907, 0.0457], device='cuda:3'), in_proj_covar=tensor([0.0382, 0.0426, 0.0413, 0.0385, 0.0459, 0.0433, 0.0511, 0.0349], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 08:50:46,001 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:50:50,491 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-01 08:50:58,005 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0279, 1.8492, 1.6695, 1.5220, 1.9895, 1.6940, 1.5653, 1.9245], device='cuda:3'), covar=tensor([0.0165, 0.0311, 0.0416, 0.0352, 0.0226, 0.0299, 0.0137, 0.0213], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0224, 0.0217, 0.0216, 0.0224, 0.0223, 0.0220, 0.0217], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:51:31,402 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6605, 3.7986, 2.9059, 2.2320, 2.3956, 2.4867, 4.0514, 3.3154], device='cuda:3'), covar=tensor([0.2968, 0.0578, 0.1817, 0.3025, 0.3021, 0.2146, 0.0343, 0.1292], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0260, 0.0297, 0.0303, 0.0284, 0.0251, 0.0284, 0.0324], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 08:52:10,270 INFO [train.py:904] (3/8) Epoch 21, batch 9750, loss[loss=0.1674, simple_loss=0.2673, pruned_loss=0.03369, over 16371.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2616, pruned_loss=0.03683, over 3052748.73 frames. ], batch size: 146, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:52:20,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1981, 4.2961, 4.0977, 3.8202, 3.8577, 4.2253, 3.8942, 3.9462], device='cuda:3'), covar=tensor([0.0585, 0.0518, 0.0317, 0.0291, 0.0714, 0.0448, 0.0796, 0.0636], device='cuda:3'), in_proj_covar=tensor([0.0275, 0.0397, 0.0321, 0.0315, 0.0326, 0.0368, 0.0222, 0.0383], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-05-01 08:52:23,091 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1602, 3.5022, 3.7362, 2.0924, 3.1897, 2.4615, 3.6263, 3.5588], device='cuda:3'), covar=tensor([0.0267, 0.0827, 0.0522, 0.2071, 0.0727, 0.0955, 0.0640, 0.1026], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0155, 0.0161, 0.0148, 0.0140, 0.0125, 0.0138, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 08:52:25,420 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:27,033 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212760.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:52:35,024 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1054, 2.0275, 2.4787, 2.9626, 2.8198, 3.4265, 2.2288, 3.3798], device='cuda:3'), covar=tensor([0.0227, 0.0601, 0.0406, 0.0323, 0.0330, 0.0159, 0.0545, 0.0154], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0188, 0.0172, 0.0176, 0.0190, 0.0146, 0.0190, 0.0143], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:52:49,243 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:53,837 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:46,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0151, 5.3312, 5.1317, 5.1216, 4.8846, 4.8092, 4.6968, 5.4378], device='cuda:3'), covar=tensor([0.1171, 0.0800, 0.0889, 0.0756, 0.0739, 0.0925, 0.1203, 0.0781], device='cuda:3'), in_proj_covar=tensor([0.0640, 0.0778, 0.0646, 0.0589, 0.0496, 0.0505, 0.0650, 0.0610], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 08:53:48,982 INFO [train.py:904] (3/8) Epoch 21, batch 9800, loss[loss=0.1561, simple_loss=0.2518, pruned_loss=0.03015, over 17117.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2614, pruned_loss=0.03583, over 3071238.83 frames. ], batch size: 47, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:54:01,428 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212808.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:54:33,330 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 08:54:51,570 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212835.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:55:31,550 INFO [train.py:904] (3/8) Epoch 21, batch 9850, loss[loss=0.156, simple_loss=0.2549, pruned_loss=0.02856, over 16298.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2627, pruned_loss=0.03549, over 3079616.35 frames. ], batch size: 146, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:07,910 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.176e+02 2.812e+02 3.349e+02 5.684e+02, threshold=5.624e+02, percent-clipped=7.0 2023-05-01 08:57:21,378 INFO [train.py:904] (3/8) Epoch 21, batch 9900, loss[loss=0.1872, simple_loss=0.295, pruned_loss=0.03968, over 16918.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2634, pruned_loss=0.03566, over 3082847.18 frames. ], batch size: 116, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:38,625 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:58:49,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 08:59:17,571 INFO [train.py:904] (3/8) Epoch 21, batch 9950, loss[loss=0.1834, simple_loss=0.2806, pruned_loss=0.04309, over 16188.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2649, pruned_loss=0.03601, over 3071751.21 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:59:29,578 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:00:58,646 INFO [optim.py:368] (3/8) 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 09:01:17,069 INFO [train.py:904] (3/8) Epoch 21, batch 10000, loss[loss=0.1522, simple_loss=0.2596, pruned_loss=0.02239, over 16847.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2638, pruned_loss=0.03576, over 3084344.41 frames. ], batch size: 90, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:01:51,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4801, 3.7543, 2.7497, 2.1338, 2.2909, 2.4071, 3.9788, 3.1445], device='cuda:3'), covar=tensor([0.3088, 0.0508, 0.1868, 0.2964, 0.2799, 0.2103, 0.0387, 0.1325], device='cuda:3'), in_proj_covar=tensor([0.0315, 0.0258, 0.0295, 0.0300, 0.0280, 0.0249, 0.0282, 0.0319], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:02:58,012 INFO [train.py:904] (3/8) Epoch 21, batch 10050, loss[loss=0.1652, simple_loss=0.2657, pruned_loss=0.03236, over 16813.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2639, pruned_loss=0.03564, over 3096799.99 frames. ], batch size: 83, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:03:02,065 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:04:21,048 INFO [optim.py:368] (3/8) 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 09:04:21,587 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 21, batch 10100, loss[loss=0.1453, simple_loss=0.2407, pruned_loss=0.02493, over 15241.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2643, pruned_loss=0.03568, over 3097899.28 frames. ], batch size: 191, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:05:27,297 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:51,219 INFO [train.py:904] (3/8) Epoch 21, batch 10150, loss[loss=0.1822, simple_loss=0.2591, pruned_loss=0.05261, over 12068.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2632, pruned_loss=0.03594, over 3079290.07 frames. ], batch size: 247, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:06:16,171 INFO [train.py:904] (3/8) Epoch 22, batch 0, loss[loss=0.2046, simple_loss=0.2896, pruned_loss=0.05977, over 16509.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2896, pruned_loss=0.05977, over 16509.00 frames. ], batch size: 68, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:06:16,172 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 09:06:23,636 INFO [train.py:938] (3/8) Epoch 22, validation: loss=0.1457, simple_loss=0.2489, pruned_loss=0.0212, over 944034.00 frames. 2023-05-01 09:06:23,636 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 09:07:26,338 INFO [optim.py:368] (3/8) 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 09:07:34,020 INFO [train.py:904] (3/8) Epoch 22, batch 50, loss[loss=0.1823, simple_loss=0.2538, pruned_loss=0.05547, over 16972.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2652, pruned_loss=0.04778, over 751047.36 frames. ], batch size: 109, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:08:41,859 INFO [train.py:904] (3/8) Epoch 22, batch 100, loss[loss=0.1883, simple_loss=0.2958, pruned_loss=0.04044, over 17126.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2659, pruned_loss=0.04743, over 1308415.69 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:09:44,121 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6756, 2.4584, 2.5164, 4.5576, 2.4890, 2.9261, 2.5624, 2.6572], device='cuda:3'), covar=tensor([0.1179, 0.3554, 0.2825, 0.0478, 0.4068, 0.2453, 0.3292, 0.3441], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0438, 0.0361, 0.0320, 0.0431, 0.0499, 0.0409, 0.0511], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:09:44,710 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.194e+02 2.643e+02 3.119e+02 6.629e+02, threshold=5.286e+02, percent-clipped=1.0 2023-05-01 09:09:51,952 INFO [train.py:904] (3/8) Epoch 22, batch 150, loss[loss=0.1898, simple_loss=0.2709, pruned_loss=0.05436, over 16901.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2645, pruned_loss=0.04591, over 1752709.55 frames. ], batch size: 109, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:00,679 INFO [train.py:904] (3/8) Epoch 22, batch 200, loss[loss=0.1626, simple_loss=0.2634, pruned_loss=0.03087, over 16658.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2649, pruned_loss=0.0461, over 2104082.81 frames. ], batch size: 62, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:02,346 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:11:20,422 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 09:12:07,193 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213402.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:08,109 INFO [train.py:904] (3/8) Epoch 22, batch 250, loss[loss=0.1664, simple_loss=0.2455, pruned_loss=0.04369, over 16784.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.263, pruned_loss=0.04557, over 2385073.04 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:12:24,686 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8745, 4.4893, 3.2302, 2.3349, 2.7082, 2.6025, 4.7912, 3.5722], device='cuda:3'), covar=tensor([0.2864, 0.0464, 0.1721, 0.2984, 0.2884, 0.2107, 0.0297, 0.1367], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0264, 0.0301, 0.0307, 0.0288, 0.0255, 0.0288, 0.0328], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:12:25,570 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213415.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:46,382 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:47,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5260, 3.5427, 2.0799, 3.6837, 2.8172, 3.6728, 2.2497, 2.8167], device='cuda:3'), covar=tensor([0.0274, 0.0416, 0.1594, 0.0413, 0.0787, 0.0840, 0.1476, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0159, 0.0177, 0.0214, 0.0202, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:12:55,176 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6324, 3.6348, 2.2477, 3.8575, 2.9408, 3.8120, 2.3539, 2.9149], device='cuda:3'), covar=tensor([0.0262, 0.0430, 0.1489, 0.0412, 0.0747, 0.0828, 0.1364, 0.0715], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0159, 0.0177, 0.0214, 0.0202, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:13:17,488 INFO [train.py:904] (3/8) Epoch 22, batch 300, loss[loss=0.1794, simple_loss=0.2646, pruned_loss=0.04703, over 16853.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2613, pruned_loss=0.04523, over 2593779.63 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:13:52,065 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:14:19,526 INFO [optim.py:368] (3/8) 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 09:14:25,242 INFO [train.py:904] (3/8) Epoch 22, batch 350, loss[loss=0.1761, simple_loss=0.2642, pruned_loss=0.04403, over 17049.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2588, pruned_loss=0.04369, over 2759314.04 frames. ], batch size: 50, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:14:30,283 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3107, 4.6478, 4.4820, 4.4593, 4.1915, 4.1683, 4.1292, 4.7158], device='cuda:3'), covar=tensor([0.1344, 0.0972, 0.1011, 0.0830, 0.0842, 0.1583, 0.1176, 0.0965], device='cuda:3'), in_proj_covar=tensor([0.0658, 0.0804, 0.0663, 0.0610, 0.0511, 0.0520, 0.0672, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:14:43,978 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 09:15:16,949 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7246, 2.5014, 2.5100, 4.1012, 3.3823, 4.0546, 1.6221, 2.9071], device='cuda:3'), covar=tensor([0.1407, 0.0743, 0.1161, 0.0170, 0.0143, 0.0359, 0.1540, 0.0851], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0172, 0.0192, 0.0185, 0.0198, 0.0211, 0.0200, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:15:34,170 INFO [train.py:904] (3/8) Epoch 22, batch 400, loss[loss=0.1849, simple_loss=0.2592, pruned_loss=0.05525, over 16738.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2566, pruned_loss=0.04321, over 2892045.16 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:15:40,343 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2045, 5.8808, 5.9289, 5.6602, 5.7616, 6.3242, 5.8506, 5.5839], device='cuda:3'), covar=tensor([0.0885, 0.1980, 0.2735, 0.2272, 0.2757, 0.1015, 0.1610, 0.2354], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0591, 0.0654, 0.0493, 0.0652, 0.0688, 0.0514, 0.0658], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:15:52,114 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:16:15,105 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1929, 5.2213, 5.6142, 5.5753, 5.6207, 5.2851, 5.1999, 5.0516], device='cuda:3'), covar=tensor([0.0349, 0.0537, 0.0349, 0.0481, 0.0504, 0.0371, 0.0946, 0.0402], device='cuda:3'), in_proj_covar=tensor([0.0397, 0.0443, 0.0427, 0.0401, 0.0475, 0.0450, 0.0531, 0.0363], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 09:16:36,629 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.265e+02 2.651e+02 3.318e+02 1.993e+03, threshold=5.302e+02, percent-clipped=3.0 2023-05-01 09:16:38,887 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8413, 2.5610, 2.4827, 3.8734, 3.2058, 3.9437, 1.5926, 2.9579], device='cuda:3'), covar=tensor([0.1352, 0.0693, 0.1172, 0.0170, 0.0134, 0.0376, 0.1536, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0172, 0.0192, 0.0185, 0.0199, 0.0212, 0.0200, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:16:43,116 INFO [train.py:904] (3/8) Epoch 22, batch 450, loss[loss=0.1636, simple_loss=0.2501, pruned_loss=0.03849, over 16858.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2551, pruned_loss=0.04204, over 2993200.68 frames. ], batch size: 90, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:17:16,870 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213627.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:17:53,028 INFO [train.py:904] (3/8) Epoch 22, batch 500, loss[loss=0.1584, simple_loss=0.2424, pruned_loss=0.03718, over 15563.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2537, pruned_loss=0.04114, over 3055080.43 frames. ], batch size: 191, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:18:54,885 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.187e+02 2.448e+02 2.877e+02 6.308e+02, threshold=4.896e+02, percent-clipped=1.0 2023-05-01 09:19:01,576 INFO [train.py:904] (3/8) Epoch 22, batch 550, loss[loss=0.1823, simple_loss=0.2582, pruned_loss=0.05321, over 16880.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2531, pruned_loss=0.04085, over 3110941.81 frames. ], batch size: 116, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:20:10,672 INFO [train.py:904] (3/8) Epoch 22, batch 600, loss[loss=0.1579, simple_loss=0.2277, pruned_loss=0.04406, over 16862.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2526, pruned_loss=0.04161, over 3156229.17 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:20:47,821 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 09:21:13,515 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 650, loss[loss=0.1559, simple_loss=0.2557, pruned_loss=0.0281, over 17269.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2504, pruned_loss=0.04105, over 3181491.46 frames. ], batch size: 52, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:21:54,530 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 09:22:30,249 INFO [train.py:904] (3/8) Epoch 22, batch 700, loss[loss=0.1347, simple_loss=0.2214, pruned_loss=0.02401, over 16796.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.25, pruned_loss=0.04071, over 3206524.80 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:23:15,778 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6050, 4.6870, 4.8331, 4.7042, 4.6711, 5.3261, 4.8227, 4.5096], device='cuda:3'), covar=tensor([0.1488, 0.2087, 0.2919, 0.2225, 0.2969, 0.1160, 0.1827, 0.2500], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0595, 0.0658, 0.0495, 0.0655, 0.0691, 0.0517, 0.0658], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:23:35,483 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 750, loss[loss=0.1671, simple_loss=0.2589, pruned_loss=0.03764, over 17012.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2506, pruned_loss=0.041, over 3232422.34 frames. ], batch size: 53, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:23:55,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9698, 3.0971, 3.3334, 2.1349, 2.8647, 2.2441, 3.4654, 3.4211], device='cuda:3'), covar=tensor([0.0243, 0.0911, 0.0547, 0.1844, 0.0844, 0.0990, 0.0518, 0.0891], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0161, 0.0166, 0.0153, 0.0144, 0.0129, 0.0143, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:23:58,463 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8558, 5.1904, 5.2891, 5.0780, 5.1129, 5.7713, 5.2677, 4.9779], device='cuda:3'), covar=tensor([0.1167, 0.2107, 0.2390, 0.2151, 0.2861, 0.1106, 0.1710, 0.2285], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0596, 0.0658, 0.0496, 0.0656, 0.0692, 0.0518, 0.0658], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:24:08,394 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:24:26,188 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7016, 3.8700, 2.5372, 4.4227, 3.0081, 4.3651, 2.6628, 3.1104], device='cuda:3'), covar=tensor([0.0342, 0.0415, 0.1546, 0.0328, 0.0887, 0.0572, 0.1402, 0.0783], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0196, 0.0163, 0.0178, 0.0217, 0.0203, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:24:53,399 INFO [train.py:904] (3/8) Epoch 22, batch 800, loss[loss=0.1701, simple_loss=0.2628, pruned_loss=0.03871, over 16629.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2501, pruned_loss=0.04047, over 3250629.61 frames. ], batch size: 62, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:25:56,892 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.094e+02 2.441e+02 2.760e+02 5.264e+02, threshold=4.882e+02, percent-clipped=1.0 2023-05-01 09:26:06,616 INFO [train.py:904] (3/8) Epoch 22, batch 850, loss[loss=0.1755, simple_loss=0.2514, pruned_loss=0.0498, over 16650.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2492, pruned_loss=0.03983, over 3266720.45 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:26:17,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8356, 2.7658, 2.5647, 4.4471, 3.7223, 4.2139, 1.6291, 3.0367], device='cuda:3'), covar=tensor([0.1328, 0.0709, 0.1170, 0.0184, 0.0211, 0.0414, 0.1544, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0188, 0.0202, 0.0215, 0.0202, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:27:17,240 INFO [train.py:904] (3/8) Epoch 22, batch 900, loss[loss=0.1991, simple_loss=0.2695, pruned_loss=0.0644, over 16877.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2492, pruned_loss=0.03964, over 3281601.07 frames. ], batch size: 109, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:28:19,799 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 1.995e+02 2.382e+02 2.694e+02 6.707e+02, threshold=4.763e+02, percent-clipped=1.0 2023-05-01 09:28:24,213 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7571, 5.1341, 4.8994, 4.9103, 4.6890, 4.6436, 4.5803, 5.1950], device='cuda:3'), covar=tensor([0.1380, 0.0890, 0.1064, 0.0823, 0.0824, 0.1141, 0.1137, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0676, 0.0821, 0.0679, 0.0626, 0.0525, 0.0532, 0.0690, 0.0646], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:28:27,529 INFO [train.py:904] (3/8) Epoch 22, batch 950, loss[loss=0.1624, simple_loss=0.2581, pruned_loss=0.03335, over 17114.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2493, pruned_loss=0.04, over 3285971.73 frames. ], batch size: 48, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:29:03,342 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 1000, loss[loss=0.1734, simple_loss=0.2441, pruned_loss=0.05135, over 16914.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2492, pruned_loss=0.03974, over 3293581.04 frames. ], batch size: 116, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:15,709 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4657, 2.2840, 1.8881, 2.0302, 2.6086, 2.3844, 2.5107, 2.7035], device='cuda:3'), covar=tensor([0.0279, 0.0444, 0.0520, 0.0487, 0.0227, 0.0319, 0.0254, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0212, 0.0238, 0.0228, 0.0229, 0.0239, 0.0237, 0.0238, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:30:20,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4176, 5.4350, 5.1933, 4.7093, 5.2737, 2.2673, 4.9881, 5.1471], device='cuda:3'), covar=tensor([0.0086, 0.0075, 0.0194, 0.0393, 0.0100, 0.2378, 0.0138, 0.0187], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0155, 0.0197, 0.0174, 0.0175, 0.0208, 0.0186, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:30:29,078 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214190.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:30:39,708 INFO [optim.py:368] (3/8) 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 09:30:46,636 INFO [train.py:904] (3/8) Epoch 22, batch 1050, loss[loss=0.1422, simple_loss=0.2291, pruned_loss=0.02768, over 17206.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2484, pruned_loss=0.03945, over 3296312.93 frames. ], batch size: 44, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:51,613 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3270, 3.5882, 3.7810, 2.0146, 2.9620, 2.4636, 3.7336, 3.7039], device='cuda:3'), covar=tensor([0.0293, 0.0938, 0.0569, 0.2181, 0.0968, 0.1041, 0.0648, 0.1099], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0154, 0.0145, 0.0130, 0.0144, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:31:07,763 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:13,781 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:56,093 INFO [train.py:904] (3/8) Epoch 22, batch 1100, loss[loss=0.1803, simple_loss=0.2746, pruned_loss=0.04301, over 16650.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2484, pruned_loss=0.03918, over 3303960.14 frames. ], batch size: 57, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:32:19,670 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214270.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:31,237 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:57,752 INFO [optim.py:368] (3/8) 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 09:33:03,836 INFO [train.py:904] (3/8) Epoch 22, batch 1150, loss[loss=0.1708, simple_loss=0.2673, pruned_loss=0.03712, over 17033.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2471, pruned_loss=0.03826, over 3313105.39 frames. ], batch size: 53, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:34:15,636 INFO [train.py:904] (3/8) Epoch 22, batch 1200, loss[loss=0.1731, simple_loss=0.2506, pruned_loss=0.04783, over 12715.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2466, pruned_loss=0.03799, over 3307584.20 frames. ], batch size: 247, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:34:54,167 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0204, 3.8788, 4.0884, 4.1935, 4.2575, 3.8438, 4.0767, 4.2687], device='cuda:3'), covar=tensor([0.1507, 0.1200, 0.1184, 0.0656, 0.0593, 0.1595, 0.2651, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0648, 0.0796, 0.0927, 0.0809, 0.0611, 0.0641, 0.0665, 0.0771], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:35:18,108 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 1250, loss[loss=0.1573, simple_loss=0.2378, pruned_loss=0.03835, over 12098.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2464, pruned_loss=0.03858, over 3301322.90 frames. ], batch size: 247, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:36:35,039 INFO [train.py:904] (3/8) Epoch 22, batch 1300, loss[loss=0.1721, simple_loss=0.2668, pruned_loss=0.03873, over 17093.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2468, pruned_loss=0.03886, over 3289698.44 frames. ], batch size: 55, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:37:18,456 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:37:34,616 INFO [optim.py:368] (3/8) 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 09:37:42,347 INFO [train.py:904] (3/8) Epoch 22, batch 1350, loss[loss=0.1486, simple_loss=0.2447, pruned_loss=0.02619, over 17153.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2473, pruned_loss=0.03829, over 3302619.15 frames. ], batch size: 48, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:37:55,036 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7877, 4.3772, 3.1309, 2.2960, 2.7022, 2.6543, 4.7036, 3.5829], device='cuda:3'), covar=tensor([0.3030, 0.0544, 0.1797, 0.3013, 0.2980, 0.2170, 0.0410, 0.1466], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0271, 0.0307, 0.0313, 0.0297, 0.0261, 0.0295, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:38:48,216 INFO [train.py:904] (3/8) Epoch 22, batch 1400, loss[loss=0.1605, simple_loss=0.235, pruned_loss=0.04303, over 16237.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2478, pruned_loss=0.03862, over 3315732.33 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:39:17,565 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 22, batch 1450, loss[loss=0.1681, simple_loss=0.2614, pruned_loss=0.0374, over 16993.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2473, pruned_loss=0.03824, over 3315121.88 frames. ], batch size: 50, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:41:07,195 INFO [train.py:904] (3/8) Epoch 22, batch 1500, loss[loss=0.1587, simple_loss=0.2641, pruned_loss=0.02665, over 17099.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2467, pruned_loss=0.03846, over 3317443.57 frames. ], batch size: 49, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:07,406 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 09:42:10,190 INFO [optim.py:368] (3/8) 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 09:42:16,362 INFO [train.py:904] (3/8) Epoch 22, batch 1550, loss[loss=0.1907, simple_loss=0.2626, pruned_loss=0.05935, over 16563.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2489, pruned_loss=0.04027, over 3314491.68 frames. ], batch size: 146, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:43:04,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8804, 2.7650, 2.5120, 2.8034, 3.1447, 2.8868, 3.5050, 3.3465], device='cuda:3'), covar=tensor([0.0143, 0.0448, 0.0485, 0.0385, 0.0266, 0.0376, 0.0219, 0.0263], device='cuda:3'), in_proj_covar=tensor([0.0216, 0.0240, 0.0230, 0.0230, 0.0240, 0.0239, 0.0242, 0.0236], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:43:27,995 INFO [train.py:904] (3/8) Epoch 22, batch 1600, loss[loss=0.178, simple_loss=0.2509, pruned_loss=0.05253, over 16788.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2512, pruned_loss=0.04124, over 3312809.92 frames. ], batch size: 89, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:44:12,456 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 09:44:32,956 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 09:44:37,869 INFO [train.py:904] (3/8) Epoch 22, batch 1650, loss[loss=0.1992, simple_loss=0.2774, pruned_loss=0.06047, over 15410.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2527, pruned_loss=0.04173, over 3319965.17 frames. ], batch size: 190, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:44:46,765 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6440, 4.4761, 4.6807, 4.8225, 4.9823, 4.4621, 4.8752, 4.9386], device='cuda:3'), covar=tensor([0.1886, 0.1528, 0.1654, 0.0879, 0.0720, 0.1109, 0.2099, 0.1349], device='cuda:3'), in_proj_covar=tensor([0.0651, 0.0800, 0.0935, 0.0814, 0.0614, 0.0642, 0.0665, 0.0774], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:45:20,797 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:45:47,645 INFO [train.py:904] (3/8) Epoch 22, batch 1700, loss[loss=0.1448, simple_loss=0.244, pruned_loss=0.02279, over 17186.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2545, pruned_loss=0.04186, over 3323189.48 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:46:07,568 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0437, 4.7684, 5.0306, 5.2476, 5.4867, 4.7529, 5.4355, 5.4338], device='cuda:3'), covar=tensor([0.1749, 0.1464, 0.1852, 0.0804, 0.0553, 0.0919, 0.0515, 0.0698], device='cuda:3'), in_proj_covar=tensor([0.0652, 0.0802, 0.0937, 0.0815, 0.0614, 0.0644, 0.0666, 0.0775], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:46:18,567 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:46:23,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0054, 2.2531, 2.6351, 2.9314, 2.8670, 3.5062, 2.4945, 3.4592], device='cuda:3'), covar=tensor([0.0289, 0.0504, 0.0360, 0.0359, 0.0346, 0.0196, 0.0465, 0.0185], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0185, 0.0199, 0.0155, 0.0197, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:46:45,048 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-05-01 09:46:53,102 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 1750, loss[loss=0.1862, simple_loss=0.2581, pruned_loss=0.0572, over 16872.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2555, pruned_loss=0.0418, over 3316054.10 frames. ], batch size: 109, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:47:25,706 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:48:06,740 INFO [train.py:904] (3/8) Epoch 22, batch 1800, loss[loss=0.1731, simple_loss=0.2635, pruned_loss=0.04133, over 16117.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2563, pruned_loss=0.04174, over 3306989.53 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:48:37,865 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 09:48:46,009 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4411, 5.8420, 5.5857, 5.6669, 5.2581, 5.2977, 5.2265, 6.0096], device='cuda:3'), covar=tensor([0.1465, 0.1053, 0.1085, 0.0900, 0.0953, 0.0799, 0.1210, 0.1009], device='cuda:3'), in_proj_covar=tensor([0.0684, 0.0839, 0.0692, 0.0638, 0.0534, 0.0540, 0.0704, 0.0657], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:49:04,891 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1796, 5.2189, 5.6797, 5.6492, 5.6615, 5.2857, 5.2521, 5.0904], device='cuda:3'), covar=tensor([0.0367, 0.0583, 0.0365, 0.0444, 0.0482, 0.0395, 0.0912, 0.0413], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0459, 0.0447, 0.0415, 0.0493, 0.0470, 0.0552, 0.0376], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 09:49:13,316 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 1850, loss[loss=0.1625, simple_loss=0.255, pruned_loss=0.03496, over 16708.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2563, pruned_loss=0.04137, over 3312579.64 frames. ], batch size: 62, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:49:46,630 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5810, 4.7171, 4.8663, 4.6241, 4.6789, 5.3189, 4.7630, 4.3798], device='cuda:3'), covar=tensor([0.1640, 0.1916, 0.2075, 0.2386, 0.2889, 0.1118, 0.1770, 0.2995], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0614, 0.0675, 0.0511, 0.0680, 0.0709, 0.0533, 0.0679], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:50:27,231 INFO [train.py:904] (3/8) Epoch 22, batch 1900, loss[loss=0.175, simple_loss=0.2666, pruned_loss=0.04165, over 15505.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2551, pruned_loss=0.04041, over 3315447.23 frames. ], batch size: 191, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:50:39,573 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 09:51:18,116 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5708, 2.2878, 1.8433, 2.1003, 2.6711, 2.4211, 2.6263, 2.7738], device='cuda:3'), covar=tensor([0.0238, 0.0429, 0.0594, 0.0447, 0.0240, 0.0349, 0.0226, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0216, 0.0240, 0.0230, 0.0231, 0.0241, 0.0240, 0.0243, 0.0236], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 09:51:31,792 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 1950, loss[loss=0.1878, simple_loss=0.2644, pruned_loss=0.05554, over 16452.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2554, pruned_loss=0.04084, over 3293905.72 frames. ], batch size: 146, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:51:41,257 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0837, 3.6761, 4.2811, 1.9772, 4.4396, 4.6185, 3.2968, 3.4361], device='cuda:3'), covar=tensor([0.0698, 0.0278, 0.0221, 0.1232, 0.0086, 0.0158, 0.0426, 0.0404], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0140, 0.0081, 0.0127, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:52:22,885 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-01 09:52:40,914 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7643, 4.9186, 5.0790, 4.8142, 4.9012, 5.5172, 4.9926, 4.6915], device='cuda:3'), covar=tensor([0.1480, 0.2022, 0.2362, 0.2183, 0.2611, 0.1036, 0.1616, 0.2502], device='cuda:3'), in_proj_covar=tensor([0.0420, 0.0614, 0.0673, 0.0509, 0.0677, 0.0707, 0.0531, 0.0679], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:52:44,784 INFO [train.py:904] (3/8) Epoch 22, batch 2000, loss[loss=0.1434, simple_loss=0.2352, pruned_loss=0.02576, over 17211.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2545, pruned_loss=0.04057, over 3305593.79 frames. ], batch size: 44, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:53:48,736 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2050, loss[loss=0.1674, simple_loss=0.2625, pruned_loss=0.03609, over 16683.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2553, pruned_loss=0.04117, over 3305385.41 frames. ], batch size: 62, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:54:31,385 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215230.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:54:34,058 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 09:55:01,196 INFO [train.py:904] (3/8) Epoch 22, batch 2100, loss[loss=0.1945, simple_loss=0.2653, pruned_loss=0.0618, over 16710.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2571, pruned_loss=0.04183, over 3312143.20 frames. ], batch size: 89, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:55:37,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3105, 3.5684, 3.7828, 2.1794, 3.1317, 2.4992, 3.8180, 3.7908], device='cuda:3'), covar=tensor([0.0329, 0.0978, 0.0581, 0.2088, 0.0814, 0.1017, 0.0645, 0.1140], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0164, 0.0167, 0.0155, 0.0145, 0.0131, 0.0144, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:55:54,404 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215291.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:56:04,629 INFO [optim.py:368] (3/8) 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 09:56:09,029 INFO [train.py:904] (3/8) Epoch 22, batch 2150, loss[loss=0.2476, simple_loss=0.3184, pruned_loss=0.08837, over 12157.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2584, pruned_loss=0.0427, over 3307802.26 frames. ], batch size: 246, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:56:16,245 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7746, 4.7112, 4.6696, 4.3561, 4.4133, 4.7508, 4.5413, 4.4614], device='cuda:3'), covar=tensor([0.0692, 0.0929, 0.0313, 0.0323, 0.0859, 0.0532, 0.0480, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0444, 0.0357, 0.0356, 0.0366, 0.0411, 0.0245, 0.0428], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 09:56:39,625 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215325.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:56:49,045 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 09:56:51,019 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:57:18,609 INFO [train.py:904] (3/8) Epoch 22, batch 2200, loss[loss=0.1719, simple_loss=0.2616, pruned_loss=0.04112, over 15565.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2582, pruned_loss=0.0422, over 3312658.40 frames. ], batch size: 190, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:02,255 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215386.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 09:58:14,329 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215394.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:58:20,773 INFO [optim.py:368] (3/8) 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 09:58:24,671 INFO [train.py:904] (3/8) Epoch 22, batch 2250, loss[loss=0.1774, simple_loss=0.2682, pruned_loss=0.04328, over 17041.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2589, pruned_loss=0.04215, over 3315192.21 frames. ], batch size: 55, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:33,598 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 09:59:35,079 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9337, 2.7299, 2.8657, 2.1402, 2.6389, 2.1668, 2.7782, 2.8768], device='cuda:3'), covar=tensor([0.0301, 0.0767, 0.0483, 0.1681, 0.0764, 0.0880, 0.0576, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0164, 0.0167, 0.0154, 0.0145, 0.0130, 0.0144, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 09:59:36,620 INFO [train.py:904] (3/8) Epoch 22, batch 2300, loss[loss=0.2252, simple_loss=0.3062, pruned_loss=0.0721, over 11896.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2593, pruned_loss=0.04234, over 3306892.19 frames. ], batch size: 246, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:59:43,408 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8506, 4.6448, 4.8870, 5.0844, 5.2995, 4.6846, 5.2833, 5.2943], device='cuda:3'), covar=tensor([0.1975, 0.1349, 0.1855, 0.0805, 0.0549, 0.0969, 0.0586, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0659, 0.0817, 0.0954, 0.0825, 0.0623, 0.0656, 0.0677, 0.0784], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:00:27,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6512, 2.5769, 2.5189, 4.5934, 2.4526, 2.9176, 2.6071, 2.7706], device='cuda:3'), covar=tensor([0.1204, 0.3449, 0.3014, 0.0476, 0.4152, 0.2565, 0.3540, 0.3520], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0452, 0.0372, 0.0331, 0.0439, 0.0519, 0.0423, 0.0530], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:00:42,879 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2350, loss[loss=0.1702, simple_loss=0.262, pruned_loss=0.03921, over 17111.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2595, pruned_loss=0.04226, over 3316841.32 frames. ], batch size: 49, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:46,889 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:01:48,915 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:01:55,003 INFO [train.py:904] (3/8) Epoch 22, batch 2400, loss[loss=0.1886, simple_loss=0.2614, pruned_loss=0.0579, over 16806.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2604, pruned_loss=0.04224, over 3312783.02 frames. ], batch size: 124, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:02:11,603 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215564.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:12,727 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:29,185 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 10:02:41,691 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215586.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:02:59,712 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2450, loss[loss=0.1539, simple_loss=0.2496, pruned_loss=0.02913, over 17094.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2611, pruned_loss=0.04231, over 3312073.61 frames. ], batch size: 47, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:03:08,423 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7171, 2.3850, 1.9861, 2.1899, 2.7788, 2.5736, 2.7764, 2.8684], device='cuda:3'), covar=tensor([0.0215, 0.0445, 0.0557, 0.0532, 0.0253, 0.0363, 0.0245, 0.0301], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0241, 0.0231, 0.0232, 0.0242, 0.0240, 0.0243, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:03:11,991 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215609.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:03:15,456 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-05-01 10:03:35,651 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 2500, loss[loss=0.1817, simple_loss=0.2585, pruned_loss=0.05242, over 16740.00 frames. ], tot_loss[loss=0.173, simple_loss=0.261, pruned_loss=0.04252, over 3310116.08 frames. ], batch size: 124, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:04:34,299 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 10:04:36,331 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-05-01 10:04:38,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8825, 2.0086, 2.5005, 2.7894, 2.6332, 3.1984, 2.2901, 3.2301], device='cuda:3'), covar=tensor([0.0261, 0.0545, 0.0361, 0.0346, 0.0361, 0.0241, 0.0488, 0.0184], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0196, 0.0182, 0.0187, 0.0200, 0.0157, 0.0198, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:04:51,429 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215681.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:05:02,670 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215689.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 10:05:15,511 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2550, loss[loss=0.1888, simple_loss=0.2847, pruned_loss=0.04645, over 16769.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2614, pruned_loss=0.04245, over 3315688.02 frames. ], batch size: 57, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:05:25,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2792, 5.8446, 5.9834, 5.5826, 5.6713, 6.2781, 5.7682, 5.4158], device='cuda:3'), covar=tensor([0.0996, 0.1843, 0.2029, 0.2083, 0.2703, 0.0975, 0.1379, 0.2442], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0618, 0.0679, 0.0514, 0.0684, 0.0714, 0.0535, 0.0684], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:06:09,934 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3392, 5.6657, 5.3935, 5.5019, 5.0727, 5.0753, 5.1420, 5.7911], device='cuda:3'), covar=tensor([0.1382, 0.0900, 0.1213, 0.0894, 0.0949, 0.0828, 0.1183, 0.0953], device='cuda:3'), in_proj_covar=tensor([0.0688, 0.0844, 0.0696, 0.0643, 0.0535, 0.0541, 0.0706, 0.0660], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:06:30,382 INFO [train.py:904] (3/8) Epoch 22, batch 2600, loss[loss=0.1638, simple_loss=0.2621, pruned_loss=0.03281, over 16637.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2611, pruned_loss=0.04196, over 3315120.53 frames. ], batch size: 62, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:07:36,116 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2650, loss[loss=0.1458, simple_loss=0.2375, pruned_loss=0.02702, over 16997.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2608, pruned_loss=0.04154, over 3320066.33 frames. ], batch size: 41, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:07:48,598 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8973, 2.8427, 2.5032, 2.9350, 3.2311, 3.0291, 3.5111, 3.4183], device='cuda:3'), covar=tensor([0.0143, 0.0422, 0.0508, 0.0386, 0.0274, 0.0373, 0.0258, 0.0249], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0241, 0.0231, 0.0232, 0.0242, 0.0241, 0.0244, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:08:16,694 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215829.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:08:48,040 INFO [train.py:904] (3/8) Epoch 22, batch 2700, loss[loss=0.193, simple_loss=0.2863, pruned_loss=0.04983, over 16723.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2609, pruned_loss=0.04099, over 3316720.18 frames. ], batch size: 57, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:57,507 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:20,429 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 10:09:22,235 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-01 10:09:23,573 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-01 10:09:34,853 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215886.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:09:39,311 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:53,924 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.142e+02 2.465e+02 2.972e+02 5.160e+02, threshold=4.929e+02, percent-clipped=1.0 2023-05-01 10:09:57,375 INFO [train.py:904] (3/8) Epoch 22, batch 2750, loss[loss=0.1663, simple_loss=0.2612, pruned_loss=0.03572, over 17160.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2602, pruned_loss=0.04037, over 3317213.59 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:09:59,451 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215904.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:10:21,927 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:10:38,795 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 2800, loss[loss=0.1635, simple_loss=0.2572, pruned_loss=0.03491, over 17110.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2602, pruned_loss=0.03988, over 3326694.78 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:11:25,073 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 10:11:42,387 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215981.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:11:53,837 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215989.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:12:07,418 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2850, loss[loss=0.175, simple_loss=0.2525, pruned_loss=0.04875, over 16796.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2594, pruned_loss=0.03974, over 3334055.23 frames. ], batch size: 134, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:12:51,446 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216028.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:12:52,493 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216029.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:13:04,675 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 2900, loss[loss=0.1771, simple_loss=0.2761, pruned_loss=0.03899, over 16679.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2576, pruned_loss=0.04001, over 3340705.36 frames. ], batch size: 57, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:14:03,364 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7722, 3.8935, 2.9671, 2.2969, 2.5888, 2.4890, 4.0371, 3.4321], device='cuda:3'), covar=tensor([0.2700, 0.0564, 0.1757, 0.3188, 0.2693, 0.2025, 0.0496, 0.1349], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0272, 0.0307, 0.0314, 0.0298, 0.0262, 0.0296, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:14:17,823 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:14:31,571 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 2950, loss[loss=0.1745, simple_loss=0.2548, pruned_loss=0.04711, over 16787.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2568, pruned_loss=0.04008, over 3346991.24 frames. ], batch size: 83, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:00,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7573, 4.8615, 4.6895, 4.4185, 4.0341, 4.9065, 4.8080, 4.4745], device='cuda:3'), covar=tensor([0.0961, 0.0908, 0.0482, 0.0451, 0.1784, 0.0535, 0.0417, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0448, 0.0358, 0.0359, 0.0370, 0.0415, 0.0247, 0.0433], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:15:45,387 INFO [train.py:904] (3/8) Epoch 22, batch 3000, loss[loss=0.1598, simple_loss=0.2391, pruned_loss=0.0402, over 16047.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2561, pruned_loss=0.04041, over 3348292.82 frames. ], batch size: 35, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,387 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 10:15:54,107 INFO [train.py:938] (3/8) Epoch 22, validation: loss=0.1347, simple_loss=0.2399, pruned_loss=0.0148, over 944034.00 frames. 2023-05-01 10:15:54,108 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 10:16:02,688 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:38,785 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:59,841 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.146e+02 2.550e+02 3.086e+02 1.158e+03, threshold=5.100e+02, percent-clipped=1.0 2023-05-01 10:17:03,725 INFO [train.py:904] (3/8) Epoch 22, batch 3050, loss[loss=0.1734, simple_loss=0.265, pruned_loss=0.0409, over 17031.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2559, pruned_loss=0.04085, over 3343353.58 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:17:05,302 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216204.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:17:09,670 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216207.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:22,516 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-01 10:17:29,347 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:32,097 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 10:17:32,926 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4876, 3.6294, 3.8425, 2.6252, 3.5109, 3.9198, 3.5933, 2.2368], device='cuda:3'), covar=tensor([0.0499, 0.0171, 0.0061, 0.0399, 0.0126, 0.0104, 0.0102, 0.0471], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0132, 0.0099, 0.0109, 0.0094, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:17:36,559 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2128, 2.2516, 2.3772, 3.9295, 2.2646, 2.5481, 2.3208, 2.4144], device='cuda:3'), covar=tensor([0.1397, 0.3670, 0.2797, 0.0585, 0.3645, 0.2613, 0.3725, 0.3067], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0452, 0.0371, 0.0332, 0.0439, 0.0520, 0.0423, 0.0530], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:18:09,848 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8082, 3.9451, 2.6284, 4.5829, 3.0510, 4.5313, 2.6435, 3.2604], device='cuda:3'), covar=tensor([0.0336, 0.0435, 0.1528, 0.0284, 0.0876, 0.0518, 0.1566, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0180, 0.0197, 0.0169, 0.0180, 0.0224, 0.0205, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:18:10,761 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216252.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:18:12,221 INFO [train.py:904] (3/8) Epoch 22, batch 3100, loss[loss=0.1459, simple_loss=0.2243, pruned_loss=0.03377, over 16784.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2552, pruned_loss=0.04049, over 3349403.23 frames. ], batch size: 39, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:18:17,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5784, 2.5075, 2.5408, 4.6046, 2.4323, 2.8979, 2.5392, 2.6876], device='cuda:3'), covar=tensor([0.1317, 0.3569, 0.3010, 0.0495, 0.4179, 0.2511, 0.3574, 0.3722], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0452, 0.0372, 0.0332, 0.0440, 0.0520, 0.0423, 0.0531], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:18:28,505 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 10:18:33,892 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216269.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:18:56,480 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8489, 2.4663, 2.3676, 3.6305, 2.7784, 3.7094, 1.5572, 2.7015], device='cuda:3'), covar=tensor([0.1374, 0.0817, 0.1332, 0.0228, 0.0194, 0.0482, 0.1717, 0.0900], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0193, 0.0204, 0.0217, 0.0202, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:19:16,924 INFO [optim.py:368] (3/8) 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 10:19:21,073 INFO [train.py:904] (3/8) Epoch 22, batch 3150, loss[loss=0.1865, simple_loss=0.2604, pruned_loss=0.05634, over 16497.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2549, pruned_loss=0.04046, over 3342062.05 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:20:29,459 INFO [train.py:904] (3/8) Epoch 22, batch 3200, loss[loss=0.1504, simple_loss=0.234, pruned_loss=0.03345, over 16932.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2541, pruned_loss=0.0403, over 3329345.94 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:20:48,545 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3430, 3.6281, 3.9257, 2.1136, 3.2242, 2.4974, 3.7524, 3.8650], device='cuda:3'), covar=tensor([0.0267, 0.0870, 0.0504, 0.2080, 0.0778, 0.0972, 0.0597, 0.0999], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0165, 0.0167, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:21:13,833 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216384.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:21:23,310 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1593, 3.3522, 3.3562, 2.1419, 2.8579, 2.3610, 3.5742, 3.7093], device='cuda:3'), covar=tensor([0.0244, 0.0878, 0.0717, 0.2031, 0.0938, 0.1063, 0.0539, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:21:36,408 INFO [optim.py:368] (3/8) 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 10:21:40,317 INFO [train.py:904] (3/8) Epoch 22, batch 3250, loss[loss=0.1814, simple_loss=0.2632, pruned_loss=0.04982, over 16862.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2547, pruned_loss=0.04033, over 3329611.18 frames. ], batch size: 102, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:22:52,219 INFO [train.py:904] (3/8) Epoch 22, batch 3300, loss[loss=0.1947, simple_loss=0.2847, pruned_loss=0.05242, over 16448.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2556, pruned_loss=0.0409, over 3319171.13 frames. ], batch size: 68, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:23:36,591 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:23:56,629 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 3350, loss[loss=0.1633, simple_loss=0.2581, pruned_loss=0.03424, over 17164.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2567, pruned_loss=0.04099, over 3325334.98 frames. ], batch size: 46, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:24:42,589 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:25:10,643 INFO [train.py:904] (3/8) Epoch 22, batch 3400, loss[loss=0.1653, simple_loss=0.2615, pruned_loss=0.03457, over 17043.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.257, pruned_loss=0.04063, over 3324191.66 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:01,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6016, 2.5754, 2.6456, 4.6694, 3.6475, 4.2764, 1.6462, 2.9642], device='cuda:3'), covar=tensor([0.1872, 0.1065, 0.1437, 0.0222, 0.0281, 0.0403, 0.2123, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0195, 0.0206, 0.0219, 0.0203, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:26:02,834 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:26:15,739 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.115e+02 2.508e+02 2.941e+02 4.033e+02, threshold=5.017e+02, percent-clipped=0.0 2023-05-01 10:26:19,664 INFO [train.py:904] (3/8) Epoch 22, batch 3450, loss[loss=0.1718, simple_loss=0.2481, pruned_loss=0.04777, over 16779.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2567, pruned_loss=0.04042, over 3318319.74 frames. ], batch size: 134, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:36,440 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6835, 3.3778, 3.7296, 2.0781, 3.8057, 3.8428, 3.1275, 2.8754], device='cuda:3'), covar=tensor([0.0692, 0.0242, 0.0185, 0.1108, 0.0104, 0.0202, 0.0399, 0.0426], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0109, 0.0100, 0.0140, 0.0082, 0.0130, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:26:51,293 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0999, 5.0596, 4.8471, 4.3524, 4.9383, 1.7800, 4.6452, 4.5898], device='cuda:3'), covar=tensor([0.0090, 0.0087, 0.0232, 0.0395, 0.0114, 0.3141, 0.0149, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0160, 0.0205, 0.0183, 0.0183, 0.0212, 0.0194, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:27:29,205 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 3500, loss[loss=0.1723, simple_loss=0.2593, pruned_loss=0.04266, over 16250.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2557, pruned_loss=0.04012, over 3319744.18 frames. ], batch size: 164, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:27:32,820 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216655.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:27:38,649 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 10:28:06,664 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9764, 2.0556, 2.5315, 2.7812, 2.7685, 3.1227, 2.2297, 3.1944], device='cuda:3'), covar=tensor([0.0202, 0.0501, 0.0348, 0.0344, 0.0340, 0.0272, 0.0523, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0198, 0.0184, 0.0190, 0.0202, 0.0158, 0.0200, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:28:12,552 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:28:38,631 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8967, 2.0912, 2.5854, 2.8172, 2.7266, 3.3043, 2.3995, 3.2660], device='cuda:3'), covar=tensor([0.0249, 0.0501, 0.0334, 0.0342, 0.0344, 0.0202, 0.0454, 0.0182], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0197, 0.0184, 0.0189, 0.0201, 0.0158, 0.0200, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:28:39,276 INFO [train.py:904] (3/8) Epoch 22, batch 3550, loss[loss=0.1925, simple_loss=0.2778, pruned_loss=0.05364, over 16756.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2552, pruned_loss=0.04024, over 3312521.50 frames. ], batch size: 57, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:28:42,521 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:57,365 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216716.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:29:20,501 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216732.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:29:45,270 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8557, 2.3106, 2.3297, 3.2410, 2.4442, 3.5225, 1.6530, 2.6260], device='cuda:3'), covar=tensor([0.1339, 0.0858, 0.1310, 0.0245, 0.0180, 0.0443, 0.1604, 0.0925], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0195, 0.0206, 0.0219, 0.0203, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:29:49,493 INFO [train.py:904] (3/8) Epoch 22, batch 3600, loss[loss=0.1655, simple_loss=0.2621, pruned_loss=0.03443, over 17247.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2537, pruned_loss=0.04003, over 3314128.43 frames. ], batch size: 52, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:29:52,298 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8937, 2.4869, 1.9505, 2.3356, 2.8815, 2.6437, 2.9245, 2.9974], device='cuda:3'), covar=tensor([0.0231, 0.0441, 0.0643, 0.0470, 0.0258, 0.0362, 0.0226, 0.0292], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0242, 0.0243, 0.0247, 0.0240], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:30:08,392 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:30:51,333 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 10:31:00,614 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 3650, loss[loss=0.1698, simple_loss=0.2437, pruned_loss=0.04792, over 16847.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2526, pruned_loss=0.04093, over 3312538.32 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:32:13,500 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-05-01 10:32:18,426 INFO [train.py:904] (3/8) Epoch 22, batch 3700, loss[loss=0.19, simple_loss=0.2611, pruned_loss=0.05941, over 11678.00 frames. ], tot_loss[loss=0.168, simple_loss=0.251, pruned_loss=0.04246, over 3292616.31 frames. ], batch size: 248, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:32:52,429 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-05-01 10:32:54,280 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5609, 4.5164, 4.5049, 4.0580, 4.5208, 1.8471, 4.3307, 4.1357], device='cuda:3'), covar=tensor([0.0136, 0.0125, 0.0173, 0.0279, 0.0098, 0.2621, 0.0132, 0.0217], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0161, 0.0205, 0.0182, 0.0183, 0.0212, 0.0194, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:33:31,443 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.238e+02 2.459e+02 3.071e+02 7.166e+02, threshold=4.918e+02, percent-clipped=1.0 2023-05-01 10:33:32,645 INFO [train.py:904] (3/8) Epoch 22, batch 3750, loss[loss=0.1643, simple_loss=0.2411, pruned_loss=0.04371, over 16848.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2521, pruned_loss=0.04404, over 3268943.47 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:34:11,431 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3942, 4.4032, 4.7454, 4.7454, 4.7864, 4.4454, 4.4701, 4.3114], device='cuda:3'), covar=tensor([0.0395, 0.0719, 0.0384, 0.0415, 0.0477, 0.0441, 0.0847, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0467, 0.0453, 0.0420, 0.0502, 0.0477, 0.0562, 0.0382], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 10:34:31,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3202, 2.3216, 2.3586, 4.2465, 2.3597, 2.6627, 2.4171, 2.5686], device='cuda:3'), covar=tensor([0.1469, 0.3782, 0.2989, 0.0522, 0.3855, 0.2508, 0.3795, 0.2981], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0455, 0.0374, 0.0334, 0.0442, 0.0525, 0.0426, 0.0534], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:34:36,638 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 3800, loss[loss=0.1477, simple_loss=0.2304, pruned_loss=0.03249, over 16746.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2535, pruned_loss=0.04503, over 3263600.09 frames. ], batch size: 89, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:35:03,068 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:35:55,032 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 3850, loss[loss=0.2084, simple_loss=0.2888, pruned_loss=0.06398, over 12323.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2536, pruned_loss=0.04547, over 3262526.97 frames. ], batch size: 248, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:36:08,428 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217011.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:36:29,865 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:09,497 INFO [train.py:904] (3/8) Epoch 22, batch 3900, loss[loss=0.1745, simple_loss=0.2604, pruned_loss=0.04431, over 17115.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.253, pruned_loss=0.04575, over 3258782.98 frames. ], batch size: 48, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:37:22,072 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217061.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:40,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7660, 5.0508, 4.8416, 4.8742, 4.6498, 4.4929, 4.5295, 5.1416], device='cuda:3'), covar=tensor([0.1137, 0.0807, 0.0901, 0.0836, 0.0743, 0.1267, 0.1142, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0696, 0.0853, 0.0701, 0.0647, 0.0540, 0.0548, 0.0714, 0.0667], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:38:21,557 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 3950, loss[loss=0.1632, simple_loss=0.2478, pruned_loss=0.03931, over 16268.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2524, pruned_loss=0.04626, over 3266101.69 frames. ], batch size: 35, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:38:41,445 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6900, 4.7856, 4.9614, 4.8097, 4.8113, 5.3970, 4.9486, 4.6519], device='cuda:3'), covar=tensor([0.1496, 0.1937, 0.2045, 0.1958, 0.2661, 0.1030, 0.1512, 0.2378], device='cuda:3'), in_proj_covar=tensor([0.0420, 0.0607, 0.0665, 0.0504, 0.0672, 0.0699, 0.0522, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:38:49,559 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 10:39:35,757 INFO [train.py:904] (3/8) Epoch 22, batch 4000, loss[loss=0.1712, simple_loss=0.2574, pruned_loss=0.04249, over 16862.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2524, pruned_loss=0.04696, over 3276904.84 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:39:38,889 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-01 10:40:48,066 INFO [optim.py:368] (3/8) 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 10:40:49,969 INFO [train.py:904] (3/8) Epoch 22, batch 4050, loss[loss=0.1642, simple_loss=0.2507, pruned_loss=0.03887, over 16854.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2531, pruned_loss=0.04618, over 3274524.82 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:41:49,074 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 10:41:55,312 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:42:04,909 INFO [train.py:904] (3/8) Epoch 22, batch 4100, loss[loss=0.214, simple_loss=0.2971, pruned_loss=0.06543, over 15375.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2553, pruned_loss=0.04587, over 3272198.91 frames. ], batch size: 190, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:43:11,028 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 22, batch 4150, loss[loss=0.1936, simple_loss=0.2922, pruned_loss=0.04752, over 15334.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.262, pruned_loss=0.04821, over 3226617.18 frames. ], batch size: 190, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:43:36,163 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217311.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:43:52,135 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:44:27,282 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 10:44:28,698 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2269, 1.5440, 1.9769, 2.0603, 2.2440, 2.3860, 1.7409, 2.3002], device='cuda:3'), covar=tensor([0.0242, 0.0487, 0.0274, 0.0339, 0.0297, 0.0195, 0.0515, 0.0156], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0195, 0.0181, 0.0187, 0.0198, 0.0156, 0.0197, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:44:39,667 INFO [train.py:904] (3/8) Epoch 22, batch 4200, loss[loss=0.1978, simple_loss=0.2904, pruned_loss=0.05256, over 16430.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2686, pruned_loss=0.04957, over 3189942.39 frames. ], batch size: 35, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:44:50,070 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217359.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:44:52,725 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:32,834 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217388.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:42,714 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-01 10:45:53,905 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 4250, loss[loss=0.179, simple_loss=0.2782, pruned_loss=0.0399, over 16597.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2717, pruned_loss=0.04925, over 3164006.54 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:46:04,661 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217409.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:46:44,490 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8422, 1.3353, 1.7426, 1.6779, 1.8267, 1.9764, 1.6597, 1.8244], device='cuda:3'), covar=tensor([0.0257, 0.0437, 0.0236, 0.0329, 0.0268, 0.0165, 0.0432, 0.0153], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0186, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:47:04,237 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 4300, loss[loss=0.1848, simple_loss=0.2791, pruned_loss=0.04529, over 15415.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2725, pruned_loss=0.04812, over 3171145.91 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:48:07,813 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.190e+02 2.513e+02 2.861e+02 5.953e+02, threshold=5.026e+02, percent-clipped=1.0 2023-05-01 10:48:24,299 INFO [train.py:904] (3/8) Epoch 22, batch 4350, loss[loss=0.1837, simple_loss=0.2763, pruned_loss=0.04556, over 16341.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2756, pruned_loss=0.04954, over 3139892.88 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:39,977 INFO [train.py:904] (3/8) Epoch 22, batch 4400, loss[loss=0.1903, simple_loss=0.2793, pruned_loss=0.05064, over 16617.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2778, pruned_loss=0.05058, over 3169002.77 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:41,081 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 22, batch 4450, loss[loss=0.2007, simple_loss=0.2819, pruned_loss=0.0598, over 11876.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.281, pruned_loss=0.05194, over 3168783.80 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:51:20,404 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:07,082 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5792, 5.5634, 5.3138, 4.6889, 5.5293, 2.0172, 5.2588, 4.9478], device='cuda:3'), covar=tensor([0.0042, 0.0038, 0.0128, 0.0286, 0.0038, 0.2834, 0.0067, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0159, 0.0203, 0.0182, 0.0181, 0.0212, 0.0193, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 10:52:08,356 INFO [train.py:904] (3/8) Epoch 22, batch 4500, loss[loss=0.1926, simple_loss=0.2823, pruned_loss=0.05144, over 16989.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2818, pruned_loss=0.0526, over 3177658.12 frames. ], batch size: 41, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:52:24,069 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:53:01,049 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 10:53:11,165 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6241, 3.6797, 2.8216, 2.2774, 2.5257, 2.2931, 4.0529, 3.3197], device='cuda:3'), covar=tensor([0.2839, 0.0603, 0.1729, 0.2524, 0.2529, 0.2240, 0.0416, 0.1195], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0271, 0.0306, 0.0316, 0.0300, 0.0261, 0.0296, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:53:18,212 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 4550, loss[loss=0.2199, simple_loss=0.3026, pruned_loss=0.06865, over 16926.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2832, pruned_loss=0.05355, over 3195648.62 frames. ], batch size: 116, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:53:52,628 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:19,806 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:27,337 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7344, 3.7992, 4.1389, 2.3005, 3.3929, 2.6685, 3.9594, 3.9810], device='cuda:3'), covar=tensor([0.0170, 0.0731, 0.0481, 0.2013, 0.0780, 0.0978, 0.0507, 0.0878], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0153, 0.0145, 0.0130, 0.0144, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:54:31,797 INFO [train.py:904] (3/8) Epoch 22, batch 4600, loss[loss=0.2017, simple_loss=0.2861, pruned_loss=0.05863, over 16758.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2842, pruned_loss=0.05352, over 3215540.55 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:55:20,852 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3361, 3.3255, 1.9295, 3.8399, 2.4930, 3.8275, 2.2178, 2.7442], device='cuda:3'), covar=tensor([0.0352, 0.0390, 0.1945, 0.0173, 0.1027, 0.0537, 0.1602, 0.0870], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0164, 0.0178, 0.0221, 0.0203, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:55:41,527 INFO [optim.py:368] (3/8) 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 10:55:42,804 INFO [train.py:904] (3/8) Epoch 22, batch 4650, loss[loss=0.1778, simple_loss=0.2651, pruned_loss=0.04529, over 16497.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2833, pruned_loss=0.05366, over 3221116.36 frames. ], batch size: 68, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:56:46,085 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217848.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:56:52,991 INFO [train.py:904] (3/8) Epoch 22, batch 4700, loss[loss=0.2105, simple_loss=0.2844, pruned_loss=0.06827, over 11468.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2805, pruned_loss=0.05251, over 3224981.77 frames. ], batch size: 246, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:57:01,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3855, 3.5591, 3.8732, 1.8858, 4.1488, 4.1913, 3.1080, 2.9167], device='cuda:3'), covar=tensor([0.1264, 0.0265, 0.0215, 0.1477, 0.0086, 0.0154, 0.0452, 0.0622], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0139, 0.0081, 0.0127, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:57:16,765 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6241, 3.7305, 2.8822, 2.3082, 2.5114, 2.3987, 4.1733, 3.2866], device='cuda:3'), covar=tensor([0.3013, 0.0723, 0.1886, 0.2686, 0.2512, 0.2165, 0.0477, 0.1333], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0270, 0.0306, 0.0316, 0.0299, 0.0261, 0.0296, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 10:57:38,794 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0353, 4.1362, 2.5524, 5.0608, 3.2981, 4.8731, 2.7323, 3.3718], device='cuda:3'), covar=tensor([0.0287, 0.0376, 0.1740, 0.0093, 0.0786, 0.0380, 0.1483, 0.0730], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0164, 0.0178, 0.0220, 0.0202, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 10:57:59,155 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217898.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:58:04,666 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 4750, loss[loss=0.169, simple_loss=0.2589, pruned_loss=0.0396, over 16457.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2763, pruned_loss=0.05016, over 3228150.47 frames. ], batch size: 75, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:58:49,641 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 10:59:17,334 INFO [train.py:904] (3/8) Epoch 22, batch 4800, loss[loss=0.1703, simple_loss=0.2558, pruned_loss=0.04235, over 16474.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2727, pruned_loss=0.04829, over 3207518.18 frames. ], batch size: 68, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:27,604 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217959.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:00:36,339 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.772e+02 2.165e+02 2.536e+02 4.360e+02, threshold=4.330e+02, percent-clipped=2.0 2023-05-01 11:00:36,354 INFO [train.py:904] (3/8) Epoch 22, batch 4850, loss[loss=0.1802, simple_loss=0.2779, pruned_loss=0.04126, over 16217.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2728, pruned_loss=0.0476, over 3188007.75 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:01:01,609 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218020.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:38,583 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:51,558 INFO [train.py:904] (3/8) Epoch 22, batch 4900, loss[loss=0.1728, simple_loss=0.2565, pruned_loss=0.04455, over 12203.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2716, pruned_loss=0.04595, over 3179307.02 frames. ], batch size: 247, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:02:49,274 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218092.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:01,556 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 11:03:05,462 INFO [train.py:904] (3/8) Epoch 22, batch 4950, loss[loss=0.1963, simple_loss=0.2925, pruned_loss=0.05002, over 16735.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2711, pruned_loss=0.04538, over 3188831.20 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:03:20,858 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 11:04:11,456 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:04:12,887 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1771, 2.2564, 2.8983, 3.0937, 3.0699, 3.8937, 2.4536, 3.7610], device='cuda:3'), covar=tensor([0.0224, 0.0509, 0.0322, 0.0321, 0.0313, 0.0127, 0.0538, 0.0130], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0197, 0.0182, 0.0188, 0.0202, 0.0156, 0.0200, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:04:18,892 INFO [train.py:904] (3/8) Epoch 22, batch 5000, loss[loss=0.1766, simple_loss=0.2737, pruned_loss=0.03974, over 16758.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2729, pruned_loss=0.04553, over 3202638.20 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:04:30,122 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:04:30,268 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 11:04:56,624 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9528, 3.7694, 3.7151, 2.4136, 3.3523, 3.7934, 3.3683, 2.0103], device='cuda:3'), covar=tensor([0.0622, 0.0051, 0.0051, 0.0404, 0.0109, 0.0087, 0.0110, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0082, 0.0084, 0.0132, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:05:01,531 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218182.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:21,356 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218196.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:32,359 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 5050, loss[loss=0.1978, simple_loss=0.2861, pruned_loss=0.05472, over 12382.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2731, pruned_loss=0.04531, over 3203408.16 frames. ], batch size: 248, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:05:35,933 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2963, 4.2705, 4.2375, 3.3770, 4.2446, 1.6690, 3.9529, 3.8888], device='cuda:3'), covar=tensor([0.0130, 0.0130, 0.0169, 0.0470, 0.0114, 0.2894, 0.0161, 0.0266], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0155, 0.0198, 0.0178, 0.0176, 0.0207, 0.0188, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:05:42,331 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:06:30,532 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 5100, loss[loss=0.1658, simple_loss=0.26, pruned_loss=0.03578, over 16694.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2727, pruned_loss=0.04542, over 3203815.69 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:06:45,787 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218254.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:07:10,612 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218271.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:07:22,622 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1500, 3.9994, 4.2155, 4.3361, 4.4842, 4.0806, 4.4071, 4.5078], device='cuda:3'), covar=tensor([0.1583, 0.1244, 0.1366, 0.0732, 0.0473, 0.1242, 0.0785, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0631, 0.0784, 0.0907, 0.0795, 0.0598, 0.0630, 0.0649, 0.0752], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:07:24,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7425, 3.1680, 3.1616, 1.8929, 2.7733, 2.2285, 3.1928, 3.4151], device='cuda:3'), covar=tensor([0.0381, 0.0776, 0.0716, 0.2112, 0.0928, 0.1005, 0.0767, 0.0839], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0164, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:07:46,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4856, 2.1545, 1.8235, 1.9293, 2.5154, 2.1958, 2.1323, 2.6458], device='cuda:3'), covar=tensor([0.0187, 0.0503, 0.0637, 0.0511, 0.0262, 0.0390, 0.0254, 0.0297], device='cuda:3'), in_proj_covar=tensor([0.0210, 0.0233, 0.0225, 0.0225, 0.0234, 0.0234, 0.0236, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:07:57,331 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 5150, loss[loss=0.1786, simple_loss=0.2775, pruned_loss=0.03988, over 16701.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2728, pruned_loss=0.04509, over 3191481.88 frames. ], batch size: 124, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:08:08,796 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0848, 2.0104, 2.2043, 3.6809, 1.9621, 2.3041, 2.1264, 2.1672], device='cuda:3'), covar=tensor([0.1655, 0.4326, 0.3183, 0.0675, 0.5121, 0.3265, 0.4161, 0.4043], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0449, 0.0368, 0.0328, 0.0436, 0.0517, 0.0421, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:08:23,130 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:08:58,541 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7292, 2.4442, 2.2097, 3.1927, 2.0886, 3.5384, 1.5156, 2.8122], device='cuda:3'), covar=tensor([0.1351, 0.0806, 0.1343, 0.0147, 0.0122, 0.0351, 0.1709, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0189, 0.0204, 0.0213, 0.0200, 0.0190], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:08:58,584 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4651, 3.3033, 2.7375, 2.2186, 2.2573, 2.3192, 3.4225, 2.9901], device='cuda:3'), covar=tensor([0.2700, 0.0598, 0.1745, 0.2811, 0.2525, 0.2080, 0.0545, 0.1252], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0269, 0.0304, 0.0314, 0.0298, 0.0259, 0.0295, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:08:59,855 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-05-01 11:09:11,242 INFO [train.py:904] (3/8) Epoch 22, batch 5200, loss[loss=0.1723, simple_loss=0.2594, pruned_loss=0.04261, over 16789.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2725, pruned_loss=0.0453, over 3183712.59 frames. ], batch size: 83, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:09:24,738 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7682, 4.8655, 5.1549, 5.1287, 5.1173, 4.8714, 4.7670, 4.6362], device='cuda:3'), covar=tensor([0.0296, 0.0464, 0.0297, 0.0333, 0.0385, 0.0304, 0.0829, 0.0409], device='cuda:3'), in_proj_covar=tensor([0.0395, 0.0438, 0.0425, 0.0393, 0.0470, 0.0444, 0.0530, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 11:09:28,266 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-01 11:09:33,041 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218368.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:10:23,969 INFO [optim.py:368] (3/8) 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 11:10:23,991 INFO [train.py:904] (3/8) Epoch 22, batch 5250, loss[loss=0.1519, simple_loss=0.2461, pruned_loss=0.02885, over 16802.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2691, pruned_loss=0.04444, over 3182960.76 frames. ], batch size: 102, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:37,150 INFO [train.py:904] (3/8) Epoch 22, batch 5300, loss[loss=0.1464, simple_loss=0.2386, pruned_loss=0.02709, over 16500.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2653, pruned_loss=0.04313, over 3199534.12 frames. ], batch size: 75, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:41,560 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218456.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:11:51,784 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7270, 1.8106, 2.3572, 2.6723, 2.6591, 3.1436, 2.1240, 3.0235], device='cuda:3'), covar=tensor([0.0204, 0.0592, 0.0325, 0.0337, 0.0340, 0.0164, 0.0525, 0.0126], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0196, 0.0181, 0.0186, 0.0199, 0.0154, 0.0199, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:12:51,253 INFO [optim.py:368] (3/8) 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 11:12:51,268 INFO [train.py:904] (3/8) Epoch 22, batch 5350, loss[loss=0.2076, simple_loss=0.2799, pruned_loss=0.06762, over 12353.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2638, pruned_loss=0.04267, over 3188351.06 frames. ], batch size: 246, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:13:00,625 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:13:21,742 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 11:13:43,231 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 5400, loss[loss=0.1791, simple_loss=0.2768, pruned_loss=0.04069, over 16502.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.266, pruned_loss=0.04291, over 3197345.79 frames. ], batch size: 75, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:14:06,161 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218554.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:14:23,141 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218566.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:14:28,894 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:34,748 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:47,865 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 11:15:18,766 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 22, batch 5450, loss[loss=0.2292, simple_loss=0.3114, pruned_loss=0.07347, over 16161.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2682, pruned_loss=0.04402, over 3188024.79 frames. ], batch size: 165, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:16:10,919 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218635.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:16:37,991 INFO [train.py:904] (3/8) Epoch 22, batch 5500, loss[loss=0.2117, simple_loss=0.2989, pruned_loss=0.06219, over 17276.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2752, pruned_loss=0.04807, over 3161138.36 frames. ], batch size: 52, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:17:57,929 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 5550, loss[loss=0.2904, simple_loss=0.3473, pruned_loss=0.1167, over 11460.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.283, pruned_loss=0.05364, over 3116541.33 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:13,553 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2157, 4.2625, 4.5855, 4.5404, 4.5692, 4.3072, 4.3007, 4.2082], device='cuda:3'), covar=tensor([0.0344, 0.0615, 0.0381, 0.0415, 0.0448, 0.0421, 0.0891, 0.0544], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0446, 0.0432, 0.0399, 0.0478, 0.0452, 0.0537, 0.0362], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 11:19:20,438 INFO [train.py:904] (3/8) Epoch 22, batch 5600, loss[loss=0.2194, simple_loss=0.2947, pruned_loss=0.07205, over 17040.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2887, pruned_loss=0.05848, over 3059381.12 frames. ], batch size: 53, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:25,776 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 22, batch 5650, loss[loss=0.2443, simple_loss=0.332, pruned_loss=0.07835, over 16471.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2933, pruned_loss=0.06214, over 3051684.44 frames. ], batch size: 75, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:20:44,086 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218804.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:35,407 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218838.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:37,100 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 11:21:56,896 INFO [train.py:904] (3/8) Epoch 22, batch 5700, loss[loss=0.1944, simple_loss=0.291, pruned_loss=0.04894, over 16680.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2945, pruned_loss=0.06359, over 3032009.24 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:22:16,160 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218865.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,549 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:48,657 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218886.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:13,875 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 5750, loss[loss=0.1876, simple_loss=0.2771, pruned_loss=0.04903, over 15313.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2968, pruned_loss=0.06461, over 3029967.68 frames. ], batch size: 191, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:23:32,116 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:33,946 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2497, 4.4055, 4.1825, 3.8874, 3.7237, 4.3205, 4.0773, 3.9578], device='cuda:3'), covar=tensor([0.0862, 0.0834, 0.0452, 0.0456, 0.1126, 0.0649, 0.0931, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0425, 0.0339, 0.0337, 0.0348, 0.0392, 0.0233, 0.0405], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:23:57,643 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:24:32,259 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8624, 4.9583, 4.7848, 4.4346, 4.2568, 4.8683, 4.7407, 4.4373], device='cuda:3'), covar=tensor([0.0840, 0.0870, 0.0366, 0.0439, 0.1327, 0.0646, 0.0531, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0290, 0.0425, 0.0338, 0.0337, 0.0348, 0.0391, 0.0233, 0.0404], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:24:34,384 INFO [train.py:904] (3/8) Epoch 22, batch 5800, loss[loss=0.1975, simple_loss=0.2887, pruned_loss=0.05317, over 15399.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2964, pruned_loss=0.06349, over 3039323.43 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:24:54,969 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:24,051 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:52,947 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5579, 3.6574, 2.7378, 2.2482, 2.4612, 2.4003, 3.9221, 3.2497], device='cuda:3'), covar=tensor([0.3025, 0.0644, 0.1853, 0.2832, 0.2671, 0.2081, 0.0463, 0.1342], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0269, 0.0305, 0.0314, 0.0297, 0.0259, 0.0296, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:25:53,553 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 5850, loss[loss=0.1936, simple_loss=0.2871, pruned_loss=0.04998, over 16911.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2943, pruned_loss=0.0619, over 3049437.17 frames. ], batch size: 109, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:26:03,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4652, 5.7741, 5.4980, 5.5773, 5.2350, 5.1941, 5.2038, 5.9043], device='cuda:3'), covar=tensor([0.1251, 0.0828, 0.1033, 0.0882, 0.0791, 0.0667, 0.1156, 0.0775], device='cuda:3'), in_proj_covar=tensor([0.0666, 0.0812, 0.0674, 0.0618, 0.0514, 0.0522, 0.0677, 0.0635], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:26:21,521 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:26:22,750 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9417, 4.1972, 4.0054, 4.0759, 3.7778, 3.8301, 3.8526, 4.1852], device='cuda:3'), covar=tensor([0.1147, 0.0869, 0.1032, 0.0823, 0.0780, 0.1660, 0.0930, 0.0999], device='cuda:3'), in_proj_covar=tensor([0.0665, 0.0811, 0.0674, 0.0618, 0.0514, 0.0522, 0.0676, 0.0634], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:26:29,997 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:02,418 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:14,927 INFO [train.py:904] (3/8) Epoch 22, batch 5900, loss[loss=0.2299, simple_loss=0.3066, pruned_loss=0.0766, over 17008.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2938, pruned_loss=0.06163, over 3057613.18 frames. ], batch size: 53, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:27:49,355 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:00,022 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 11:28:04,708 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:16,231 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 11:28:35,884 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 22, batch 5950, loss[loss=0.203, simple_loss=0.2973, pruned_loss=0.05433, over 16689.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2946, pruned_loss=0.0602, over 3089416.33 frames. ], batch size: 124, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:28:56,926 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 11:29:24,482 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:29:46,105 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 11:29:57,662 INFO [train.py:904] (3/8) Epoch 22, batch 6000, loss[loss=0.2098, simple_loss=0.2902, pruned_loss=0.06474, over 16650.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2933, pruned_loss=0.05956, over 3102751.04 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:29:57,662 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 11:30:07,617 INFO [train.py:938] (3/8) Epoch 22, validation: loss=0.1507, simple_loss=0.2632, pruned_loss=0.01907, over 944034.00 frames. 2023-05-01 11:30:07,618 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 11:30:27,783 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219165.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:28,922 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.979e+02 2.852e+02 3.411e+02 4.220e+02 7.320e+02, threshold=6.821e+02, percent-clipped=6.0 2023-05-01 11:31:28,937 INFO [train.py:904] (3/8) Epoch 22, batch 6050, loss[loss=0.1905, simple_loss=0.2905, pruned_loss=0.04522, over 16648.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2917, pruned_loss=0.05866, over 3119237.61 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:31:29,584 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:44,910 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:11,166 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219230.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:18,483 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 11:32:46,221 INFO [train.py:904] (3/8) Epoch 22, batch 6100, loss[loss=0.2351, simple_loss=0.3021, pruned_loss=0.08405, over 11369.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2912, pruned_loss=0.05811, over 3105037.43 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:33:05,426 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219264.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:26,349 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:34:04,107 INFO [train.py:904] (3/8) Epoch 22, batch 6150, loss[loss=0.1881, simple_loss=0.2788, pruned_loss=0.0487, over 16256.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2895, pruned_loss=0.05748, over 3113328.77 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:34:05,864 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:03,740 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:06,789 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219342.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:35:22,805 INFO [train.py:904] (3/8) Epoch 22, batch 6200, loss[loss=0.2131, simple_loss=0.301, pruned_loss=0.06256, over 16557.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2875, pruned_loss=0.05694, over 3125092.10 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:35:25,114 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 11:36:02,464 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 6250, loss[loss=0.1945, simple_loss=0.283, pruned_loss=0.05295, over 15540.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2872, pruned_loss=0.0569, over 3108631.40 frames. ], batch size: 191, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:43,133 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219403.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:36:43,732 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.827e+02 2.722e+02 3.215e+02 3.900e+02 7.497e+02, threshold=6.430e+02, percent-clipped=6.0 2023-05-01 11:37:20,351 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:37:41,083 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 11:37:57,562 INFO [train.py:904] (3/8) Epoch 22, batch 6300, loss[loss=0.1801, simple_loss=0.2752, pruned_loss=0.04254, over 16906.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2874, pruned_loss=0.05669, over 3104597.99 frames. ], batch size: 109, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:38:21,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2346, 4.0733, 3.8542, 4.3945, 4.4472, 4.2093, 4.4381, 4.4870], device='cuda:3'), covar=tensor([0.1850, 0.1593, 0.3042, 0.1186, 0.1205, 0.1679, 0.1331, 0.1279], device='cuda:3'), in_proj_covar=tensor([0.0629, 0.0780, 0.0903, 0.0789, 0.0597, 0.0626, 0.0650, 0.0750], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:38:34,925 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-05-01 11:38:51,343 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3093, 3.4367, 3.5804, 3.5451, 3.5784, 3.3834, 3.4319, 3.4873], device='cuda:3'), covar=tensor([0.0446, 0.0724, 0.0489, 0.0540, 0.0524, 0.0608, 0.0890, 0.0580], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0450, 0.0434, 0.0403, 0.0482, 0.0457, 0.0541, 0.0366], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 11:38:56,019 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219490.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:38:58,823 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 11:39:03,300 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3347, 5.6692, 5.1818, 5.5436, 5.1882, 4.9160, 5.1855, 5.7168], device='cuda:3'), covar=tensor([0.2207, 0.1553, 0.3029, 0.1233, 0.1486, 0.1525, 0.2189, 0.2008], device='cuda:3'), in_proj_covar=tensor([0.0668, 0.0816, 0.0678, 0.0621, 0.0517, 0.0525, 0.0682, 0.0639], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 11:39:15,298 INFO [train.py:904] (3/8) Epoch 22, batch 6350, loss[loss=0.1968, simple_loss=0.2811, pruned_loss=0.05629, over 16621.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2882, pruned_loss=0.05745, over 3110872.11 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:39:16,413 INFO [optim.py:368] (3/8) 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 11:40:28,613 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 6400, loss[loss=0.2068, simple_loss=0.2869, pruned_loss=0.06335, over 15270.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2884, pruned_loss=0.05857, over 3095257.25 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:40:40,461 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219559.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:41:00,814 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1693, 3.5540, 3.5417, 2.3625, 3.2758, 3.6558, 3.3590, 2.1190], device='cuda:3'), covar=tensor([0.0528, 0.0066, 0.0066, 0.0404, 0.0112, 0.0112, 0.0102, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0083, 0.0084, 0.0133, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:41:45,411 INFO [train.py:904] (3/8) Epoch 22, batch 6450, loss[loss=0.1971, simple_loss=0.2825, pruned_loss=0.05589, over 16211.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2885, pruned_loss=0.05794, over 3096829.44 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:41:47,189 INFO [optim.py:368] (3/8) 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 11:41:55,809 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 11:42:13,294 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:42:45,357 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 6500, loss[loss=0.1947, simple_loss=0.2813, pruned_loss=0.05404, over 16219.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2868, pruned_loss=0.05752, over 3086781.24 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:43:30,546 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:43,521 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:19,771 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 6550, loss[loss=0.2273, simple_loss=0.3226, pruned_loss=0.06595, over 16659.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2897, pruned_loss=0.05818, over 3092538.81 frames. ], batch size: 76, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:44:28,427 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.583e+02 3.105e+02 3.738e+02 8.268e+02, threshold=6.210e+02, percent-clipped=2.0 2023-05-01 11:45:03,112 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:08,242 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:46,181 INFO [train.py:904] (3/8) Epoch 22, batch 6600, loss[loss=0.2682, simple_loss=0.3266, pruned_loss=0.1049, over 11813.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2916, pruned_loss=0.05836, over 3106944.10 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:46:02,120 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219763.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:23,294 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:47:06,507 INFO [train.py:904] (3/8) Epoch 22, batch 6650, loss[loss=0.2721, simple_loss=0.3285, pruned_loss=0.1078, over 11425.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2927, pruned_loss=0.05983, over 3090108.89 frames. ], batch size: 247, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:47:07,642 INFO [optim.py:368] (3/8) 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 11:47:19,670 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7123, 3.9145, 2.4519, 4.6683, 3.0846, 4.5440, 2.5185, 3.1828], device='cuda:3'), covar=tensor([0.0325, 0.0365, 0.1721, 0.0142, 0.0772, 0.0440, 0.1553, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0161, 0.0175, 0.0215, 0.0200, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:47:39,840 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219824.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:10,768 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 6700, loss[loss=0.2102, simple_loss=0.2852, pruned_loss=0.06763, over 16393.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2906, pruned_loss=0.05974, over 3081475.88 frames. ], batch size: 35, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:48:27,532 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:30,103 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:49:05,528 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4583, 3.3115, 2.6388, 2.1852, 2.2129, 2.2621, 3.4017, 3.0526], device='cuda:3'), covar=tensor([0.2972, 0.0705, 0.1906, 0.2892, 0.2696, 0.2261, 0.0617, 0.1417], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0269, 0.0305, 0.0314, 0.0298, 0.0259, 0.0296, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:49:12,222 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0979, 5.1187, 5.4333, 5.4485, 5.5290, 5.1611, 5.1096, 4.7790], device='cuda:3'), covar=tensor([0.0307, 0.0536, 0.0424, 0.0417, 0.0484, 0.0356, 0.1000, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0450, 0.0435, 0.0404, 0.0484, 0.0458, 0.0541, 0.0367], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 11:49:36,042 INFO [train.py:904] (3/8) Epoch 22, batch 6750, loss[loss=0.1682, simple_loss=0.2659, pruned_loss=0.03527, over 16786.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2903, pruned_loss=0.06013, over 3073264.47 frames. ], batch size: 102, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:49:37,626 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-01 11:49:37,871 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.924e+02 3.493e+02 4.502e+02 8.889e+02, threshold=6.985e+02, percent-clipped=2.0 2023-05-01 11:49:43,850 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:00,673 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:50,299 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6398, 4.1939, 4.1134, 2.8778, 3.6963, 4.1728, 3.7202, 2.5714], device='cuda:3'), covar=tensor([0.0472, 0.0046, 0.0053, 0.0361, 0.0109, 0.0112, 0.0095, 0.0399], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0082, 0.0083, 0.0132, 0.0097, 0.0109, 0.0094, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:50:54,124 INFO [train.py:904] (3/8) Epoch 22, batch 6800, loss[loss=0.231, simple_loss=0.2994, pruned_loss=0.08136, over 11655.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2898, pruned_loss=0.06003, over 3058943.83 frames. ], batch size: 246, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:51:33,484 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 11:52:05,040 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219998.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:52:15,005 INFO [train.py:904] (3/8) Epoch 22, batch 6850, loss[loss=0.2158, simple_loss=0.3189, pruned_loss=0.05634, over 16831.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2916, pruned_loss=0.06034, over 3063938.86 frames. ], batch size: 116, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:52:16,798 INFO [optim.py:368] (3/8) 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 11:52:17,540 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-01 11:52:56,917 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 11:53:20,649 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220046.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:53:24,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7153, 2.6521, 2.4785, 4.2945, 2.9968, 4.0154, 1.5233, 2.8140], device='cuda:3'), covar=tensor([0.1423, 0.0793, 0.1233, 0.0153, 0.0230, 0.0390, 0.1746, 0.0857], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0187, 0.0204, 0.0213, 0.0200, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:53:31,114 INFO [train.py:904] (3/8) Epoch 22, batch 6900, loss[loss=0.2246, simple_loss=0.3042, pruned_loss=0.07247, over 15403.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2936, pruned_loss=0.05967, over 3074437.97 frames. ], batch size: 191, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:53:59,135 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6846, 2.5310, 2.2875, 3.6815, 2.5052, 3.7327, 1.4322, 2.5828], device='cuda:3'), covar=tensor([0.1465, 0.0809, 0.1356, 0.0177, 0.0220, 0.0427, 0.1872, 0.0960], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0172, 0.0193, 0.0188, 0.0204, 0.0213, 0.0200, 0.0191], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:54:14,037 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 11:54:24,092 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220087.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:41,881 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220098.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:49,991 INFO [train.py:904] (3/8) Epoch 22, batch 6950, loss[loss=0.1969, simple_loss=0.2814, pruned_loss=0.05619, over 16591.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2951, pruned_loss=0.06126, over 3069024.81 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:54:51,079 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220119.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:22,839 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4389, 3.2817, 2.6673, 2.2096, 2.1938, 2.2759, 3.4162, 2.9979], device='cuda:3'), covar=tensor([0.2908, 0.0700, 0.1742, 0.2645, 0.2682, 0.2215, 0.0489, 0.1313], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0267, 0.0303, 0.0312, 0.0296, 0.0258, 0.0293, 0.0334], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:55:54,590 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:57,641 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:03,774 INFO [train.py:904] (3/8) Epoch 22, batch 7000, loss[loss=0.2081, simple_loss=0.3052, pruned_loss=0.05554, over 16395.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2948, pruned_loss=0.06036, over 3070635.04 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 11:56:13,443 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 11:56:14,275 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:05,649 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220194.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:18,843 INFO [train.py:904] (3/8) Epoch 22, batch 7050, loss[loss=0.1992, simple_loss=0.2872, pruned_loss=0.05556, over 17183.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2955, pruned_loss=0.06041, over 3070211.35 frames. ], batch size: 46, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:57:21,954 INFO [optim.py:368] (3/8) 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 11:57:23,755 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2146, 2.8349, 3.1137, 1.8403, 3.2337, 3.3222, 2.6170, 2.5364], device='cuda:3'), covar=tensor([0.0826, 0.0292, 0.0242, 0.1140, 0.0099, 0.0204, 0.0512, 0.0482], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0107, 0.0098, 0.0138, 0.0080, 0.0125, 0.0128, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:57:35,285 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:35,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3642, 2.9554, 2.7068, 2.3202, 2.2827, 2.3444, 2.9973, 2.8311], device='cuda:3'), covar=tensor([0.2520, 0.0797, 0.1596, 0.2330, 0.2342, 0.2136, 0.0514, 0.1363], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0267, 0.0303, 0.0312, 0.0296, 0.0258, 0.0293, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 11:57:59,218 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:58:19,113 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 11:58:36,504 INFO [train.py:904] (3/8) Epoch 22, batch 7100, loss[loss=0.2376, simple_loss=0.3006, pruned_loss=0.08735, over 11837.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2936, pruned_loss=0.05982, over 3074360.04 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:58:45,522 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:07,799 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 11:59:32,873 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8842, 2.7358, 2.7298, 2.0865, 2.5768, 2.0911, 2.7205, 2.9336], device='cuda:3'), covar=tensor([0.0327, 0.0824, 0.0637, 0.1887, 0.0919, 0.1020, 0.0633, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0164, 0.0168, 0.0154, 0.0146, 0.0130, 0.0143, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 11:59:35,753 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:54,482 INFO [train.py:904] (3/8) Epoch 22, batch 7150, loss[loss=0.1947, simple_loss=0.2821, pruned_loss=0.05362, over 16451.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2917, pruned_loss=0.05963, over 3078153.23 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:59:58,136 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.530e+02 2.701e+02 3.416e+02 4.420e+02 1.024e+03, threshold=6.833e+02, percent-clipped=4.0 2023-05-01 12:00:20,926 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:01:07,948 INFO [train.py:904] (3/8) Epoch 22, batch 7200, loss[loss=0.187, simple_loss=0.275, pruned_loss=0.04946, over 16683.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2898, pruned_loss=0.05826, over 3073994.29 frames. ], batch size: 62, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:28,580 INFO [train.py:904] (3/8) Epoch 22, batch 7250, loss[loss=0.1801, simple_loss=0.2676, pruned_loss=0.04629, over 16231.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2877, pruned_loss=0.05716, over 3065755.36 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:30,892 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.548e+02 3.054e+02 3.716e+02 9.083e+02, threshold=6.108e+02, percent-clipped=2.0 2023-05-01 12:02:53,041 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220419.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:30,322 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 7300, loss[loss=0.2115, simple_loss=0.3035, pruned_loss=0.05978, over 15291.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2872, pruned_loss=0.05707, over 3065601.42 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:03:46,865 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220454.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:50,954 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4706, 3.5385, 2.7174, 2.1521, 2.4147, 2.3464, 3.8754, 3.2520], device='cuda:3'), covar=tensor([0.3163, 0.0698, 0.1917, 0.2818, 0.2698, 0.2184, 0.0491, 0.1309], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0267, 0.0303, 0.0313, 0.0297, 0.0258, 0.0294, 0.0335], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:04:07,574 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:05:02,411 INFO [train.py:904] (3/8) Epoch 22, batch 7350, loss[loss=0.1933, simple_loss=0.2808, pruned_loss=0.05289, over 17265.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2879, pruned_loss=0.05755, over 3060993.10 frames. ], batch size: 52, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:05:05,572 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.460e+02 2.713e+02 3.260e+02 3.989e+02 1.125e+03, threshold=6.520e+02, percent-clipped=5.0 2023-05-01 12:05:18,400 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220513.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:05:33,328 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6671, 3.7856, 2.3836, 4.3865, 2.8638, 4.3083, 2.3816, 3.0531], device='cuda:3'), covar=tensor([0.0278, 0.0340, 0.1519, 0.0158, 0.0767, 0.0448, 0.1538, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0173, 0.0191, 0.0159, 0.0173, 0.0213, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 12:06:06,088 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8438, 5.2979, 5.4398, 5.1824, 5.2132, 5.7849, 5.2718, 5.0577], device='cuda:3'), covar=tensor([0.0975, 0.1684, 0.2318, 0.1708, 0.2357, 0.0971, 0.1648, 0.2197], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0587, 0.0650, 0.0488, 0.0654, 0.0678, 0.0512, 0.0657], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:06:20,489 INFO [train.py:904] (3/8) Epoch 22, batch 7400, loss[loss=0.219, simple_loss=0.3127, pruned_loss=0.06266, over 16978.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2893, pruned_loss=0.05806, over 3068098.28 frames. ], batch size: 41, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:06:25,443 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:07:13,079 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:07:38,779 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8370, 2.0419, 2.3920, 3.1183, 2.1335, 2.2840, 2.3054, 2.2143], device='cuda:3'), covar=tensor([0.1429, 0.3543, 0.2462, 0.0741, 0.4161, 0.2416, 0.3206, 0.3188], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0445, 0.0363, 0.0323, 0.0434, 0.0513, 0.0417, 0.0520], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:07:41,479 INFO [train.py:904] (3/8) Epoch 22, batch 7450, loss[loss=0.2436, simple_loss=0.3064, pruned_loss=0.09042, over 11511.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2905, pruned_loss=0.05966, over 3038878.43 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:07:43,937 INFO [optim.py:368] (3/8) 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 12:08:01,211 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:08:04,702 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220616.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:09:00,783 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 7500, loss[loss=0.2068, simple_loss=0.2907, pruned_loss=0.06148, over 16588.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2901, pruned_loss=0.05854, over 3049021.55 frames. ], batch size: 62, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:09:29,058 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 12:10:21,149 INFO [train.py:904] (3/8) Epoch 22, batch 7550, loss[loss=0.1888, simple_loss=0.2688, pruned_loss=0.05435, over 16508.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2893, pruned_loss=0.0591, over 3051279.40 frames. ], batch size: 68, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:24,492 INFO [optim.py:368] (3/8) 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 12:10:35,472 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:10:35,606 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4788, 2.4786, 2.5460, 4.3858, 2.3421, 2.8807, 2.5856, 2.7100], device='cuda:3'), covar=tensor([0.1221, 0.3222, 0.2673, 0.0419, 0.3887, 0.2347, 0.3153, 0.3091], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0446, 0.0364, 0.0324, 0.0435, 0.0514, 0.0418, 0.0520], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:10:47,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5745, 5.9006, 5.6556, 5.6992, 5.3176, 5.2352, 5.3570, 6.0496], device='cuda:3'), covar=tensor([0.1283, 0.0844, 0.0950, 0.0857, 0.0873, 0.0686, 0.1189, 0.0787], device='cuda:3'), in_proj_covar=tensor([0.0670, 0.0818, 0.0677, 0.0622, 0.0516, 0.0530, 0.0684, 0.0640], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:11:20,680 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0150, 2.0264, 2.0682, 3.6158, 1.9985, 2.3545, 2.1867, 2.1858], device='cuda:3'), covar=tensor([0.1490, 0.3932, 0.3219, 0.0622, 0.4646, 0.2783, 0.3675, 0.3658], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0446, 0.0364, 0.0324, 0.0435, 0.0514, 0.0418, 0.0520], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:11:23,177 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:11:38,532 INFO [train.py:904] (3/8) Epoch 22, batch 7600, loss[loss=0.1916, simple_loss=0.2807, pruned_loss=0.05126, over 16472.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2878, pruned_loss=0.05843, over 3073442.56 frames. ], batch size: 68, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:11:40,757 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:37,570 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:55,414 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220802.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:56,365 INFO [train.py:904] (3/8) Epoch 22, batch 7650, loss[loss=0.2036, simple_loss=0.2865, pruned_loss=0.06032, over 16938.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2884, pruned_loss=0.0588, over 3079320.84 frames. ], batch size: 109, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:12:59,169 INFO [optim.py:368] (3/8) 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 12:13:27,098 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:13:50,660 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-05-01 12:14:11,869 INFO [train.py:904] (3/8) Epoch 22, batch 7700, loss[loss=0.2389, simple_loss=0.3102, pruned_loss=0.08378, over 11811.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2888, pruned_loss=0.05963, over 3070969.43 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:14:58,220 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:01,930 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220885.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:29,157 INFO [train.py:904] (3/8) Epoch 22, batch 7750, loss[loss=0.2078, simple_loss=0.2967, pruned_loss=0.05948, over 16939.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2885, pruned_loss=0.059, over 3085519.15 frames. ], batch size: 109, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:15:30,834 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:15:40,092 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220911.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:15:45,260 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:15,122 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:32,056 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9344, 5.4376, 5.6357, 5.3482, 5.3701, 5.9715, 5.3812, 5.1733], device='cuda:3'), covar=tensor([0.1034, 0.1750, 0.2155, 0.1772, 0.2503, 0.0873, 0.1622, 0.2396], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0593, 0.0655, 0.0492, 0.0658, 0.0684, 0.0518, 0.0662], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:16:42,219 INFO [train.py:904] (3/8) Epoch 22, batch 7800, loss[loss=0.2285, simple_loss=0.3072, pruned_loss=0.07485, over 16380.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2896, pruned_loss=0.05929, over 3096994.27 frames. ], batch size: 146, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:16:56,567 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:00,199 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3337, 4.2020, 4.3758, 4.5076, 4.6928, 4.1927, 4.6410, 4.6710], device='cuda:3'), covar=tensor([0.1897, 0.1247, 0.1519, 0.0814, 0.0613, 0.1275, 0.0826, 0.0775], device='cuda:3'), in_proj_covar=tensor([0.0629, 0.0776, 0.0899, 0.0783, 0.0597, 0.0623, 0.0650, 0.0749], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:17:00,291 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:23,602 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 12:17:55,724 INFO [train.py:904] (3/8) Epoch 22, batch 7850, loss[loss=0.1845, simple_loss=0.2731, pruned_loss=0.04794, over 16737.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2904, pruned_loss=0.0595, over 3072654.24 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:17:58,015 INFO [optim.py:368] (3/8) 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 12:18:02,926 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:18:03,118 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3037, 2.5725, 2.2576, 2.3541, 2.9642, 2.5469, 2.9415, 3.1092], device='cuda:3'), covar=tensor([0.0138, 0.0428, 0.0508, 0.0451, 0.0260, 0.0392, 0.0214, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0207, 0.0232, 0.0224, 0.0223, 0.0234, 0.0231, 0.0232, 0.0227], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:19:09,447 INFO [train.py:904] (3/8) Epoch 22, batch 7900, loss[loss=0.2358, simple_loss=0.299, pruned_loss=0.08631, over 11268.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2899, pruned_loss=0.059, over 3090364.61 frames. ], batch size: 247, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:19:30,727 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 12:19:50,537 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0529, 2.3125, 2.2928, 2.7681, 1.7514, 3.1709, 1.8726, 2.6851], device='cuda:3'), covar=tensor([0.1175, 0.0699, 0.1047, 0.0233, 0.0119, 0.0376, 0.1479, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0190, 0.0207, 0.0214, 0.0202, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 12:20:07,698 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3951, 3.3689, 3.4089, 3.4837, 3.5248, 3.3188, 3.5177, 3.5713], device='cuda:3'), covar=tensor([0.1234, 0.0929, 0.1026, 0.0611, 0.0692, 0.2026, 0.1001, 0.0814], device='cuda:3'), in_proj_covar=tensor([0.0630, 0.0778, 0.0903, 0.0785, 0.0598, 0.0625, 0.0652, 0.0752], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:20:27,100 INFO [train.py:904] (3/8) Epoch 22, batch 7950, loss[loss=0.1827, simple_loss=0.2699, pruned_loss=0.04768, over 16877.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2907, pruned_loss=0.05955, over 3084312.20 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:20:32,010 INFO [optim.py:368] (3/8) 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 12:21:31,819 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5372, 4.7519, 4.9195, 4.6977, 4.7819, 5.3410, 4.7558, 4.4978], device='cuda:3'), covar=tensor([0.1344, 0.1782, 0.2175, 0.1869, 0.2238, 0.0940, 0.1687, 0.2501], device='cuda:3'), in_proj_covar=tensor([0.0412, 0.0592, 0.0655, 0.0491, 0.0655, 0.0684, 0.0517, 0.0660], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:21:41,819 INFO [train.py:904] (3/8) Epoch 22, batch 8000, loss[loss=0.1997, simple_loss=0.289, pruned_loss=0.05525, over 16828.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2905, pruned_loss=0.05957, over 3088405.74 frames. ], batch size: 83, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:22:19,111 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:22:55,166 INFO [train.py:904] (3/8) Epoch 22, batch 8050, loss[loss=0.2183, simple_loss=0.3077, pruned_loss=0.06439, over 15498.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2911, pruned_loss=0.05977, over 3085540.17 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:23:01,252 INFO [optim.py:368] (3/8) 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 12:23:07,168 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:24:08,783 INFO [train.py:904] (3/8) Epoch 22, batch 8100, loss[loss=0.1941, simple_loss=0.2807, pruned_loss=0.05373, over 17010.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2909, pruned_loss=0.05933, over 3072119.00 frames. ], batch size: 41, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:24:17,302 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:24:19,975 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221260.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:24:24,898 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 12:25:24,332 INFO [train.py:904] (3/8) Epoch 22, batch 8150, loss[loss=0.2099, simple_loss=0.2876, pruned_loss=0.06615, over 16395.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2877, pruned_loss=0.05785, over 3092411.56 frames. ], batch size: 35, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:25:31,011 INFO [optim.py:368] (3/8) 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 12:25:31,340 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221307.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:26:41,083 INFO [train.py:904] (3/8) Epoch 22, batch 8200, loss[loss=0.1883, simple_loss=0.2812, pruned_loss=0.04768, over 16865.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2851, pruned_loss=0.05665, over 3108416.83 frames. ], batch size: 116, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:26:44,114 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 8250, loss[loss=0.1612, simple_loss=0.2463, pruned_loss=0.0381, over 12100.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2848, pruned_loss=0.05475, over 3085186.61 frames. ], batch size: 247, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:28:05,605 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.638e+02 3.060e+02 3.657e+02 7.056e+02, threshold=6.120e+02, percent-clipped=1.0 2023-05-01 12:29:17,814 INFO [train.py:904] (3/8) Epoch 22, batch 8300, loss[loss=0.1958, simple_loss=0.2714, pruned_loss=0.06009, over 11992.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2818, pruned_loss=0.05181, over 3067092.98 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:29:57,535 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:30:38,296 INFO [train.py:904] (3/8) Epoch 22, batch 8350, loss[loss=0.2067, simple_loss=0.2862, pruned_loss=0.06356, over 12082.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2805, pruned_loss=0.04987, over 3051588.00 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:30:43,698 INFO [optim.py:368] (3/8) 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 12:30:44,205 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:15,660 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221526.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:29,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2338, 4.3043, 4.1135, 3.8215, 3.7851, 4.2257, 3.8628, 3.9605], device='cuda:3'), covar=tensor([0.0571, 0.0633, 0.0331, 0.0326, 0.0882, 0.0545, 0.0877, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0422, 0.0336, 0.0331, 0.0341, 0.0387, 0.0231, 0.0401], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:31:56,139 INFO [train.py:904] (3/8) Epoch 22, batch 8400, loss[loss=0.1845, simple_loss=0.2825, pruned_loss=0.04328, over 16653.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2782, pruned_loss=0.04793, over 3060369.26 frames. ], batch size: 134, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:32:08,954 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:32:19,762 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4680, 5.8138, 5.5337, 5.6199, 5.2710, 5.2535, 5.2085, 5.9243], device='cuda:3'), covar=tensor([0.1245, 0.0839, 0.1115, 0.0863, 0.0847, 0.0673, 0.1162, 0.0938], device='cuda:3'), in_proj_covar=tensor([0.0662, 0.0803, 0.0666, 0.0610, 0.0505, 0.0521, 0.0672, 0.0629], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:32:21,003 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:33:17,378 INFO [train.py:904] (3/8) Epoch 22, batch 8450, loss[loss=0.178, simple_loss=0.2729, pruned_loss=0.04158, over 15331.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2766, pruned_loss=0.04669, over 3039234.24 frames. ], batch size: 190, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:33:24,323 INFO [optim.py:368] (3/8) 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 12:33:26,153 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221608.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:34:38,832 INFO [train.py:904] (3/8) Epoch 22, batch 8500, loss[loss=0.1655, simple_loss=0.2582, pruned_loss=0.03637, over 16457.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2733, pruned_loss=0.04486, over 3044510.73 frames. ], batch size: 75, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:35:39,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4629, 4.4149, 4.7953, 4.7680, 4.7573, 4.5433, 4.4488, 4.4143], device='cuda:3'), covar=tensor([0.0423, 0.0849, 0.0523, 0.0534, 0.0628, 0.0510, 0.1196, 0.0573], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0457, 0.0440, 0.0408, 0.0487, 0.0463, 0.0547, 0.0371], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 12:35:44,621 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 12:35:59,837 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 12:36:02,531 INFO [train.py:904] (3/8) Epoch 22, batch 8550, loss[loss=0.1769, simple_loss=0.279, pruned_loss=0.03736, over 16748.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2706, pruned_loss=0.04366, over 3041174.98 frames. ], batch size: 89, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:10,062 INFO [optim.py:368] (3/8) 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 12:37:41,421 INFO [train.py:904] (3/8) Epoch 22, batch 8600, loss[loss=0.2001, simple_loss=0.2981, pruned_loss=0.05103, over 16418.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.271, pruned_loss=0.04285, over 3044837.63 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:37:53,232 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:39:21,614 INFO [train.py:904] (3/8) Epoch 22, batch 8650, loss[loss=0.1558, simple_loss=0.2556, pruned_loss=0.02798, over 15313.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2691, pruned_loss=0.04161, over 3040677.08 frames. ], batch size: 190, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:39:30,557 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.228e+02 2.506e+02 3.024e+02 6.242e+02, threshold=5.012e+02, percent-clipped=2.0 2023-05-01 12:39:58,956 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:40:39,765 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 12:41:06,645 INFO [train.py:904] (3/8) Epoch 22, batch 8700, loss[loss=0.1741, simple_loss=0.2629, pruned_loss=0.04268, over 16865.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2659, pruned_loss=0.04003, over 3040688.72 frames. ], batch size: 116, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:41:27,670 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221863.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:41:42,976 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5313, 4.5274, 4.8773, 4.8563, 4.8498, 4.5986, 4.5501, 4.4831], device='cuda:3'), covar=tensor([0.0364, 0.0670, 0.0479, 0.0592, 0.0587, 0.0639, 0.0964, 0.0464], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0449, 0.0434, 0.0400, 0.0480, 0.0454, 0.0538, 0.0365], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 12:42:16,869 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 8750, loss[loss=0.1877, simple_loss=0.283, pruned_loss=0.04626, over 17001.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.266, pruned_loss=0.03984, over 3047180.65 frames. ], batch size: 109, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:42:53,179 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.058e+02 2.561e+02 3.126e+02 5.404e+02, threshold=5.122e+02, percent-clipped=2.0 2023-05-01 12:44:31,263 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:44:34,618 INFO [train.py:904] (3/8) Epoch 22, batch 8800, loss[loss=0.173, simple_loss=0.2652, pruned_loss=0.04039, over 12846.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2654, pruned_loss=0.0391, over 3066603.10 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:44:50,906 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8416, 2.6869, 2.9012, 2.1176, 2.6577, 2.1785, 2.7857, 2.9005], device='cuda:3'), covar=tensor([0.0270, 0.0910, 0.0531, 0.1836, 0.0817, 0.0959, 0.0599, 0.0830], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0159, 0.0163, 0.0151, 0.0142, 0.0127, 0.0140, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 12:44:54,618 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8286, 3.7216, 3.8371, 3.9869, 4.0571, 3.6700, 4.0447, 4.1082], device='cuda:3'), covar=tensor([0.1601, 0.1211, 0.1516, 0.0770, 0.0705, 0.1982, 0.0812, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0615, 0.0761, 0.0880, 0.0768, 0.0586, 0.0612, 0.0637, 0.0738], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:45:49,204 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 12:45:56,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8823, 3.8149, 3.9671, 4.0331, 4.1035, 3.6944, 4.1062, 4.1514], device='cuda:3'), covar=tensor([0.1415, 0.1016, 0.1064, 0.0598, 0.0541, 0.1856, 0.0556, 0.0621], device='cuda:3'), in_proj_covar=tensor([0.0613, 0.0759, 0.0878, 0.0766, 0.0584, 0.0611, 0.0634, 0.0735], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:46:22,303 INFO [train.py:904] (3/8) Epoch 22, batch 8850, loss[loss=0.1643, simple_loss=0.27, pruned_loss=0.02931, over 15572.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2672, pruned_loss=0.03836, over 3050512.02 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:46:28,904 INFO [optim.py:368] (3/8) 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 12:46:35,315 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:47:15,554 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0784, 2.1162, 2.0684, 3.6562, 2.0398, 2.3866, 2.2161, 2.2001], device='cuda:3'), covar=tensor([0.1340, 0.3695, 0.3414, 0.0573, 0.4685, 0.2817, 0.3870, 0.3937], device='cuda:3'), in_proj_covar=tensor([0.0394, 0.0441, 0.0363, 0.0319, 0.0432, 0.0508, 0.0414, 0.0514], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:48:07,872 INFO [train.py:904] (3/8) Epoch 22, batch 8900, loss[loss=0.1771, simple_loss=0.2751, pruned_loss=0.03957, over 15315.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2681, pruned_loss=0.0382, over 3056730.79 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:48:42,798 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222070.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:48:50,380 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8556, 2.8001, 2.7243, 1.9479, 2.5991, 2.8069, 2.6838, 1.9528], device='cuda:3'), covar=tensor([0.0449, 0.0067, 0.0069, 0.0378, 0.0113, 0.0097, 0.0093, 0.0413], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0081, 0.0082, 0.0131, 0.0096, 0.0107, 0.0093, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:50:11,677 INFO [train.py:904] (3/8) Epoch 22, batch 8950, loss[loss=0.1823, simple_loss=0.269, pruned_loss=0.04779, over 12790.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2674, pruned_loss=0.03829, over 3061734.84 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:50:23,607 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:51:12,419 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 12:51:36,416 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4469, 3.3165, 3.5767, 1.8735, 3.7238, 3.8073, 2.9861, 2.9021], device='cuda:3'), covar=tensor([0.0766, 0.0247, 0.0173, 0.1198, 0.0076, 0.0126, 0.0403, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0104, 0.0093, 0.0134, 0.0077, 0.0118, 0.0124, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 12:51:44,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1556, 3.4899, 3.5614, 2.2453, 2.9827, 2.5252, 3.6790, 3.6531], device='cuda:3'), covar=tensor([0.0267, 0.0839, 0.0622, 0.2056, 0.0846, 0.0899, 0.0589, 0.0928], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0158, 0.0161, 0.0149, 0.0141, 0.0126, 0.0138, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 12:52:03,799 INFO [train.py:904] (3/8) Epoch 22, batch 9000, loss[loss=0.1576, simple_loss=0.2485, pruned_loss=0.03332, over 16917.00 frames. ], tot_loss[loss=0.169, simple_loss=0.264, pruned_loss=0.03697, over 3066669.52 frames. ], batch size: 116, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:52:03,799 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 12:52:14,709 INFO [train.py:938] (3/8) Epoch 22, validation: loss=0.1453, simple_loss=0.2492, pruned_loss=0.0207, over 944034.00 frames. 2023-05-01 12:52:14,710 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 12:52:36,386 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:22,216 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:58,629 INFO [train.py:904] (3/8) Epoch 22, batch 9050, loss[loss=0.1545, simple_loss=0.2443, pruned_loss=0.03238, over 16784.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2658, pruned_loss=0.03762, over 3057535.05 frames. ], batch size: 90, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:54:04,034 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,498 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,630 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222246.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:55:44,665 INFO [train.py:904] (3/8) Epoch 22, batch 9100, loss[loss=0.1597, simple_loss=0.2616, pruned_loss=0.02893, over 16602.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2645, pruned_loss=0.03742, over 3070017.30 frames. ], batch size: 62, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:56:11,336 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 9150, loss[loss=0.1616, simple_loss=0.2525, pruned_loss=0.03533, over 12368.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2654, pruned_loss=0.03729, over 3065288.62 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:57:53,760 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.177e+02 2.613e+02 3.280e+02 4.905e+02, threshold=5.227e+02, percent-clipped=0.0 2023-05-01 12:58:44,285 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4687, 1.8156, 2.1408, 2.4297, 2.4515, 2.7196, 1.9007, 2.6475], device='cuda:3'), covar=tensor([0.0262, 0.0596, 0.0353, 0.0385, 0.0391, 0.0249, 0.0581, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0183, 0.0188, 0.0174, 0.0178, 0.0192, 0.0148, 0.0191, 0.0146], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:59:19,548 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0159, 2.1261, 2.2225, 3.4309, 2.1006, 2.3453, 2.2774, 2.2402], device='cuda:3'), covar=tensor([0.1468, 0.3843, 0.3260, 0.0742, 0.4573, 0.2904, 0.3434, 0.3874], device='cuda:3'), in_proj_covar=tensor([0.0391, 0.0438, 0.0360, 0.0317, 0.0428, 0.0504, 0.0410, 0.0510], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:59:25,382 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2597, 4.3737, 4.4549, 4.2731, 4.3243, 4.8556, 4.3859, 4.0531], device='cuda:3'), covar=tensor([0.1576, 0.1703, 0.2009, 0.2032, 0.2591, 0.0933, 0.1600, 0.2312], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0568, 0.0630, 0.0472, 0.0628, 0.0658, 0.0498, 0.0634], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:59:26,803 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7101, 5.0604, 4.8397, 4.8677, 4.5977, 4.5952, 4.4299, 5.1432], device='cuda:3'), covar=tensor([0.1262, 0.0927, 0.0998, 0.0855, 0.0779, 0.0974, 0.1284, 0.0878], device='cuda:3'), in_proj_covar=tensor([0.0650, 0.0790, 0.0652, 0.0600, 0.0497, 0.0511, 0.0661, 0.0619], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 12:59:28,938 INFO [train.py:904] (3/8) Epoch 22, batch 9200, loss[loss=0.1576, simple_loss=0.2454, pruned_loss=0.03488, over 12220.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2605, pruned_loss=0.03586, over 3068503.80 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:59:30,254 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2181, 5.8296, 5.9062, 5.6265, 5.6511, 6.2280, 5.7428, 5.4487], device='cuda:3'), covar=tensor([0.0748, 0.1535, 0.1725, 0.1766, 0.2180, 0.0742, 0.1467, 0.2041], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0568, 0.0629, 0.0472, 0.0628, 0.0658, 0.0498, 0.0634], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:59:34,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0361, 5.4900, 5.5415, 5.3575, 5.3300, 5.9249, 5.3887, 5.0810], device='cuda:3'), covar=tensor([0.0840, 0.1515, 0.1843, 0.1801, 0.2276, 0.0719, 0.1526, 0.2100], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0568, 0.0629, 0.0472, 0.0628, 0.0658, 0.0498, 0.0633], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 12:59:52,286 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:00:15,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6779, 4.6630, 4.4138, 3.8142, 4.5315, 1.7491, 4.2748, 4.2514], device='cuda:3'), covar=tensor([0.0090, 0.0094, 0.0213, 0.0314, 0.0112, 0.2605, 0.0135, 0.0216], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0152, 0.0193, 0.0171, 0.0171, 0.0204, 0.0182, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:00:38,082 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 13:01:05,892 INFO [train.py:904] (3/8) Epoch 22, batch 9250, loss[loss=0.1673, simple_loss=0.2605, pruned_loss=0.03704, over 16563.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2604, pruned_loss=0.03607, over 3078119.32 frames. ], batch size: 148, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:01:16,244 INFO [optim.py:368] (3/8) 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 13:01:28,845 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:02:56,973 INFO [train.py:904] (3/8) Epoch 22, batch 9300, loss[loss=0.1649, simple_loss=0.2451, pruned_loss=0.04235, over 12584.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2586, pruned_loss=0.03574, over 3060128.03 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:03:16,991 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:04:19,646 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0755, 4.0509, 3.9396, 3.2827, 4.0024, 1.8440, 3.8058, 3.5909], device='cuda:3'), covar=tensor([0.0110, 0.0115, 0.0184, 0.0264, 0.0119, 0.2684, 0.0135, 0.0254], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0151, 0.0192, 0.0170, 0.0170, 0.0204, 0.0181, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:04:40,773 INFO [train.py:904] (3/8) Epoch 22, batch 9350, loss[loss=0.165, simple_loss=0.2572, pruned_loss=0.03639, over 16370.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2589, pruned_loss=0.03595, over 3061855.53 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:04:49,912 INFO [optim.py:368] (3/8) 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 13:05:23,386 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:05:57,777 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222541.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:06:08,602 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:20,192 INFO [train.py:904] (3/8) Epoch 22, batch 9400, loss[loss=0.1576, simple_loss=0.249, pruned_loss=0.03312, over 16935.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2591, pruned_loss=0.03555, over 3076275.10 frames. ], batch size: 41, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:06:38,150 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:55,308 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:24,535 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222584.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:07:43,446 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:08:00,592 INFO [train.py:904] (3/8) Epoch 22, batch 9450, loss[loss=0.1605, simple_loss=0.2582, pruned_loss=0.03142, over 16301.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2604, pruned_loss=0.03587, over 3055031.90 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:08:08,413 INFO [optim.py:368] (3/8) 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 13:08:20,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7724, 4.7634, 4.5292, 3.9360, 4.6363, 1.8030, 4.3830, 4.3730], device='cuda:3'), covar=tensor([0.0077, 0.0116, 0.0167, 0.0304, 0.0094, 0.2569, 0.0122, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0151, 0.0191, 0.0169, 0.0170, 0.0203, 0.0181, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:08:56,817 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:09:08,359 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 13:09:40,758 INFO [train.py:904] (3/8) Epoch 22, batch 9500, loss[loss=0.1606, simple_loss=0.2557, pruned_loss=0.03278, over 16909.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2597, pruned_loss=0.03554, over 3060005.05 frames. ], batch size: 116, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:09:53,255 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0984, 2.3022, 1.7606, 1.9864, 2.6708, 2.3303, 2.7772, 2.9293], device='cuda:3'), covar=tensor([0.0188, 0.0543, 0.0822, 0.0653, 0.0355, 0.0533, 0.0233, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0228, 0.0220, 0.0220, 0.0228, 0.0227, 0.0224, 0.0220], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:10:06,943 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:10:31,518 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 13:10:54,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7529, 2.4492, 2.2639, 3.5744, 2.0338, 3.6736, 1.4911, 2.8155], device='cuda:3'), covar=tensor([0.1455, 0.0873, 0.1321, 0.0147, 0.0090, 0.0351, 0.1898, 0.0761], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0171, 0.0190, 0.0181, 0.0197, 0.0209, 0.0199, 0.0188], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:11:22,395 INFO [train.py:904] (3/8) Epoch 22, batch 9550, loss[loss=0.1611, simple_loss=0.254, pruned_loss=0.03412, over 12504.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2592, pruned_loss=0.03549, over 3063738.32 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:11:34,510 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:13:00,661 INFO [train.py:904] (3/8) Epoch 22, batch 9600, loss[loss=0.1628, simple_loss=0.2469, pruned_loss=0.03942, over 12810.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2607, pruned_loss=0.03622, over 3065573.05 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:14:14,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8173, 3.7141, 3.8820, 3.9984, 4.0932, 3.6430, 4.0787, 4.1136], device='cuda:3'), covar=tensor([0.1623, 0.1228, 0.1499, 0.0745, 0.0604, 0.1809, 0.0663, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0607, 0.0752, 0.0868, 0.0758, 0.0576, 0.0602, 0.0630, 0.0729], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:14:44,326 INFO [train.py:904] (3/8) Epoch 22, batch 9650, loss[loss=0.1796, simple_loss=0.278, pruned_loss=0.0406, over 16290.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2626, pruned_loss=0.03653, over 3058271.53 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:14:49,475 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 13:14:58,744 INFO [optim.py:368] (3/8) 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 13:16:03,721 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222841.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:16:27,395 INFO [train.py:904] (3/8) Epoch 22, batch 9700, loss[loss=0.1711, simple_loss=0.2628, pruned_loss=0.0397, over 16305.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.262, pruned_loss=0.03642, over 3072907.79 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:16:43,287 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:17:22,888 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:08,540 INFO [train.py:904] (3/8) Epoch 22, batch 9750, loss[loss=0.1528, simple_loss=0.2453, pruned_loss=0.03016, over 16662.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2606, pruned_loss=0.03653, over 3071769.78 frames. ], batch size: 62, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:18:18,158 INFO [optim.py:368] (3/8) 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 13:18:19,216 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:52,780 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222926.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:24,097 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:36,643 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 13:19:45,653 INFO [train.py:904] (3/8) Epoch 22, batch 9800, loss[loss=0.176, simple_loss=0.2821, pruned_loss=0.03492, over 15235.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2608, pruned_loss=0.03574, over 3078003.02 frames. ], batch size: 190, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:20:15,153 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:40,822 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:23,020 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:26,491 INFO [train.py:904] (3/8) Epoch 22, batch 9850, loss[loss=0.1805, simple_loss=0.2765, pruned_loss=0.04227, over 15369.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2619, pruned_loss=0.03547, over 3076379.57 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:21:37,442 INFO [optim.py:368] (3/8) 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 13:21:54,135 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 13:22:00,932 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 13:22:19,829 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223030.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:22:54,006 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0668, 4.0251, 3.9378, 3.2887, 3.9469, 1.7772, 3.7330, 3.4894], device='cuda:3'), covar=tensor([0.0120, 0.0133, 0.0185, 0.0256, 0.0136, 0.2747, 0.0146, 0.0285], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0150, 0.0190, 0.0167, 0.0170, 0.0203, 0.0180, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:23:01,405 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 9900, loss[loss=0.1604, simple_loss=0.2695, pruned_loss=0.02572, over 15487.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2618, pruned_loss=0.03527, over 3062268.47 frames. ], batch size: 192, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:23:43,530 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:24:08,761 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9229, 2.6686, 2.9063, 2.0323, 2.7086, 2.1498, 2.6440, 2.8420], device='cuda:3'), covar=tensor([0.0285, 0.1004, 0.0440, 0.1968, 0.0759, 0.0929, 0.0630, 0.0961], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0157, 0.0162, 0.0150, 0.0141, 0.0127, 0.0139, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:24:32,075 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 13:25:04,988 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-01 13:25:13,510 INFO [train.py:904] (3/8) Epoch 22, batch 9950, loss[loss=0.1651, simple_loss=0.2647, pruned_loss=0.03274, over 16560.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2641, pruned_loss=0.03564, over 3058359.44 frames. ], batch size: 57, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:25:27,892 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.209e+02 2.588e+02 3.009e+02 4.332e+02, threshold=5.177e+02, percent-clipped=0.0 2023-05-01 13:25:37,295 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8822, 4.8894, 4.7399, 4.3176, 4.4266, 4.7893, 4.6544, 4.4590], device='cuda:3'), covar=tensor([0.0522, 0.0619, 0.0319, 0.0316, 0.0868, 0.0576, 0.0340, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0277, 0.0404, 0.0324, 0.0319, 0.0327, 0.0374, 0.0222, 0.0384], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-05-01 13:26:07,003 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:26:44,996 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0302, 2.7751, 2.9076, 2.1737, 2.6797, 2.1533, 2.7546, 2.9240], device='cuda:3'), covar=tensor([0.0243, 0.0742, 0.0526, 0.1642, 0.0726, 0.0868, 0.0578, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0156, 0.0162, 0.0150, 0.0141, 0.0126, 0.0139, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:27:13,200 INFO [train.py:904] (3/8) Epoch 22, batch 10000, loss[loss=0.1498, simple_loss=0.2446, pruned_loss=0.02745, over 17002.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2622, pruned_loss=0.03494, over 3072000.49 frames. ], batch size: 55, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:27:51,703 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223173.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:28:04,813 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 22, batch 10050, loss[loss=0.1674, simple_loss=0.2583, pruned_loss=0.03829, over 12135.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2621, pruned_loss=0.03462, over 3075643.29 frames. ], batch size: 248, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:29:04,328 INFO [optim.py:368] (3/8) 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 13:29:40,706 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:42,095 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223227.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:42,769 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 13:29:54,210 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223234.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:30:11,363 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 13:30:16,592 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-01 13:30:27,326 INFO [train.py:904] (3/8) Epoch 22, batch 10100, loss[loss=0.1598, simple_loss=0.2535, pruned_loss=0.03305, over 16598.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2624, pruned_loss=0.03491, over 3076611.01 frames. ], batch size: 57, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:31:10,809 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:39,310 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223296.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:32:13,430 INFO [train.py:904] (3/8) Epoch 23, batch 0, loss[loss=0.2392, simple_loss=0.321, pruned_loss=0.07869, over 12219.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.321, pruned_loss=0.07869, over 12219.00 frames. ], batch size: 247, lr: 2.97e-03, grad_scale: 8.0 2023-05-01 13:32:13,430 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 13:32:20,849 INFO [train.py:938] (3/8) Epoch 23, validation: loss=0.1454, simple_loss=0.2487, pruned_loss=0.02111, over 944034.00 frames. 2023-05-01 13:32:20,849 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 13:32:28,407 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.627e+02 3.041e+02 3.739e+02 7.225e+02, threshold=6.083e+02, percent-clipped=7.0 2023-05-01 13:32:40,767 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5581, 3.8039, 3.9960, 3.9712, 3.9989, 3.7869, 3.5311, 3.8020], device='cuda:3'), covar=tensor([0.0664, 0.0908, 0.0655, 0.0717, 0.0847, 0.0741, 0.1524, 0.0672], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0436, 0.0426, 0.0393, 0.0467, 0.0445, 0.0524, 0.0357], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 13:32:49,519 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:26,109 INFO [train.py:904] (3/8) Epoch 23, batch 50, loss[loss=0.1947, simple_loss=0.2842, pruned_loss=0.05262, over 16676.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2714, pruned_loss=0.05031, over 744597.61 frames. ], batch size: 62, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:32,069 INFO [train.py:904] (3/8) Epoch 23, batch 100, loss[loss=0.1614, simple_loss=0.2588, pruned_loss=0.03196, over 17140.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2678, pruned_loss=0.04762, over 1320381.06 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:33,723 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8682, 5.1673, 5.2952, 5.0721, 5.1106, 5.7333, 5.1997, 4.9225], device='cuda:3'), covar=tensor([0.1254, 0.2013, 0.2337, 0.2387, 0.2987, 0.1058, 0.1694, 0.2548], device='cuda:3'), in_proj_covar=tensor([0.0390, 0.0567, 0.0627, 0.0468, 0.0629, 0.0653, 0.0494, 0.0627], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 13:34:42,058 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.242e+02 2.764e+02 3.330e+02 6.883e+02, threshold=5.527e+02, percent-clipped=1.0 2023-05-01 13:34:55,417 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:35:03,882 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9027, 3.0198, 3.3052, 2.1087, 2.7810, 2.1166, 3.3848, 3.3969], device='cuda:3'), covar=tensor([0.0289, 0.0911, 0.0617, 0.1910, 0.0939, 0.1092, 0.0606, 0.0998], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0159, 0.0164, 0.0151, 0.0143, 0.0128, 0.0140, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:35:38,698 INFO [train.py:904] (3/8) Epoch 23, batch 150, loss[loss=0.1971, simple_loss=0.2808, pruned_loss=0.05674, over 16485.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2657, pruned_loss=0.04589, over 1770699.26 frames. ], batch size: 68, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:35:40,371 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3599, 4.2956, 4.3207, 3.4184, 4.2960, 1.6249, 3.9448, 3.9181], device='cuda:3'), covar=tensor([0.0270, 0.0203, 0.0261, 0.0653, 0.0214, 0.3521, 0.0273, 0.0430], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0153, 0.0193, 0.0169, 0.0171, 0.0205, 0.0183, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:35:53,429 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5169, 3.6683, 4.1418, 2.3475, 3.2311, 2.4562, 3.8623, 3.8703], device='cuda:3'), covar=tensor([0.0283, 0.0886, 0.0478, 0.1874, 0.0831, 0.1012, 0.0638, 0.1085], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0159, 0.0164, 0.0151, 0.0143, 0.0128, 0.0140, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:36:47,606 INFO [train.py:904] (3/8) Epoch 23, batch 200, loss[loss=0.1601, simple_loss=0.2448, pruned_loss=0.03773, over 16140.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2645, pruned_loss=0.04489, over 2110908.85 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:57,903 INFO [optim.py:368] (3/8) 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 13:37:21,173 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223529.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 13:37:52,727 INFO [train.py:904] (3/8) Epoch 23, batch 250, loss[loss=0.1453, simple_loss=0.2291, pruned_loss=0.03071, over 16988.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2614, pruned_loss=0.04464, over 2366245.98 frames. ], batch size: 41, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:38:06,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 13:38:29,103 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:38:54,845 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 300, loss[loss=0.1588, simple_loss=0.2526, pruned_loss=0.03251, over 17133.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2581, pruned_loss=0.04292, over 2583336.08 frames. ], batch size: 48, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:39:14,756 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.219e+02 2.672e+02 3.069e+02 6.824e+02, threshold=5.344e+02, percent-clipped=3.0 2023-05-01 13:39:35,577 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:54,168 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:55,234 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:01,693 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 350, loss[loss=0.2038, simple_loss=0.2733, pruned_loss=0.06712, over 16698.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2556, pruned_loss=0.04184, over 2745437.56 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:40:38,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0936, 5.1563, 5.5783, 5.5488, 5.5535, 5.2155, 5.1362, 4.9986], device='cuda:3'), covar=tensor([0.0339, 0.0551, 0.0366, 0.0419, 0.0472, 0.0386, 0.0957, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0451, 0.0439, 0.0405, 0.0482, 0.0460, 0.0540, 0.0368], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 13:40:42,277 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:03,084 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:15,701 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8098, 4.0065, 2.6235, 4.5914, 3.1885, 4.5279, 2.6982, 3.3018], device='cuda:3'), covar=tensor([0.0378, 0.0429, 0.1627, 0.0438, 0.0882, 0.0583, 0.1493, 0.0791], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0177, 0.0193, 0.0163, 0.0177, 0.0215, 0.0203, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:41:22,666 INFO [train.py:904] (3/8) Epoch 23, batch 400, loss[loss=0.1684, simple_loss=0.2427, pruned_loss=0.04705, over 16844.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2547, pruned_loss=0.04174, over 2871723.27 frames. ], batch size: 83, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:41:24,787 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-01 13:41:34,906 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.103e+02 2.400e+02 2.968e+02 1.328e+03, threshold=4.800e+02, percent-clipped=3.0 2023-05-01 13:41:40,090 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7923, 4.4458, 4.4396, 4.9254, 5.1917, 4.6077, 5.1882, 5.1264], device='cuda:3'), covar=tensor([0.2030, 0.1769, 0.3271, 0.1452, 0.0906, 0.1338, 0.0921, 0.1226], device='cuda:3'), in_proj_covar=tensor([0.0634, 0.0779, 0.0903, 0.0789, 0.0598, 0.0625, 0.0653, 0.0754], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:41:46,587 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:42:32,062 INFO [train.py:904] (3/8) Epoch 23, batch 450, loss[loss=0.1726, simple_loss=0.265, pruned_loss=0.04012, over 17082.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2528, pruned_loss=0.0412, over 2969261.91 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:42:52,142 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:43:30,237 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 13:43:40,975 INFO [train.py:904] (3/8) Epoch 23, batch 500, loss[loss=0.154, simple_loss=0.2388, pruned_loss=0.03465, over 17208.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2515, pruned_loss=0.04057, over 3052809.95 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:43:50,503 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-01 13:43:52,982 INFO [optim.py:368] (3/8) 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 13:44:16,834 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223829.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:44:49,713 INFO [train.py:904] (3/8) Epoch 23, batch 550, loss[loss=0.1654, simple_loss=0.2592, pruned_loss=0.03582, over 17131.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2508, pruned_loss=0.04029, over 3110056.12 frames. ], batch size: 48, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:45:22,792 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7608, 2.8778, 2.7574, 4.9746, 4.1265, 4.3841, 1.6620, 3.1486], device='cuda:3'), covar=tensor([0.1416, 0.0803, 0.1194, 0.0255, 0.0201, 0.0421, 0.1662, 0.0795], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0188, 0.0202, 0.0215, 0.0203, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 13:45:24,265 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223877.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:45:59,107 INFO [train.py:904] (3/8) Epoch 23, batch 600, loss[loss=0.1598, simple_loss=0.2354, pruned_loss=0.04208, over 16824.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2503, pruned_loss=0.04083, over 3155019.77 frames. ], batch size: 102, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:46:10,991 INFO [optim.py:368] (3/8) 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 13:46:15,777 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2807, 2.2145, 2.3820, 4.0055, 2.3337, 2.5768, 2.3273, 2.4402], device='cuda:3'), covar=tensor([0.1412, 0.3865, 0.3059, 0.0608, 0.3901, 0.2622, 0.3907, 0.3041], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0451, 0.0371, 0.0328, 0.0439, 0.0518, 0.0422, 0.0527], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:46:23,810 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-05-01 13:46:28,911 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-01 13:46:41,947 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 650, loss[loss=0.157, simple_loss=0.2417, pruned_loss=0.03616, over 17024.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2485, pruned_loss=0.04006, over 3191973.28 frames. ], batch size: 41, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:22,013 INFO [train.py:904] (3/8) Epoch 23, batch 700, loss[loss=0.1628, simple_loss=0.2396, pruned_loss=0.04303, over 16050.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2486, pruned_loss=0.03986, over 3229990.10 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:35,477 INFO [optim.py:368] (3/8) 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 13:49:33,339 INFO [train.py:904] (3/8) Epoch 23, batch 750, loss[loss=0.1563, simple_loss=0.236, pruned_loss=0.03832, over 16844.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2497, pruned_loss=0.04025, over 3246247.02 frames. ], batch size: 102, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:50:41,982 INFO [train.py:904] (3/8) Epoch 23, batch 800, loss[loss=0.1787, simple_loss=0.2533, pruned_loss=0.05205, over 16683.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.249, pruned_loss=0.03983, over 3269229.50 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:50:47,325 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-05-01 13:50:54,821 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 2.213e+02 2.619e+02 3.114e+02 4.585e+02, threshold=5.238e+02, percent-clipped=0.0 2023-05-01 13:51:51,762 INFO [train.py:904] (3/8) Epoch 23, batch 850, loss[loss=0.1707, simple_loss=0.2483, pruned_loss=0.04659, over 16606.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2486, pruned_loss=0.03966, over 3269121.88 frames. ], batch size: 68, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:00,735 INFO [train.py:904] (3/8) Epoch 23, batch 900, loss[loss=0.1548, simple_loss=0.2365, pruned_loss=0.0365, over 16917.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.248, pruned_loss=0.03887, over 3273971.66 frames. ], batch size: 96, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:14,895 INFO [optim.py:368] (3/8) 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 13:53:15,211 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224212.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:53:46,029 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 950, loss[loss=0.1507, simple_loss=0.239, pruned_loss=0.03117, over 17218.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.248, pruned_loss=0.03913, over 3283669.99 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:54:36,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3991, 4.2256, 4.4547, 4.5882, 4.7387, 4.2916, 4.6222, 4.6980], device='cuda:3'), covar=tensor([0.2015, 0.1569, 0.1534, 0.0907, 0.0717, 0.1214, 0.2064, 0.1182], device='cuda:3'), in_proj_covar=tensor([0.0656, 0.0807, 0.0935, 0.0819, 0.0619, 0.0645, 0.0674, 0.0782], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 13:54:38,516 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224273.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:47,564 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 13:54:49,983 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 13:54:50,479 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:55:20,104 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 13:55:20,379 INFO [train.py:904] (3/8) Epoch 23, batch 1000, loss[loss=0.1636, simple_loss=0.2574, pruned_loss=0.03494, over 17115.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2469, pruned_loss=0.03846, over 3298949.26 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:55:33,530 INFO [optim.py:368] (3/8) 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 13:56:26,361 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 13:56:31,386 INFO [train.py:904] (3/8) Epoch 23, batch 1050, loss[loss=0.1478, simple_loss=0.2332, pruned_loss=0.03115, over 17201.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2469, pruned_loss=0.03829, over 3304807.40 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:42,152 INFO [train.py:904] (3/8) Epoch 23, batch 1100, loss[loss=0.1752, simple_loss=0.2474, pruned_loss=0.05154, over 16754.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2466, pruned_loss=0.03835, over 3312296.58 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:54,066 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 2.008e+02 2.334e+02 2.683e+02 4.688e+02, threshold=4.667e+02, percent-clipped=0.0 2023-05-01 13:58:51,564 INFO [train.py:904] (3/8) Epoch 23, batch 1150, loss[loss=0.1584, simple_loss=0.2401, pruned_loss=0.03828, over 16512.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2462, pruned_loss=0.038, over 3319673.79 frames. ], batch size: 146, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:59:32,701 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-01 14:00:00,970 INFO [train.py:904] (3/8) Epoch 23, batch 1200, loss[loss=0.1521, simple_loss=0.2544, pruned_loss=0.02484, over 17118.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2456, pruned_loss=0.03804, over 3302530.90 frames. ], batch size: 47, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:00:14,523 INFO [optim.py:368] (3/8) 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 14:01:10,513 INFO [train.py:904] (3/8) Epoch 23, batch 1250, loss[loss=0.1747, simple_loss=0.2482, pruned_loss=0.05058, over 16692.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2462, pruned_loss=0.0385, over 3316382.70 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:01:12,197 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4185, 2.2025, 2.2044, 4.2000, 2.1769, 2.5335, 2.3460, 2.3335], device='cuda:3'), covar=tensor([0.1285, 0.4195, 0.3442, 0.0527, 0.4628, 0.2936, 0.3968, 0.4051], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0456, 0.0375, 0.0332, 0.0442, 0.0525, 0.0427, 0.0533], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:01:22,973 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224562.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:24,274 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 14:01:31,867 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:02:16,010 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 14:02:20,417 INFO [train.py:904] (3/8) Epoch 23, batch 1300, loss[loss=0.1312, simple_loss=0.216, pruned_loss=0.0232, over 15795.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2469, pruned_loss=0.03861, over 3315015.29 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:02:33,658 INFO [optim.py:368] (3/8) 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 14:02:49,066 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224623.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:03:04,973 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 14:03:29,376 INFO [train.py:904] (3/8) Epoch 23, batch 1350, loss[loss=0.188, simple_loss=0.2577, pruned_loss=0.05909, over 16676.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2463, pruned_loss=0.03832, over 3308103.35 frames. ], batch size: 89, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:10,782 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9333, 2.5549, 1.9787, 2.4045, 2.9667, 2.7521, 2.9685, 3.0573], device='cuda:3'), covar=tensor([0.0256, 0.0395, 0.0589, 0.0430, 0.0230, 0.0342, 0.0236, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0242, 0.0233, 0.0233, 0.0243, 0.0242, 0.0243, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:04:40,082 INFO [train.py:904] (3/8) Epoch 23, batch 1400, loss[loss=0.175, simple_loss=0.2519, pruned_loss=0.04902, over 16754.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2472, pruned_loss=0.039, over 3317129.85 frames. ], batch size: 102, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:52,698 INFO [optim.py:368] (3/8) 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 14:05:13,934 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:33,047 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:50,557 INFO [train.py:904] (3/8) Epoch 23, batch 1450, loss[loss=0.1469, simple_loss=0.2274, pruned_loss=0.0332, over 15894.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2465, pruned_loss=0.03858, over 3319028.24 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:06:39,711 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9500, 2.1155, 2.5480, 2.9032, 2.7996, 3.4682, 2.3447, 3.3939], device='cuda:3'), covar=tensor([0.0274, 0.0525, 0.0372, 0.0374, 0.0366, 0.0203, 0.0487, 0.0197], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0193, 0.0181, 0.0186, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:06:39,720 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:06:59,988 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 1500, loss[loss=0.1493, simple_loss=0.2362, pruned_loss=0.03121, over 17220.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2463, pruned_loss=0.03853, over 3320195.75 frames. ], batch size: 43, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:07:02,297 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9274, 4.0490, 2.6740, 4.6722, 3.2712, 4.6488, 2.8097, 3.3914], device='cuda:3'), covar=tensor([0.0344, 0.0441, 0.1532, 0.0318, 0.0795, 0.0497, 0.1408, 0.0717], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0181, 0.0197, 0.0170, 0.0179, 0.0221, 0.0207, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:07:12,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3800, 5.3498, 5.1636, 4.5902, 5.1475, 2.1970, 4.9450, 5.1123], device='cuda:3'), covar=tensor([0.0077, 0.0071, 0.0197, 0.0410, 0.0105, 0.2452, 0.0140, 0.0194], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0163, 0.0205, 0.0181, 0.0183, 0.0215, 0.0195, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:07:16,528 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.087e+02 2.500e+02 3.086e+02 8.479e+02, threshold=5.000e+02, percent-clipped=3.0 2023-05-01 14:07:38,563 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9252, 2.0463, 2.4357, 2.7947, 2.8302, 2.8469, 2.0746, 3.0539], device='cuda:3'), covar=tensor([0.0194, 0.0466, 0.0357, 0.0282, 0.0315, 0.0307, 0.0520, 0.0163], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0193, 0.0181, 0.0185, 0.0198, 0.0155, 0.0197, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:08:14,249 INFO [train.py:904] (3/8) Epoch 23, batch 1550, loss[loss=0.1936, simple_loss=0.2582, pruned_loss=0.06449, over 16898.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2474, pruned_loss=0.03924, over 3315349.89 frames. ], batch size: 109, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 14:08:34,607 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224868.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:08:53,943 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9588, 4.7082, 4.9946, 5.1643, 5.4100, 4.7089, 5.3701, 5.3762], device='cuda:3'), covar=tensor([0.2037, 0.1613, 0.1904, 0.0900, 0.0545, 0.0951, 0.0583, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0673, 0.0830, 0.0962, 0.0839, 0.0633, 0.0664, 0.0690, 0.0801], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:09:23,916 INFO [train.py:904] (3/8) Epoch 23, batch 1600, loss[loss=0.1657, simple_loss=0.2466, pruned_loss=0.04239, over 16671.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2486, pruned_loss=0.03969, over 3298907.50 frames. ], batch size: 134, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:09:36,822 INFO [optim.py:368] (3/8) 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 14:09:41,252 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:54,949 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 14:10:32,920 INFO [train.py:904] (3/8) Epoch 23, batch 1650, loss[loss=0.1738, simple_loss=0.2706, pruned_loss=0.03855, over 17054.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2503, pruned_loss=0.0401, over 3309165.88 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:41,679 INFO [train.py:904] (3/8) Epoch 23, batch 1700, loss[loss=0.1649, simple_loss=0.2576, pruned_loss=0.03608, over 16556.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2513, pruned_loss=0.0402, over 3316920.95 frames. ], batch size: 75, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:56,160 INFO [optim.py:368] (3/8) 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 14:12:13,278 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7880, 2.5692, 2.0671, 2.3858, 2.9230, 2.7184, 2.9356, 2.9859], device='cuda:3'), covar=tensor([0.0194, 0.0354, 0.0542, 0.0416, 0.0218, 0.0291, 0.0229, 0.0257], device='cuda:3'), in_proj_covar=tensor([0.0220, 0.0242, 0.0232, 0.0232, 0.0243, 0.0242, 0.0244, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:12:52,528 INFO [train.py:904] (3/8) Epoch 23, batch 1750, loss[loss=0.1781, simple_loss=0.268, pruned_loss=0.04411, over 16668.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2516, pruned_loss=0.04036, over 3317877.90 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:13:33,033 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:13:34,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4825, 4.4468, 4.4064, 4.0658, 4.1354, 4.4672, 4.2389, 4.1913], device='cuda:3'), covar=tensor([0.0687, 0.0792, 0.0363, 0.0349, 0.0876, 0.0529, 0.0603, 0.0674], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0448, 0.0357, 0.0354, 0.0362, 0.0414, 0.0242, 0.0426], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:13:52,600 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 1800, loss[loss=0.1867, simple_loss=0.2739, pruned_loss=0.04972, over 15323.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2532, pruned_loss=0.0401, over 3312564.94 frames. ], batch size: 190, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:14:02,301 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.184e+02 2.489e+02 3.045e+02 6.303e+02, threshold=4.978e+02, percent-clipped=2.0 2023-05-01 14:14:35,780 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9033, 4.6921, 5.0073, 5.1646, 5.3997, 4.7145, 5.3528, 5.4030], device='cuda:3'), covar=tensor([0.2095, 0.1542, 0.1954, 0.0899, 0.0623, 0.1071, 0.0611, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0674, 0.0832, 0.0964, 0.0842, 0.0636, 0.0668, 0.0692, 0.0803], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:14:57,257 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:12,125 INFO [train.py:904] (3/8) Epoch 23, batch 1850, loss[loss=0.1748, simple_loss=0.271, pruned_loss=0.03931, over 16756.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2547, pruned_loss=0.04083, over 3315551.70 frames. ], batch size: 57, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:15:27,755 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:16:22,244 INFO [train.py:904] (3/8) Epoch 23, batch 1900, loss[loss=0.1935, simple_loss=0.2692, pruned_loss=0.05889, over 16478.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2542, pruned_loss=0.04045, over 3308629.93 frames. ], batch size: 146, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:16:22,791 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:16:43,951 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:17:11,957 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 14:17:31,819 INFO [train.py:904] (3/8) Epoch 23, batch 1950, loss[loss=0.1759, simple_loss=0.2688, pruned_loss=0.0415, over 15706.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2545, pruned_loss=0.04006, over 3303237.81 frames. ], batch size: 191, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:17:49,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7063, 3.7577, 2.8673, 2.2608, 2.4125, 2.3124, 3.8406, 3.2869], device='cuda:3'), covar=tensor([0.2642, 0.0590, 0.1685, 0.3123, 0.2615, 0.2154, 0.0483, 0.1474], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0271, 0.0309, 0.0317, 0.0300, 0.0265, 0.0300, 0.0344], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 14:17:50,055 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:18:24,623 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8476, 4.6219, 4.9260, 5.0726, 5.3013, 4.6534, 5.2744, 5.2778], device='cuda:3'), covar=tensor([0.2262, 0.1476, 0.1994, 0.0926, 0.0636, 0.1088, 0.0630, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0677, 0.0838, 0.0971, 0.0844, 0.0639, 0.0672, 0.0694, 0.0805], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:18:37,323 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6602, 2.6030, 2.3409, 2.6283, 2.9312, 2.7592, 3.2611, 3.1963], device='cuda:3'), covar=tensor([0.0165, 0.0482, 0.0573, 0.0447, 0.0343, 0.0410, 0.0298, 0.0297], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0243, 0.0242, 0.0243, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:18:42,383 INFO [train.py:904] (3/8) Epoch 23, batch 2000, loss[loss=0.1496, simple_loss=0.2451, pruned_loss=0.02707, over 17100.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2545, pruned_loss=0.04013, over 3303158.38 frames. ], batch size: 47, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:18:56,051 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.089e+02 2.408e+02 2.851e+02 8.100e+02, threshold=4.816e+02, percent-clipped=1.0 2023-05-01 14:19:52,847 INFO [train.py:904] (3/8) Epoch 23, batch 2050, loss[loss=0.1996, simple_loss=0.2767, pruned_loss=0.06131, over 16736.00 frames. ], tot_loss[loss=0.168, simple_loss=0.255, pruned_loss=0.04049, over 3294977.64 frames. ], batch size: 124, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:20:02,801 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-01 14:20:35,690 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:20:54,907 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225396.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:04,272 INFO [train.py:904] (3/8) Epoch 23, batch 2100, loss[loss=0.1627, simple_loss=0.2511, pruned_loss=0.03711, over 16516.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2556, pruned_loss=0.04097, over 3301893.46 frames. ], batch size: 75, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:21:09,863 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2562, 2.1935, 2.8764, 3.1761, 3.1952, 3.6505, 2.3331, 3.7368], device='cuda:3'), covar=tensor([0.0231, 0.0614, 0.0309, 0.0307, 0.0277, 0.0208, 0.0642, 0.0139], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0200, 0.0158, 0.0199, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:21:18,934 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.181e+02 2.818e+02 3.430e+02 7.396e+02, threshold=5.635e+02, percent-clipped=7.0 2023-05-01 14:21:29,387 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 14:21:44,079 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:57,162 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:03,499 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225444.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:15,763 INFO [train.py:904] (3/8) Epoch 23, batch 2150, loss[loss=0.1625, simple_loss=0.2533, pruned_loss=0.03584, over 16563.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2564, pruned_loss=0.04141, over 3297052.33 frames. ], batch size: 68, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:22:25,262 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:25,320 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3467, 4.2992, 4.2705, 3.9913, 4.0552, 4.3121, 4.0375, 4.1049], device='cuda:3'), covar=tensor([0.0599, 0.0745, 0.0295, 0.0307, 0.0675, 0.0515, 0.0688, 0.0568], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0456, 0.0363, 0.0361, 0.0367, 0.0421, 0.0246, 0.0433], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 14:22:29,383 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:57,590 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3704, 3.7163, 4.0174, 2.2189, 3.1516, 2.6496, 3.8670, 3.8868], device='cuda:3'), covar=tensor([0.0326, 0.0886, 0.0509, 0.2061, 0.0857, 0.0940, 0.0640, 0.1043], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:23:18,689 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225498.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:22,027 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:26,043 INFO [train.py:904] (3/8) Epoch 23, batch 2200, loss[loss=0.148, simple_loss=0.2384, pruned_loss=0.0288, over 17165.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2576, pruned_loss=0.04226, over 3302353.07 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:23:29,392 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:23:53,575 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225522.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:55,302 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:24:36,374 INFO [train.py:904] (3/8) Epoch 23, batch 2250, loss[loss=0.1469, simple_loss=0.2354, pruned_loss=0.02918, over 17179.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2584, pruned_loss=0.04211, over 3314223.60 frames. ], batch size: 44, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:24:39,295 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9510, 2.0637, 2.4587, 2.7576, 2.8078, 3.0532, 2.1428, 3.1341], device='cuda:3'), covar=tensor([0.0245, 0.0475, 0.0356, 0.0339, 0.0340, 0.0246, 0.0562, 0.0164], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0188, 0.0200, 0.0157, 0.0199, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:24:45,136 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225559.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:24:54,952 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:18,156 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:46,973 INFO [train.py:904] (3/8) Epoch 23, batch 2300, loss[loss=0.1615, simple_loss=0.2516, pruned_loss=0.03568, over 16695.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2578, pruned_loss=0.04181, over 3315381.73 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:25:56,930 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 14:26:01,554 INFO [optim.py:368] (3/8) 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 14:26:11,645 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225620.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:26:58,891 INFO [train.py:904] (3/8) Epoch 23, batch 2350, loss[loss=0.1628, simple_loss=0.2458, pruned_loss=0.03989, over 16413.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2571, pruned_loss=0.04181, over 3318401.61 frames. ], batch size: 146, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:27:34,810 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4146, 2.2498, 2.3655, 4.1362, 2.1582, 2.4948, 2.3555, 2.4124], device='cuda:3'), covar=tensor([0.1434, 0.4065, 0.3284, 0.0625, 0.4803, 0.3147, 0.4027, 0.4341], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0459, 0.0376, 0.0335, 0.0443, 0.0530, 0.0431, 0.0537], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:28:10,319 INFO [train.py:904] (3/8) Epoch 23, batch 2400, loss[loss=0.1692, simple_loss=0.2681, pruned_loss=0.03511, over 16527.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2583, pruned_loss=0.04218, over 3322029.67 frames. ], batch size: 68, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:28:23,215 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.227e+02 2.621e+02 3.138e+02 6.803e+02, threshold=5.242e+02, percent-clipped=3.0 2023-05-01 14:29:17,841 INFO [train.py:904] (3/8) Epoch 23, batch 2450, loss[loss=0.1441, simple_loss=0.2309, pruned_loss=0.0286, over 16828.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2587, pruned_loss=0.042, over 3317397.45 frames. ], batch size: 39, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:29:26,303 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:10,488 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-05-01 14:30:17,192 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:22,197 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 2500, loss[loss=0.1569, simple_loss=0.2529, pruned_loss=0.03042, over 17171.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2582, pruned_loss=0.04182, over 3318531.90 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:30:36,108 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.105e+02 2.544e+02 3.075e+02 7.140e+02, threshold=5.088e+02, percent-clipped=2.0 2023-05-01 14:30:51,423 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225818.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:28,974 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:38,358 INFO [train.py:904] (3/8) Epoch 23, batch 2550, loss[loss=0.1522, simple_loss=0.2481, pruned_loss=0.02811, over 17157.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.259, pruned_loss=0.04203, over 3314715.74 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:31:38,906 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7249, 1.9313, 2.2994, 2.5193, 2.6761, 2.6189, 1.9643, 2.8372], device='cuda:3'), covar=tensor([0.0170, 0.0476, 0.0321, 0.0272, 0.0287, 0.0308, 0.0499, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0188, 0.0201, 0.0158, 0.0198, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:31:49,565 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:50,933 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225862.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:14,110 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225878.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:38,023 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 2600, loss[loss=0.1772, simple_loss=0.271, pruned_loss=0.0417, over 17094.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04154, over 3326243.38 frames. ], batch size: 55, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:32:51,084 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 14:33:03,045 INFO [optim.py:368] (3/8) 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 14:33:06,084 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225915.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 14:33:07,223 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9438, 4.9994, 5.3795, 5.4030, 5.4418, 5.0859, 5.0364, 4.7624], device='cuda:3'), covar=tensor([0.0372, 0.0628, 0.0455, 0.0437, 0.0524, 0.0412, 0.1034, 0.0526], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0476, 0.0462, 0.0427, 0.0509, 0.0485, 0.0568, 0.0387], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 14:33:16,611 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:33:58,794 INFO [train.py:904] (3/8) Epoch 23, batch 2650, loss[loss=0.1602, simple_loss=0.26, pruned_loss=0.03019, over 17123.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2595, pruned_loss=0.04146, over 3327569.90 frames. ], batch size: 47, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:34:04,243 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:34:51,933 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-05-01 14:35:12,270 INFO [train.py:904] (3/8) Epoch 23, batch 2700, loss[loss=0.1391, simple_loss=0.2337, pruned_loss=0.02224, over 17155.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2588, pruned_loss=0.04075, over 3330639.58 frames. ], batch size: 47, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:35:24,996 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0663, 4.8034, 5.0930, 5.2643, 5.4845, 4.7870, 5.4658, 5.4610], device='cuda:3'), covar=tensor([0.1932, 0.1453, 0.1838, 0.0826, 0.0570, 0.0900, 0.0581, 0.0626], device='cuda:3'), in_proj_covar=tensor([0.0672, 0.0835, 0.0965, 0.0842, 0.0640, 0.0668, 0.0692, 0.0803], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:35:25,728 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.491e+02 2.127e+02 2.428e+02 2.889e+02 7.591e+02, threshold=4.856e+02, percent-clipped=2.0 2023-05-01 14:36:06,276 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:36:13,645 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2950, 5.2610, 4.9709, 4.4499, 5.1150, 1.9869, 4.8441, 4.9550], device='cuda:3'), covar=tensor([0.0091, 0.0079, 0.0223, 0.0408, 0.0118, 0.2781, 0.0143, 0.0211], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0165, 0.0207, 0.0184, 0.0185, 0.0216, 0.0197, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:36:23,069 INFO [train.py:904] (3/8) Epoch 23, batch 2750, loss[loss=0.1531, simple_loss=0.2492, pruned_loss=0.02849, over 16819.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2592, pruned_loss=0.04052, over 3320015.30 frames. ], batch size: 42, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:21,842 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:25,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7200, 2.4885, 2.0182, 2.2870, 2.8156, 2.6529, 3.2655, 3.1945], device='cuda:3'), covar=tensor([0.0188, 0.0613, 0.0808, 0.0663, 0.0425, 0.0534, 0.0341, 0.0381], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0243, 0.0231, 0.0233, 0.0244, 0.0242, 0.0244, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:37:31,987 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226102.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:32,650 INFO [train.py:904] (3/8) Epoch 23, batch 2800, loss[loss=0.1587, simple_loss=0.253, pruned_loss=0.03218, over 17019.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2584, pruned_loss=0.04023, over 3331153.13 frames. ], batch size: 50, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:47,345 INFO [optim.py:368] (3/8) 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 14:37:54,707 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:59,228 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1177, 2.1961, 2.7275, 2.9573, 2.9709, 3.5340, 2.3176, 3.5553], device='cuda:3'), covar=tensor([0.0291, 0.0563, 0.0346, 0.0395, 0.0367, 0.0223, 0.0598, 0.0198], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0190, 0.0202, 0.0160, 0.0201, 0.0156], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:38:29,330 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:38:42,389 INFO [train.py:904] (3/8) Epoch 23, batch 2850, loss[loss=0.1697, simple_loss=0.2698, pruned_loss=0.03481, over 17113.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2576, pruned_loss=0.03991, over 3325639.49 frames. ], batch size: 49, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:38:54,384 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:00,759 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:18,075 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:51,833 INFO [train.py:904] (3/8) Epoch 23, batch 2900, loss[loss=0.1614, simple_loss=0.247, pruned_loss=0.03785, over 16454.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2569, pruned_loss=0.04023, over 3332939.34 frames. ], batch size: 68, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:39:59,523 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 14:40:00,216 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226209.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:05,815 INFO [optim.py:368] (3/8) 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 14:40:09,305 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226215.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 14:40:13,513 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:34,174 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:58,436 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 2950, loss[loss=0.1611, simple_loss=0.2456, pruned_loss=0.03831, over 17008.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2571, pruned_loss=0.04128, over 3332761.94 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:41:14,089 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226263.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:41:56,413 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 3000, loss[loss=0.1801, simple_loss=0.2624, pruned_loss=0.04887, over 16361.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2564, pruned_loss=0.04123, over 3332176.55 frames. ], batch size: 146, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:42:08,778 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 14:42:17,866 INFO [train.py:938] (3/8) Epoch 23, validation: loss=0.1344, simple_loss=0.2397, pruned_loss=0.01456, over 944034.00 frames. 2023-05-01 14:42:17,867 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 14:42:30,993 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.313e+02 2.743e+02 3.282e+02 6.136e+02, threshold=5.486e+02, percent-clipped=4.0 2023-05-01 14:42:47,608 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9336, 4.8683, 4.7327, 3.8148, 4.8496, 1.8420, 4.5729, 4.5344], device='cuda:3'), covar=tensor([0.0154, 0.0140, 0.0257, 0.0594, 0.0149, 0.3060, 0.0185, 0.0292], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0164, 0.0205, 0.0183, 0.0184, 0.0213, 0.0195, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:42:54,769 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 14:43:05,597 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 14:43:16,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5894, 3.5339, 3.5014, 2.8733, 3.3389, 2.0935, 3.2426, 2.7735], device='cuda:3'), covar=tensor([0.0147, 0.0139, 0.0198, 0.0256, 0.0113, 0.2348, 0.0129, 0.0243], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0164, 0.0206, 0.0183, 0.0184, 0.0213, 0.0196, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:43:20,739 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3751, 4.2229, 4.4400, 4.5815, 4.6801, 4.2643, 4.5255, 4.6654], device='cuda:3'), covar=tensor([0.1723, 0.1201, 0.1373, 0.0691, 0.0672, 0.1210, 0.1999, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0676, 0.0841, 0.0971, 0.0847, 0.0644, 0.0674, 0.0696, 0.0809], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:43:26,885 INFO [train.py:904] (3/8) Epoch 23, batch 3050, loss[loss=0.1349, simple_loss=0.2236, pruned_loss=0.02305, over 17233.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2566, pruned_loss=0.04145, over 3319272.87 frames. ], batch size: 45, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:43:29,210 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8917, 1.4116, 1.7143, 1.7555, 1.8929, 2.0174, 1.6053, 1.8985], device='cuda:3'), covar=tensor([0.0267, 0.0419, 0.0243, 0.0308, 0.0289, 0.0221, 0.0453, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0198, 0.0186, 0.0191, 0.0204, 0.0161, 0.0201, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:44:29,373 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 3100, loss[loss=0.1538, simple_loss=0.2316, pruned_loss=0.03805, over 16833.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2559, pruned_loss=0.04118, over 3323984.17 frames. ], batch size: 102, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:44:51,603 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 23, batch 3150, loss[loss=0.1642, simple_loss=0.2563, pruned_loss=0.03601, over 17065.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2548, pruned_loss=0.04065, over 3322455.75 frames. ], batch size: 53, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:46:12,413 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6706, 3.7992, 2.0493, 4.2758, 2.9076, 4.2240, 2.1094, 2.9175], device='cuda:3'), covar=tensor([0.0300, 0.0359, 0.1847, 0.0318, 0.0796, 0.0439, 0.1953, 0.0771], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0171, 0.0180, 0.0223, 0.0207, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:46:36,290 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2646, 4.1532, 4.3602, 4.4953, 4.5694, 4.1438, 4.4014, 4.5707], device='cuda:3'), covar=tensor([0.1763, 0.1207, 0.1263, 0.0620, 0.0592, 0.1284, 0.2524, 0.0633], device='cuda:3'), in_proj_covar=tensor([0.0681, 0.0846, 0.0978, 0.0851, 0.0647, 0.0678, 0.0700, 0.0814], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:46:39,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0075, 4.2512, 4.3355, 3.0445, 3.6442, 4.3198, 3.8470, 2.6451], device='cuda:3'), covar=tensor([0.0422, 0.0083, 0.0048, 0.0379, 0.0131, 0.0096, 0.0094, 0.0444], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0086, 0.0087, 0.0136, 0.0100, 0.0113, 0.0098, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 14:46:40,732 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-05-01 14:46:54,851 INFO [train.py:904] (3/8) Epoch 23, batch 3200, loss[loss=0.1799, simple_loss=0.2639, pruned_loss=0.04794, over 12184.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2546, pruned_loss=0.04025, over 3323494.99 frames. ], batch size: 246, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:47:09,850 INFO [optim.py:368] (3/8) 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 14:47:15,796 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:47:46,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6945, 2.4633, 2.5345, 4.6063, 2.4755, 2.8429, 2.5450, 2.7044], device='cuda:3'), covar=tensor([0.1273, 0.3671, 0.3123, 0.0480, 0.4347, 0.2758, 0.3635, 0.3612], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0457, 0.0375, 0.0336, 0.0441, 0.0527, 0.0428, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:48:01,529 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 3250, loss[loss=0.1773, simple_loss=0.2657, pruned_loss=0.04446, over 17032.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2553, pruned_loss=0.04117, over 3313765.96 frames. ], batch size: 53, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:48:22,283 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:54,452 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226589.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:08,847 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 3300, loss[loss=0.1919, simple_loss=0.2666, pruned_loss=0.05857, over 16771.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2562, pruned_loss=0.04103, over 3320738.06 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:49:27,925 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.105e+02 2.619e+02 3.003e+02 4.974e+02, threshold=5.238e+02, percent-clipped=1.0 2023-05-01 14:50:23,120 INFO [train.py:904] (3/8) Epoch 23, batch 3350, loss[loss=0.1496, simple_loss=0.2454, pruned_loss=0.02695, over 17231.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2562, pruned_loss=0.04041, over 3332525.36 frames. ], batch size: 44, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:23,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8443, 2.8294, 2.6568, 4.3642, 3.4912, 4.1305, 1.7774, 3.0182], device='cuda:3'), covar=tensor([0.1325, 0.0700, 0.1091, 0.0178, 0.0187, 0.0360, 0.1479, 0.0775], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0195, 0.0206, 0.0218, 0.0204, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:51:25,719 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:51:34,605 INFO [train.py:904] (3/8) Epoch 23, batch 3400, loss[loss=0.1807, simple_loss=0.2766, pruned_loss=0.04241, over 16789.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2565, pruned_loss=0.04051, over 3336625.18 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:35,034 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5326, 5.4378, 5.2714, 4.5418, 5.3300, 2.2648, 5.0954, 5.2354], device='cuda:3'), covar=tensor([0.0087, 0.0087, 0.0210, 0.0458, 0.0100, 0.2572, 0.0137, 0.0182], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0166, 0.0207, 0.0184, 0.0185, 0.0214, 0.0197, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:51:47,766 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:52:43,450 INFO [train.py:904] (3/8) Epoch 23, batch 3450, loss[loss=0.135, simple_loss=0.2241, pruned_loss=0.02293, over 16800.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2548, pruned_loss=0.03964, over 3340635.84 frames. ], batch size: 42, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:53:52,948 INFO [train.py:904] (3/8) Epoch 23, batch 3500, loss[loss=0.189, simple_loss=0.2782, pruned_loss=0.04986, over 17091.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2533, pruned_loss=0.03928, over 3328554.76 frames. ], batch size: 55, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:54:07,217 INFO [optim.py:368] (3/8) 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 14:54:19,270 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2849, 2.2800, 2.4016, 4.0375, 2.2648, 2.6638, 2.3654, 2.4731], device='cuda:3'), covar=tensor([0.1482, 0.3632, 0.2950, 0.0607, 0.4081, 0.2693, 0.3886, 0.3149], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0456, 0.0375, 0.0335, 0.0441, 0.0526, 0.0428, 0.0534], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:54:34,870 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6888, 3.7502, 2.0887, 3.9672, 2.8628, 3.9220, 2.1156, 2.9291], device='cuda:3'), covar=tensor([0.0234, 0.0319, 0.1767, 0.0370, 0.0762, 0.0586, 0.1882, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0182, 0.0197, 0.0172, 0.0180, 0.0225, 0.0207, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:55:02,057 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6828, 2.6796, 2.3775, 2.5210, 3.0185, 2.7505, 3.3028, 3.2177], device='cuda:3'), covar=tensor([0.0177, 0.0481, 0.0555, 0.0520, 0.0335, 0.0434, 0.0281, 0.0304], device='cuda:3'), in_proj_covar=tensor([0.0226, 0.0247, 0.0233, 0.0236, 0.0246, 0.0245, 0.0247, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:55:03,886 INFO [train.py:904] (3/8) Epoch 23, batch 3550, loss[loss=0.1579, simple_loss=0.2541, pruned_loss=0.03089, over 17162.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2521, pruned_loss=0.03881, over 3337588.44 frames. ], batch size: 46, lr: 2.94e-03, grad_scale: 16.0 2023-05-01 14:55:34,931 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 14:55:52,844 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 3600, loss[loss=0.1563, simple_loss=0.2553, pruned_loss=0.02862, over 16712.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.251, pruned_loss=0.03882, over 3329538.40 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:56:16,969 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2877, 2.3223, 2.3993, 4.0611, 2.3373, 2.6356, 2.3916, 2.4803], device='cuda:3'), covar=tensor([0.1527, 0.3803, 0.2979, 0.0644, 0.3899, 0.2744, 0.3850, 0.3226], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0457, 0.0374, 0.0335, 0.0441, 0.0527, 0.0428, 0.0534], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:56:22,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3730, 3.6714, 4.0940, 2.2750, 3.2863, 2.6182, 3.8283, 3.8145], device='cuda:3'), covar=tensor([0.0306, 0.0891, 0.0463, 0.1936, 0.0798, 0.0930, 0.0622, 0.1049], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0149, 0.0133, 0.0146, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-01 14:56:28,214 INFO [optim.py:368] (3/8) 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 14:56:30,102 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9826, 1.9817, 2.6062, 2.8784, 2.8018, 3.4349, 2.3960, 3.4132], device='cuda:3'), covar=tensor([0.0277, 0.0602, 0.0353, 0.0351, 0.0372, 0.0220, 0.0508, 0.0200], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0197, 0.0186, 0.0191, 0.0204, 0.0161, 0.0201, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:56:39,306 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0506, 1.9131, 2.6393, 3.0239, 2.8378, 3.4553, 2.0302, 3.5129], device='cuda:3'), covar=tensor([0.0220, 0.0629, 0.0345, 0.0287, 0.0322, 0.0191, 0.0705, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0197, 0.0185, 0.0191, 0.0204, 0.0161, 0.0201, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:57:02,339 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226937.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:57:02,412 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7926, 5.0951, 4.8923, 4.8579, 4.6693, 4.6315, 4.5494, 5.1818], device='cuda:3'), covar=tensor([0.1213, 0.0854, 0.1041, 0.0919, 0.0833, 0.1093, 0.1236, 0.0891], device='cuda:3'), in_proj_covar=tensor([0.0703, 0.0857, 0.0705, 0.0650, 0.0540, 0.0546, 0.0719, 0.0670], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:57:24,672 INFO [train.py:904] (3/8) Epoch 23, batch 3650, loss[loss=0.1673, simple_loss=0.2423, pruned_loss=0.04613, over 16482.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.25, pruned_loss=0.03974, over 3315466.50 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:57:44,269 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 14:57:45,603 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5595, 3.6417, 2.2224, 3.8982, 2.8694, 3.8329, 2.3138, 2.9014], device='cuda:3'), covar=tensor([0.0268, 0.0430, 0.1572, 0.0344, 0.0797, 0.0909, 0.1466, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0172, 0.0180, 0.0225, 0.0207, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:58:11,645 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3354, 2.3728, 2.4376, 4.1793, 2.3356, 2.7343, 2.4204, 2.5821], device='cuda:3'), covar=tensor([0.1469, 0.3616, 0.2948, 0.0613, 0.3941, 0.2505, 0.3758, 0.2981], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0457, 0.0375, 0.0335, 0.0441, 0.0528, 0.0428, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 14:58:19,608 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 3700, loss[loss=0.1738, simple_loss=0.2497, pruned_loss=0.04899, over 16505.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2488, pruned_loss=0.04112, over 3298731.47 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:56,577 INFO [optim.py:368] (3/8) 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 14:59:47,511 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8438, 2.7371, 2.7229, 5.0218, 3.9761, 4.3395, 1.9033, 3.1946], device='cuda:3'), covar=tensor([0.1320, 0.0853, 0.1252, 0.0205, 0.0414, 0.0387, 0.1490, 0.0884], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0194, 0.0206, 0.0217, 0.0203, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:59:50,767 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227050.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:59:52,013 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7654, 3.5364, 3.9715, 2.0199, 4.0542, 4.0294, 3.2417, 3.0636], device='cuda:3'), covar=tensor([0.0711, 0.0259, 0.0165, 0.1186, 0.0082, 0.0178, 0.0359, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0139, 0.0082, 0.0128, 0.0129, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 14:59:53,686 INFO [train.py:904] (3/8) Epoch 23, batch 3750, loss[loss=0.1713, simple_loss=0.2467, pruned_loss=0.04791, over 16451.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2493, pruned_loss=0.0423, over 3295927.88 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:00:56,167 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5125, 5.5128, 5.3834, 4.8893, 4.9875, 5.4496, 5.2726, 5.1265], device='cuda:3'), covar=tensor([0.0570, 0.0335, 0.0279, 0.0321, 0.1065, 0.0314, 0.0268, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0462, 0.0367, 0.0365, 0.0371, 0.0426, 0.0249, 0.0438], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 15:01:07,855 INFO [train.py:904] (3/8) Epoch 23, batch 3800, loss[loss=0.1581, simple_loss=0.2368, pruned_loss=0.03975, over 16830.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.251, pruned_loss=0.0439, over 3288225.90 frames. ], batch size: 102, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:01:25,256 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.146e+02 2.630e+02 3.388e+02 6.503e+02, threshold=5.260e+02, percent-clipped=3.0 2023-05-01 15:02:21,617 INFO [train.py:904] (3/8) Epoch 23, batch 3850, loss[loss=0.1888, simple_loss=0.2727, pruned_loss=0.05245, over 17056.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2516, pruned_loss=0.04483, over 3282809.98 frames. ], batch size: 53, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:34,969 INFO [train.py:904] (3/8) Epoch 23, batch 3900, loss[loss=0.1369, simple_loss=0.2174, pruned_loss=0.0282, over 16859.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.251, pruned_loss=0.045, over 3284497.58 frames. ], batch size: 96, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:51,652 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.230e+02 2.530e+02 3.049e+02 6.179e+02, threshold=5.060e+02, percent-clipped=1.0 2023-05-01 15:03:59,542 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3424, 3.1077, 3.5154, 1.7336, 3.5630, 3.5878, 2.9582, 2.6805], device='cuda:3'), covar=tensor([0.0799, 0.0287, 0.0197, 0.1264, 0.0127, 0.0225, 0.0410, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0140, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 15:04:47,727 INFO [train.py:904] (3/8) Epoch 23, batch 3950, loss[loss=0.1571, simple_loss=0.2322, pruned_loss=0.04104, over 16816.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2511, pruned_loss=0.04574, over 3297103.58 frames. ], batch size: 90, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:00,634 INFO [train.py:904] (3/8) Epoch 23, batch 4000, loss[loss=0.1755, simple_loss=0.2589, pruned_loss=0.04602, over 16517.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2508, pruned_loss=0.04597, over 3304125.41 frames. ], batch size: 75, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:17,147 INFO [optim.py:368] (3/8) 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 15:07:01,804 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:07:13,384 INFO [train.py:904] (3/8) Epoch 23, batch 4050, loss[loss=0.1582, simple_loss=0.2443, pruned_loss=0.03605, over 16761.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2516, pruned_loss=0.04535, over 3293011.51 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:00,764 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 15:08:05,197 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 15:08:27,045 INFO [train.py:904] (3/8) Epoch 23, batch 4100, loss[loss=0.1835, simple_loss=0.2737, pruned_loss=0.04666, over 16643.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2533, pruned_loss=0.04504, over 3279735.65 frames. ], batch size: 76, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:39,127 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227411.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:08:39,599 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-01 15:08:42,773 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.752e+02 1.921e+02 2.258e+02 5.560e+02, threshold=3.841e+02, percent-clipped=1.0 2023-05-01 15:09:17,470 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8200, 2.0056, 2.4725, 2.7906, 2.7667, 3.1459, 1.9596, 3.1045], device='cuda:3'), covar=tensor([0.0239, 0.0516, 0.0337, 0.0354, 0.0317, 0.0231, 0.0632, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0195, 0.0183, 0.0190, 0.0203, 0.0159, 0.0200, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:09:45,380 INFO [train.py:904] (3/8) Epoch 23, batch 4150, loss[loss=0.1829, simple_loss=0.2764, pruned_loss=0.04473, over 16900.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2603, pruned_loss=0.04713, over 3260418.98 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:10:16,193 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227472.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:10:20,790 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6582, 3.0364, 3.2855, 2.0020, 2.7792, 2.1314, 3.2219, 3.2343], device='cuda:3'), covar=tensor([0.0280, 0.0790, 0.0568, 0.2046, 0.0907, 0.1008, 0.0631, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0168, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 15:11:03,934 INFO [train.py:904] (3/8) Epoch 23, batch 4200, loss[loss=0.1906, simple_loss=0.2868, pruned_loss=0.04717, over 16458.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2673, pruned_loss=0.0484, over 3235157.49 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:11:20,462 INFO [optim.py:368] (3/8) 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 15:12:08,403 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-05-01 15:12:19,943 INFO [train.py:904] (3/8) Epoch 23, batch 4250, loss[loss=0.1893, simple_loss=0.2841, pruned_loss=0.0473, over 17162.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2708, pruned_loss=0.04812, over 3215040.69 frames. ], batch size: 46, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:13:36,401 INFO [train.py:904] (3/8) Epoch 23, batch 4300, loss[loss=0.1879, simple_loss=0.2827, pruned_loss=0.04648, over 16724.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.272, pruned_loss=0.04731, over 3184189.90 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:13:55,080 INFO [optim.py:368] (3/8) 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 15:14:41,963 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=227645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:14:53,553 INFO [train.py:904] (3/8) Epoch 23, batch 4350, loss[loss=0.2146, simple_loss=0.3018, pruned_loss=0.06375, over 17258.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2756, pruned_loss=0.04853, over 3187572.26 frames. ], batch size: 52, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:15:45,301 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 15:15:55,221 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:15:55,353 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:16:09,817 INFO [train.py:904] (3/8) Epoch 23, batch 4400, loss[loss=0.2155, simple_loss=0.2997, pruned_loss=0.06571, over 16740.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2777, pruned_loss=0.04972, over 3178538.41 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:16:27,161 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.099e+02 2.601e+02 2.882e+02 6.298e+02, threshold=5.202e+02, percent-clipped=2.0 2023-05-01 15:16:41,893 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6665, 2.8664, 2.6455, 4.6422, 2.5754, 3.1906, 2.7772, 2.8306], device='cuda:3'), covar=tensor([0.1097, 0.2694, 0.2426, 0.0356, 0.3418, 0.1956, 0.2694, 0.2938], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0457, 0.0372, 0.0332, 0.0439, 0.0527, 0.0426, 0.0534], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:17:22,036 INFO [train.py:904] (3/8) Epoch 23, batch 4450, loss[loss=0.2132, simple_loss=0.3002, pruned_loss=0.06304, over 17132.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2808, pruned_loss=0.05084, over 3199250.53 frames. ], batch size: 47, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:17:23,554 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:18:23,591 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 15:18:35,087 INFO [train.py:904] (3/8) Epoch 23, batch 4500, loss[loss=0.2035, simple_loss=0.2934, pruned_loss=0.0568, over 16499.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2814, pruned_loss=0.0517, over 3221312.85 frames. ], batch size: 75, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:18:52,359 INFO [optim.py:368] (3/8) 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 15:19:11,695 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 4550, loss[loss=0.2015, simple_loss=0.2881, pruned_loss=0.05748, over 17206.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.282, pruned_loss=0.05257, over 3224785.79 frames. ], batch size: 43, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:20:39,349 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227888.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:21:00,778 INFO [train.py:904] (3/8) Epoch 23, batch 4600, loss[loss=0.1817, simple_loss=0.275, pruned_loss=0.04426, over 16347.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.283, pruned_loss=0.05301, over 3218432.46 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:21:18,012 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3452, 3.8951, 3.8194, 2.5423, 3.5510, 3.9087, 3.4892, 2.2194], device='cuda:3'), covar=tensor([0.0533, 0.0042, 0.0051, 0.0411, 0.0090, 0.0113, 0.0099, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0084, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 15:21:18,784 INFO [optim.py:368] (3/8) 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 15:21:39,976 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3437, 2.5543, 2.3889, 4.1726, 2.3390, 2.9526, 2.6098, 2.6592], device='cuda:3'), covar=tensor([0.1299, 0.2969, 0.2648, 0.0496, 0.3734, 0.2006, 0.2871, 0.3001], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0455, 0.0370, 0.0330, 0.0438, 0.0524, 0.0424, 0.0530], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:22:14,195 INFO [train.py:904] (3/8) Epoch 23, batch 4650, loss[loss=0.2011, simple_loss=0.2924, pruned_loss=0.05487, over 15495.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2818, pruned_loss=0.05261, over 3210013.21 frames. ], batch size: 191, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:28,874 INFO [train.py:904] (3/8) Epoch 23, batch 4700, loss[loss=0.1944, simple_loss=0.2763, pruned_loss=0.0562, over 11985.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2793, pruned_loss=0.05181, over 3211616.38 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:34,926 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:23:45,412 INFO [optim.py:368] (3/8) 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 15:24:34,809 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 4750, loss[loss=0.196, simple_loss=0.2755, pruned_loss=0.05828, over 12166.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2753, pruned_loss=0.04946, over 3210856.33 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:24:45,262 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9618, 3.9582, 3.9287, 2.9350, 3.9330, 1.7000, 3.6977, 3.3688], device='cuda:3'), covar=tensor([0.0185, 0.0205, 0.0207, 0.0519, 0.0151, 0.3211, 0.0182, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0162, 0.0203, 0.0180, 0.0180, 0.0209, 0.0191, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:25:02,201 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:03,388 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:19,477 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228080.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:34,861 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6769, 3.6973, 2.3476, 4.4983, 2.9486, 4.3385, 2.5765, 3.0042], device='cuda:3'), covar=tensor([0.0292, 0.0424, 0.1725, 0.0149, 0.0843, 0.0488, 0.1460, 0.0876], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0165, 0.0176, 0.0218, 0.0201, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 15:25:40,504 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 15:25:48,661 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-01 15:25:54,063 INFO [train.py:904] (3/8) Epoch 23, batch 4800, loss[loss=0.1738, simple_loss=0.2542, pruned_loss=0.04676, over 16442.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2714, pruned_loss=0.04764, over 3213443.66 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:26:10,773 INFO [optim.py:368] (3/8) 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 15:26:11,948 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228115.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:26:25,115 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 15:26:50,641 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228141.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:27:05,030 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4042, 3.5279, 3.6674, 3.6529, 3.6591, 3.5126, 3.5146, 3.5277], device='cuda:3'), covar=tensor([0.0365, 0.0594, 0.0441, 0.0438, 0.0489, 0.0455, 0.0737, 0.0480], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0454, 0.0441, 0.0410, 0.0487, 0.0461, 0.0546, 0.0369], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 15:27:07,818 INFO [train.py:904] (3/8) Epoch 23, batch 4850, loss[loss=0.1763, simple_loss=0.2739, pruned_loss=0.03936, over 15377.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2733, pruned_loss=0.04783, over 3172256.57 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:27:54,521 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228183.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:28:19,146 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2160, 3.0939, 3.4182, 1.5889, 3.5761, 3.5992, 2.8507, 2.6085], device='cuda:3'), covar=tensor([0.0929, 0.0298, 0.0190, 0.1443, 0.0086, 0.0154, 0.0421, 0.0565], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0108, 0.0099, 0.0139, 0.0081, 0.0126, 0.0128, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 15:28:23,601 INFO [train.py:904] (3/8) Epoch 23, batch 4900, loss[loss=0.1695, simple_loss=0.2671, pruned_loss=0.03593, over 16698.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2731, pruned_loss=0.04665, over 3164740.64 frames. ], batch size: 76, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:28:42,121 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.946e+02 2.242e+02 2.632e+02 4.949e+02, threshold=4.484e+02, percent-clipped=2.0 2023-05-01 15:29:36,711 INFO [train.py:904] (3/8) Epoch 23, batch 4950, loss[loss=0.2086, simple_loss=0.2912, pruned_loss=0.06303, over 11717.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2724, pruned_loss=0.04607, over 3160934.91 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:30:32,325 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0962, 2.2897, 2.2990, 3.9230, 2.1752, 2.6236, 2.3138, 2.4529], device='cuda:3'), covar=tensor([0.1514, 0.3444, 0.2938, 0.0547, 0.3906, 0.2351, 0.3620, 0.2999], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0456, 0.0372, 0.0331, 0.0438, 0.0525, 0.0426, 0.0532], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:30:51,762 INFO [train.py:904] (3/8) Epoch 23, batch 5000, loss[loss=0.164, simple_loss=0.2647, pruned_loss=0.03168, over 16854.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2736, pruned_loss=0.0456, over 3192962.42 frames. ], batch size: 102, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:30:54,348 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-01 15:31:09,998 INFO [optim.py:368] (3/8) 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 15:31:17,531 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5957, 4.5562, 4.4998, 3.6692, 4.5185, 1.5484, 4.2707, 4.1146], device='cuda:3'), covar=tensor([0.0111, 0.0109, 0.0170, 0.0480, 0.0127, 0.3039, 0.0154, 0.0288], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0177, 0.0178, 0.0207, 0.0189, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:31:19,861 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2460, 4.0642, 4.0325, 2.5617, 3.6086, 4.0593, 3.5485, 2.2629], device='cuda:3'), covar=tensor([0.0583, 0.0044, 0.0042, 0.0417, 0.0092, 0.0083, 0.0115, 0.0445], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0100, 0.0112, 0.0096, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 15:31:30,331 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:36,881 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:58,798 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:04,592 INFO [train.py:904] (3/8) Epoch 23, batch 5050, loss[loss=0.1733, simple_loss=0.2605, pruned_loss=0.04307, over 16447.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2738, pruned_loss=0.04534, over 3199909.06 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:32:19,725 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228363.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:19,843 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6075, 4.6659, 4.5126, 4.1359, 4.1762, 4.5772, 4.3209, 4.3097], device='cuda:3'), covar=tensor([0.0600, 0.0451, 0.0290, 0.0313, 0.0944, 0.0515, 0.0551, 0.0601], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0439, 0.0347, 0.0345, 0.0351, 0.0404, 0.0237, 0.0413], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:32:58,967 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228390.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:04,512 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:17,240 INFO [train.py:904] (3/8) Epoch 23, batch 5100, loss[loss=0.1645, simple_loss=0.2513, pruned_loss=0.03892, over 16647.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2722, pruned_loss=0.04468, over 3204863.13 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:33:34,756 INFO [optim.py:368] (3/8) 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 15:33:55,554 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228429.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:34:06,505 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 5150, loss[loss=0.1798, simple_loss=0.2783, pruned_loss=0.04059, over 16407.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2718, pruned_loss=0.04384, over 3210916.28 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:35:05,127 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 15:35:08,725 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 15:35:15,453 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:35:26,011 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228490.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:35:44,364 INFO [train.py:904] (3/8) Epoch 23, batch 5200, loss[loss=0.1746, simple_loss=0.2703, pruned_loss=0.03948, over 16263.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2707, pruned_loss=0.0436, over 3209447.71 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:36:01,315 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228531.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:36:57,627 INFO [train.py:904] (3/8) Epoch 23, batch 5250, loss[loss=0.1749, simple_loss=0.2666, pruned_loss=0.04158, over 16936.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2681, pruned_loss=0.04334, over 3226468.94 frames. ], batch size: 109, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:38:10,579 INFO [train.py:904] (3/8) Epoch 23, batch 5300, loss[loss=0.1793, simple_loss=0.2586, pruned_loss=0.04994, over 12341.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2648, pruned_loss=0.0423, over 3221821.82 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:38:13,273 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.25 vs. limit=5.0 2023-05-01 15:38:28,435 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 1.965e+02 2.226e+02 2.604e+02 5.338e+02, threshold=4.452e+02, percent-clipped=3.0 2023-05-01 15:39:23,394 INFO [train.py:904] (3/8) Epoch 23, batch 5350, loss[loss=0.1846, simple_loss=0.2746, pruned_loss=0.04732, over 16710.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2633, pruned_loss=0.04197, over 3221454.15 frames. ], batch size: 124, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:39:34,145 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3075, 3.3928, 1.9358, 3.7830, 2.5076, 3.7294, 2.1268, 2.7111], device='cuda:3'), covar=tensor([0.0299, 0.0412, 0.1764, 0.0201, 0.0961, 0.0626, 0.1629, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0164, 0.0176, 0.0217, 0.0201, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 15:39:37,187 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1071, 4.1417, 4.4022, 4.3901, 4.4030, 4.1620, 4.1410, 4.1252], device='cuda:3'), covar=tensor([0.0329, 0.0693, 0.0410, 0.0397, 0.0496, 0.0403, 0.0901, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0454, 0.0441, 0.0411, 0.0487, 0.0461, 0.0547, 0.0371], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 15:39:38,468 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:10,112 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228685.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:10,385 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7570, 4.0113, 2.9525, 2.3857, 2.7676, 2.5856, 4.3060, 3.6037], device='cuda:3'), covar=tensor([0.2828, 0.0626, 0.1869, 0.2773, 0.2792, 0.1934, 0.0483, 0.1212], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0271, 0.0308, 0.0316, 0.0299, 0.0263, 0.0300, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 15:40:16,732 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 5400, loss[loss=0.192, simple_loss=0.2841, pruned_loss=0.04993, over 16672.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2662, pruned_loss=0.04274, over 3198824.54 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:40:46,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9415, 2.1219, 2.2247, 3.4239, 2.0495, 2.4178, 2.2600, 2.3195], device='cuda:3'), covar=tensor([0.1508, 0.3703, 0.2948, 0.0655, 0.4173, 0.2659, 0.3720, 0.3225], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0453, 0.0371, 0.0330, 0.0436, 0.0522, 0.0424, 0.0529], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:40:48,793 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:51,849 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 15:40:54,347 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.938e+02 2.190e+02 2.538e+02 4.586e+02, threshold=4.379e+02, percent-clipped=1.0 2023-05-01 15:41:27,333 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:41:54,046 INFO [train.py:904] (3/8) Epoch 23, batch 5450, loss[loss=0.2242, simple_loss=0.3184, pruned_loss=0.065, over 16394.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2693, pruned_loss=0.0441, over 3198148.78 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:41:56,910 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5253, 4.7326, 4.4997, 4.2079, 3.8072, 4.6796, 4.5005, 4.2779], device='cuda:3'), covar=tensor([0.1004, 0.0801, 0.0476, 0.0416, 0.1794, 0.0586, 0.0554, 0.0740], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0440, 0.0347, 0.0346, 0.0353, 0.0405, 0.0237, 0.0414], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:42:42,359 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228784.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:42:44,382 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228785.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 15:43:12,590 INFO [train.py:904] (3/8) Epoch 23, batch 5500, loss[loss=0.1971, simple_loss=0.2959, pruned_loss=0.04912, over 16808.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2764, pruned_loss=0.04873, over 3155447.07 frames. ], batch size: 102, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:43:23,458 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 15:43:25,781 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 15:43:32,426 INFO [optim.py:368] (3/8) 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 15:44:00,451 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-01 15:44:14,891 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-01 15:44:31,577 INFO [train.py:904] (3/8) Epoch 23, batch 5550, loss[loss=0.1919, simple_loss=0.2794, pruned_loss=0.05217, over 16422.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.283, pruned_loss=0.053, over 3147129.31 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:44:47,159 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 15:45:42,533 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9648, 4.0255, 4.2855, 4.2546, 4.2892, 4.0358, 4.0211, 4.0216], device='cuda:3'), covar=tensor([0.0385, 0.0685, 0.0444, 0.0472, 0.0512, 0.0476, 0.0984, 0.0582], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0457, 0.0443, 0.0411, 0.0488, 0.0464, 0.0550, 0.0371], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 15:45:52,969 INFO [train.py:904] (3/8) Epoch 23, batch 5600, loss[loss=0.275, simple_loss=0.3346, pruned_loss=0.1077, over 11463.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2877, pruned_loss=0.05726, over 3104293.57 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:46:10,853 INFO [optim.py:368] (3/8) 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 15:46:45,606 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:47:13,516 INFO [train.py:904] (3/8) Epoch 23, batch 5650, loss[loss=0.2031, simple_loss=0.2946, pruned_loss=0.0558, over 16607.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2924, pruned_loss=0.06137, over 3056438.55 frames. ], batch size: 75, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:47:45,309 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 15:47:56,961 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5743, 4.5350, 4.4547, 3.6914, 4.4947, 1.6308, 4.2740, 4.0803], device='cuda:3'), covar=tensor([0.0098, 0.0090, 0.0173, 0.0336, 0.0092, 0.2833, 0.0122, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0159, 0.0199, 0.0177, 0.0177, 0.0205, 0.0188, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:48:04,589 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228985.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:10,673 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:15,819 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 15:48:22,455 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228996.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:22,668 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-01 15:48:32,719 INFO [train.py:904] (3/8) Epoch 23, batch 5700, loss[loss=0.2184, simple_loss=0.3216, pruned_loss=0.05764, over 16421.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2934, pruned_loss=0.062, over 3060291.48 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:48:50,700 INFO [optim.py:368] (3/8) 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 15:49:03,445 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 15:49:18,828 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:50,896 INFO [train.py:904] (3/8) Epoch 23, batch 5750, loss[loss=0.1942, simple_loss=0.2873, pruned_loss=0.05055, over 16721.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2955, pruned_loss=0.06282, over 3047786.57 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:50:44,176 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229085.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:51:13,970 INFO [train.py:904] (3/8) Epoch 23, batch 5800, loss[loss=0.1818, simple_loss=0.2778, pruned_loss=0.04292, over 16538.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.295, pruned_loss=0.06209, over 3027498.04 frames. ], batch size: 75, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:51:32,213 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.995e+02 2.918e+02 3.323e+02 4.052e+02 8.500e+02, threshold=6.645e+02, percent-clipped=1.0 2023-05-01 15:52:01,459 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229133.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:52:31,397 INFO [train.py:904] (3/8) Epoch 23, batch 5850, loss[loss=0.1746, simple_loss=0.2631, pruned_loss=0.04306, over 17156.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2931, pruned_loss=0.06062, over 3040749.52 frames. ], batch size: 47, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:53:53,139 INFO [train.py:904] (3/8) Epoch 23, batch 5900, loss[loss=0.19, simple_loss=0.2838, pruned_loss=0.04814, over 17162.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2924, pruned_loss=0.06027, over 3040913.90 frames. ], batch size: 46, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:54:16,064 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 23, batch 5950, loss[loss=0.19, simple_loss=0.2832, pruned_loss=0.04837, over 16719.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2928, pruned_loss=0.05864, over 3066445.98 frames. ], batch size: 134, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:06,524 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8770, 2.1488, 2.4149, 3.1019, 2.2389, 2.3640, 2.3457, 2.2634], device='cuda:3'), covar=tensor([0.1346, 0.3245, 0.2398, 0.0727, 0.3854, 0.2291, 0.3110, 0.3211], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0449, 0.0366, 0.0325, 0.0432, 0.0516, 0.0420, 0.0524], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:56:08,034 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 15:56:18,151 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229291.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:33,786 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:37,756 INFO [train.py:904] (3/8) Epoch 23, batch 6000, loss[loss=0.2105, simple_loss=0.2924, pruned_loss=0.06432, over 15362.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2919, pruned_loss=0.05858, over 3067294.56 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:37,756 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 15:56:49,494 INFO [train.py:938] (3/8) Epoch 23, validation: loss=0.1497, simple_loss=0.2623, pruned_loss=0.01859, over 944034.00 frames. 2023-05-01 15:56:49,495 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 15:57:07,745 INFO [optim.py:368] (3/8) 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 15:58:06,100 INFO [train.py:904] (3/8) Epoch 23, batch 6050, loss[loss=0.1802, simple_loss=0.2798, pruned_loss=0.04025, over 17282.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2915, pruned_loss=0.05857, over 3076754.16 frames. ], batch size: 52, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:58:20,426 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:58:46,530 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9591, 2.4061, 1.8951, 2.1830, 2.7534, 2.3576, 2.7068, 2.9167], device='cuda:3'), covar=tensor([0.0193, 0.0457, 0.0623, 0.0500, 0.0275, 0.0431, 0.0224, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0237, 0.0227, 0.0229, 0.0237, 0.0237, 0.0237, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 15:58:50,009 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 15:59:07,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4856, 4.5533, 4.8551, 4.8318, 4.8494, 4.5574, 4.4920, 4.3982], device='cuda:3'), covar=tensor([0.0398, 0.0681, 0.0482, 0.0464, 0.0557, 0.0434, 0.1144, 0.0565], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0460, 0.0446, 0.0415, 0.0491, 0.0466, 0.0552, 0.0373], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 15:59:21,990 INFO [train.py:904] (3/8) Epoch 23, batch 6100, loss[loss=0.2219, simple_loss=0.2927, pruned_loss=0.07552, over 11632.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2914, pruned_loss=0.05752, over 3097570.49 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:59:40,505 INFO [optim.py:368] (3/8) 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 15:59:55,680 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 16:00:22,345 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:00:31,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.8490, 6.1921, 5.8561, 6.0346, 5.6313, 5.4673, 5.6206, 6.3235], device='cuda:3'), covar=tensor([0.1145, 0.0744, 0.0929, 0.0796, 0.0743, 0.0641, 0.1133, 0.0778], device='cuda:3'), in_proj_covar=tensor([0.0675, 0.0820, 0.0678, 0.0625, 0.0519, 0.0528, 0.0690, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:00:36,341 INFO [train.py:904] (3/8) Epoch 23, batch 6150, loss[loss=0.2021, simple_loss=0.2836, pruned_loss=0.06029, over 16462.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.29, pruned_loss=0.05768, over 3085497.10 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 16:00:43,906 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4074, 3.2874, 2.6736, 2.1574, 2.2274, 2.3429, 3.3784, 3.0502], device='cuda:3'), covar=tensor([0.3006, 0.0675, 0.1795, 0.2694, 0.2510, 0.2146, 0.0533, 0.1363], device='cuda:3'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0317, 0.0300, 0.0263, 0.0300, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:01:15,276 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6200, loss[loss=0.1774, simple_loss=0.2699, pruned_loss=0.0425, over 16762.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2883, pruned_loss=0.05737, over 3071693.95 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:01:55,811 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:02:22,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4884, 4.4710, 4.3709, 3.6253, 4.4123, 1.7321, 4.1605, 4.0516], device='cuda:3'), covar=tensor([0.0119, 0.0107, 0.0194, 0.0386, 0.0108, 0.2777, 0.0149, 0.0257], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0160, 0.0201, 0.0179, 0.0178, 0.0207, 0.0189, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:02:49,454 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6250, loss[loss=0.2454, simple_loss=0.3121, pruned_loss=0.08932, over 11800.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2875, pruned_loss=0.0567, over 3101428.12 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:03:46,482 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3548, 4.4214, 4.2198, 3.9252, 3.9210, 4.3201, 3.9794, 4.0844], device='cuda:3'), covar=tensor([0.0705, 0.0794, 0.0315, 0.0346, 0.0854, 0.0627, 0.0899, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0436, 0.0343, 0.0340, 0.0348, 0.0399, 0.0233, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:03:48,960 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:07,466 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4077, 3.3810, 3.4588, 3.5512, 3.5728, 3.3135, 3.5599, 3.6254], device='cuda:3'), covar=tensor([0.1263, 0.1021, 0.1066, 0.0636, 0.0670, 0.2336, 0.0989, 0.0777], device='cuda:3'), in_proj_covar=tensor([0.0634, 0.0787, 0.0906, 0.0792, 0.0604, 0.0630, 0.0653, 0.0758], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:04:09,701 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6300, loss[loss=0.1877, simple_loss=0.2774, pruned_loss=0.04902, over 16729.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2868, pruned_loss=0.05582, over 3097059.04 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:04:50,639 INFO [optim.py:368] (3/8) 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 16:05:12,767 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0919, 5.0961, 4.9590, 4.5738, 4.5615, 4.9955, 4.8983, 4.7105], device='cuda:3'), covar=tensor([0.0635, 0.0590, 0.0319, 0.0349, 0.1163, 0.0542, 0.0356, 0.0698], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0436, 0.0343, 0.0340, 0.0348, 0.0399, 0.0234, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:05:23,335 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8860, 2.7308, 2.6128, 1.9493, 2.5789, 2.7241, 2.5787, 1.9832], device='cuda:3'), covar=tensor([0.0457, 0.0102, 0.0104, 0.0395, 0.0141, 0.0143, 0.0123, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0084, 0.0085, 0.0133, 0.0098, 0.0110, 0.0095, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 16:05:27,296 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:48,006 INFO [train.py:904] (3/8) Epoch 23, batch 6350, loss[loss=0.1949, simple_loss=0.2835, pruned_loss=0.0531, over 16531.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2874, pruned_loss=0.05694, over 3095015.35 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:05:53,554 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229656.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:57,369 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229659.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:06:16,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0884, 2.2348, 2.1762, 3.7089, 2.1284, 2.5552, 2.3492, 2.3215], device='cuda:3'), covar=tensor([0.1340, 0.3365, 0.2886, 0.0532, 0.4012, 0.2437, 0.3373, 0.3279], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0451, 0.0368, 0.0326, 0.0435, 0.0518, 0.0422, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:06:57,754 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4252, 3.3005, 2.7304, 2.1826, 2.2108, 2.3271, 3.4320, 3.0105], device='cuda:3'), covar=tensor([0.3104, 0.0633, 0.1859, 0.3018, 0.2717, 0.2242, 0.0549, 0.1403], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0269, 0.0306, 0.0315, 0.0297, 0.0261, 0.0299, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:07:04,876 INFO [train.py:904] (3/8) Epoch 23, batch 6400, loss[loss=0.1938, simple_loss=0.281, pruned_loss=0.05326, over 16200.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2874, pruned_loss=0.05775, over 3098214.56 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:07:05,423 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6381, 2.5334, 1.9737, 2.7081, 2.1971, 2.7468, 2.1800, 2.3683], device='cuda:3'), covar=tensor([0.0316, 0.0366, 0.1174, 0.0313, 0.0620, 0.0548, 0.1223, 0.0619], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0195, 0.0165, 0.0177, 0.0218, 0.0203, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:07:16,562 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4731, 4.6365, 4.7892, 4.5834, 4.6649, 5.1609, 4.6514, 4.4093], device='cuda:3'), covar=tensor([0.1510, 0.1823, 0.2310, 0.1785, 0.2407, 0.1010, 0.1712, 0.2382], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0598, 0.0659, 0.0492, 0.0655, 0.0690, 0.0515, 0.0662], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:07:24,219 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6760, 4.6936, 4.5409, 4.1855, 4.2173, 4.6201, 4.4064, 4.3382], device='cuda:3'), covar=tensor([0.0596, 0.0569, 0.0308, 0.0359, 0.0930, 0.0492, 0.0528, 0.0688], device='cuda:3'), in_proj_covar=tensor([0.0293, 0.0436, 0.0343, 0.0339, 0.0347, 0.0398, 0.0233, 0.0409], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:07:24,872 INFO [optim.py:368] (3/8) 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 16:07:32,377 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229720.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:07:57,567 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9044, 2.7499, 2.6339, 2.0080, 2.5946, 2.7477, 2.6049, 2.0436], device='cuda:3'), covar=tensor([0.0453, 0.0094, 0.0096, 0.0363, 0.0132, 0.0141, 0.0116, 0.0375], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0084, 0.0085, 0.0133, 0.0098, 0.0111, 0.0095, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 16:08:21,141 INFO [train.py:904] (3/8) Epoch 23, batch 6450, loss[loss=0.1777, simple_loss=0.2702, pruned_loss=0.04258, over 16575.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2869, pruned_loss=0.0568, over 3109426.52 frames. ], batch size: 62, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:34,332 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229799.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:09:40,106 INFO [train.py:904] (3/8) Epoch 23, batch 6500, loss[loss=0.2392, simple_loss=0.3017, pruned_loss=0.08839, over 11511.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2854, pruned_loss=0.05648, over 3117092.43 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:59,295 INFO [optim.py:368] (3/8) 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 16:10:25,443 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6550, loss[loss=0.1972, simple_loss=0.2933, pruned_loss=0.05056, over 15343.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2887, pruned_loss=0.05791, over 3098265.52 frames. ], batch size: 190, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:02,467 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229894.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:12:15,580 INFO [train.py:904] (3/8) Epoch 23, batch 6600, loss[loss=0.1835, simple_loss=0.2728, pruned_loss=0.04711, over 17226.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2911, pruned_loss=0.05832, over 3113181.99 frames. ], batch size: 45, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:35,465 INFO [optim.py:368] (3/8) 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 16:12:43,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8907, 2.1847, 2.3927, 3.1232, 2.2305, 2.3809, 2.3468, 2.2320], device='cuda:3'), covar=tensor([0.1448, 0.3131, 0.2444, 0.0732, 0.3969, 0.2419, 0.3156, 0.3362], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0452, 0.0369, 0.0327, 0.0437, 0.0520, 0.0424, 0.0529], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:13:01,778 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:16,257 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1249, 5.6071, 5.7822, 5.4945, 5.6091, 6.1201, 5.5862, 5.3684], device='cuda:3'), covar=tensor([0.0972, 0.1703, 0.2190, 0.2005, 0.2388, 0.0989, 0.1611, 0.2332], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0603, 0.0664, 0.0496, 0.0660, 0.0692, 0.0519, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:13:33,861 INFO [train.py:904] (3/8) Epoch 23, batch 6650, loss[loss=0.1711, simple_loss=0.2579, pruned_loss=0.04214, over 16678.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2905, pruned_loss=0.05858, over 3114398.17 frames. ], batch size: 57, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:13:37,589 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:38,753 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:14,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6417, 3.0295, 3.2369, 1.9579, 2.8327, 2.0419, 3.2199, 3.2651], device='cuda:3'), covar=tensor([0.0281, 0.0834, 0.0600, 0.2141, 0.0894, 0.1103, 0.0648, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0147, 0.0131, 0.0143, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:14:26,035 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229987.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:52,371 INFO [train.py:904] (3/8) Epoch 23, batch 6700, loss[loss=0.1893, simple_loss=0.2786, pruned_loss=0.05, over 16625.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2899, pruned_loss=0.05905, over 3123166.79 frames. ], batch size: 57, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:14:54,790 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230004.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:15:11,259 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230015.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:15:11,752 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 16:15:12,080 INFO [optim.py:368] (3/8) 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 16:15:23,291 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5785, 2.9879, 3.1702, 1.9775, 2.7778, 2.0189, 3.1807, 3.2110], device='cuda:3'), covar=tensor([0.0299, 0.0790, 0.0620, 0.2137, 0.0909, 0.1114, 0.0676, 0.0892], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0143, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:16:02,763 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:16:09,698 INFO [train.py:904] (3/8) Epoch 23, batch 6750, loss[loss=0.2123, simple_loss=0.2868, pruned_loss=0.06893, over 16512.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2892, pruned_loss=0.05945, over 3119875.66 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:16:22,529 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 16:16:39,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0565, 4.0403, 3.9718, 3.2003, 3.9788, 1.7569, 3.7852, 3.5634], device='cuda:3'), covar=tensor([0.0108, 0.0107, 0.0192, 0.0285, 0.0092, 0.2820, 0.0129, 0.0260], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0176, 0.0207, 0.0189, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:17:00,529 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1484, 3.9872, 4.2284, 4.3405, 4.4498, 4.0480, 4.4030, 4.4998], device='cuda:3'), covar=tensor([0.1711, 0.1297, 0.1407, 0.0720, 0.0576, 0.1320, 0.0799, 0.0665], device='cuda:3'), in_proj_covar=tensor([0.0628, 0.0781, 0.0898, 0.0789, 0.0599, 0.0624, 0.0649, 0.0754], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:17:19,771 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6800, loss[loss=0.2199, simple_loss=0.3063, pruned_loss=0.06679, over 16372.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2889, pruned_loss=0.05933, over 3115339.39 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:43,641 INFO [optim.py:368] (3/8) 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 16:18:11,075 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:31,926 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:36,429 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230150.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:40,518 INFO [train.py:904] (3/8) Epoch 23, batch 6850, loss[loss=0.2052, simple_loss=0.3083, pruned_loss=0.05106, over 16588.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2907, pruned_loss=0.0601, over 3091212.25 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:19:23,351 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6900, loss[loss=0.2095, simple_loss=0.3035, pruned_loss=0.05772, over 16429.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.292, pruned_loss=0.05879, over 3106642.82 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:20:10,807 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:20,076 INFO [optim.py:368] (3/8) 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 16:20:45,477 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:45,811 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 16:20:58,923 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230241.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:13,094 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 6950, loss[loss=0.2462, simple_loss=0.3107, pruned_loss=0.09087, over 11336.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2935, pruned_loss=0.0603, over 3089339.03 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:01,066 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230281.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:28,309 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 16:22:34,112 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230302.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,886 INFO [train.py:904] (3/8) Epoch 23, batch 7000, loss[loss=0.19, simple_loss=0.2963, pruned_loss=0.04186, over 16761.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2935, pruned_loss=0.05958, over 3084149.31 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:52,320 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5792, 3.6964, 2.7216, 2.2694, 2.4232, 2.3970, 3.8533, 3.2985], device='cuda:3'), covar=tensor([0.2988, 0.0646, 0.1892, 0.2868, 0.2684, 0.2162, 0.0549, 0.1327], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0268, 0.0305, 0.0314, 0.0296, 0.0260, 0.0298, 0.0338], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:22:53,492 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230315.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:22:56,094 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.703e+02 3.306e+02 4.193e+02 6.685e+02, threshold=6.612e+02, percent-clipped=2.0 2023-05-01 16:23:35,807 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 7050, loss[loss=0.1919, simple_loss=0.2814, pruned_loss=0.05124, over 16643.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2939, pruned_loss=0.05912, over 3091156.84 frames. ], batch size: 57, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:24:05,842 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230363.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:24:20,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4672, 4.4709, 4.8414, 4.7930, 4.8184, 4.5180, 4.4992, 4.3424], device='cuda:3'), covar=tensor([0.0357, 0.0629, 0.0359, 0.0391, 0.0453, 0.0385, 0.0975, 0.0556], device='cuda:3'), in_proj_covar=tensor([0.0416, 0.0463, 0.0449, 0.0416, 0.0492, 0.0471, 0.0557, 0.0376], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 16:25:07,481 INFO [train.py:904] (3/8) Epoch 23, batch 7100, loss[loss=0.2169, simple_loss=0.3017, pruned_loss=0.06605, over 16444.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2924, pruned_loss=0.05836, over 3111125.66 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:25:30,796 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.925e+02 3.270e+02 3.973e+02 7.550e+02, threshold=6.541e+02, percent-clipped=2.0 2023-05-01 16:25:47,920 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6296, 2.4281, 2.4341, 3.6530, 2.4908, 3.8386, 1.4664, 2.7527], device='cuda:3'), covar=tensor([0.1531, 0.0915, 0.1361, 0.0205, 0.0264, 0.0436, 0.1900, 0.0918], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0200, 0.0195, 0.0209, 0.0219, 0.0207, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:26:17,481 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1533, 2.2233, 2.2372, 3.7802, 2.1615, 2.5860, 2.3002, 2.3838], device='cuda:3'), covar=tensor([0.1375, 0.3486, 0.2999, 0.0574, 0.4099, 0.2424, 0.3344, 0.3373], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0450, 0.0367, 0.0325, 0.0435, 0.0517, 0.0421, 0.0526], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:26:24,957 INFO [train.py:904] (3/8) Epoch 23, batch 7150, loss[loss=0.2684, simple_loss=0.3254, pruned_loss=0.1056, over 11363.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2908, pruned_loss=0.05873, over 3092897.12 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:26:49,319 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:27:42,211 INFO [train.py:904] (3/8) Epoch 23, batch 7200, loss[loss=0.1794, simple_loss=0.2711, pruned_loss=0.04388, over 16645.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2888, pruned_loss=0.05725, over 3094146.43 frames. ], batch size: 62, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:27:47,859 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230506.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:03,477 INFO [optim.py:368] (3/8) 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 16:28:16,786 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:16,921 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9231, 2.7333, 2.8237, 2.1086, 2.6712, 2.1590, 2.7248, 2.8993], device='cuda:3'), covar=tensor([0.0279, 0.0764, 0.0492, 0.1728, 0.0770, 0.0861, 0.0547, 0.0640], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:28:22,568 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230529.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:28:30,084 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:58,904 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230550.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:02,529 INFO [train.py:904] (3/8) Epoch 23, batch 7250, loss[loss=0.1855, simple_loss=0.2752, pruned_loss=0.04788, over 16208.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2869, pruned_loss=0.05672, over 3066965.65 frames. ], batch size: 165, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:29:51,466 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:58,437 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1692, 4.2025, 4.5188, 4.4913, 4.5170, 4.2529, 4.2613, 4.1765], device='cuda:3'), covar=tensor([0.0378, 0.0646, 0.0421, 0.0443, 0.0524, 0.0400, 0.0937, 0.0550], device='cuda:3'), in_proj_covar=tensor([0.0416, 0.0462, 0.0448, 0.0416, 0.0493, 0.0470, 0.0557, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 16:30:03,505 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:08,115 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230597.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:10,098 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 7300, loss[loss=0.2059, simple_loss=0.2998, pruned_loss=0.05604, over 16212.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2865, pruned_loss=0.05704, over 3067162.61 frames. ], batch size: 165, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:30:21,153 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 16:30:39,629 INFO [optim.py:368] (3/8) 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 16:30:57,546 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7181, 3.8620, 2.9352, 2.3310, 2.7086, 2.5817, 4.3652, 3.3839], device='cuda:3'), covar=tensor([0.2777, 0.0639, 0.1788, 0.2559, 0.2586, 0.1968, 0.0362, 0.1240], device='cuda:3'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0318, 0.0300, 0.0263, 0.0300, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:30:59,104 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 16:31:19,777 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 7350, loss[loss=0.1809, simple_loss=0.275, pruned_loss=0.04342, over 16543.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2876, pruned_loss=0.05809, over 3037809.25 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:31:34,897 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7751, 1.3701, 1.6585, 1.6444, 1.7286, 1.9587, 1.6220, 1.8267], device='cuda:3'), covar=tensor([0.0264, 0.0420, 0.0239, 0.0318, 0.0289, 0.0165, 0.0435, 0.0141], device='cuda:3'), in_proj_covar=tensor([0.0190, 0.0193, 0.0179, 0.0185, 0.0199, 0.0155, 0.0197, 0.0154], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:32:33,455 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 7400, loss[loss=0.1923, simple_loss=0.2804, pruned_loss=0.05207, over 16384.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2885, pruned_loss=0.05823, over 3053817.31 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:33:16,062 INFO [optim.py:368] (3/8) 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 16:34:13,422 INFO [train.py:904] (3/8) Epoch 23, batch 7450, loss[loss=0.2555, simple_loss=0.3136, pruned_loss=0.09864, over 11625.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.289, pruned_loss=0.0589, over 3056422.13 frames. ], batch size: 248, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:18,622 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:35,341 INFO [train.py:904] (3/8) Epoch 23, batch 7500, loss[loss=0.2575, simple_loss=0.319, pruned_loss=0.09801, over 11470.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2895, pruned_loss=0.05784, over 3084850.79 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:40,979 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:53,684 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-01 16:35:58,126 INFO [optim.py:368] (3/8) 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 16:36:09,653 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230824.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:36:09,725 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3120, 3.4354, 3.6059, 3.5748, 3.5956, 3.4128, 3.4602, 3.4881], device='cuda:3'), covar=tensor([0.0448, 0.0717, 0.0444, 0.0493, 0.0599, 0.0564, 0.0847, 0.0556], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0457, 0.0443, 0.0412, 0.0488, 0.0465, 0.0549, 0.0372], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 16:36:24,340 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6741, 3.8088, 4.0716, 4.0191, 4.0697, 3.8036, 3.7567, 3.8555], device='cuda:3'), covar=tensor([0.0627, 0.1017, 0.0598, 0.0707, 0.0692, 0.0707, 0.1378, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0412, 0.0457, 0.0442, 0.0412, 0.0488, 0.0465, 0.0549, 0.0372], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 16:36:52,039 INFO [train.py:904] (3/8) Epoch 23, batch 7550, loss[loss=0.1607, simple_loss=0.2556, pruned_loss=0.03283, over 16795.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2879, pruned_loss=0.05763, over 3092112.30 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:36:52,483 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:20,007 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5503, 4.4162, 4.6123, 4.7545, 4.9101, 4.4358, 4.8963, 4.9214], device='cuda:3'), covar=tensor([0.1937, 0.1240, 0.1454, 0.0701, 0.0558, 0.1002, 0.0622, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0624, 0.0776, 0.0895, 0.0783, 0.0596, 0.0622, 0.0650, 0.0753], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:37:33,343 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:45,739 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:47,075 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:58,274 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230897.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:38:06,492 INFO [train.py:904] (3/8) Epoch 23, batch 7600, loss[loss=0.1752, simple_loss=0.2682, pruned_loss=0.04111, over 16857.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2869, pruned_loss=0.0574, over 3099355.69 frames. ], batch size: 42, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:38:27,912 INFO [optim.py:368] (3/8) 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 16:39:10,330 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:19,819 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:21,933 INFO [train.py:904] (3/8) Epoch 23, batch 7650, loss[loss=0.1944, simple_loss=0.2881, pruned_loss=0.05035, over 16470.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2884, pruned_loss=0.05899, over 3075104.68 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:39:26,838 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-01 16:39:53,071 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 16:40:13,003 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0681, 2.1518, 2.1744, 3.8022, 2.1042, 2.5050, 2.2559, 2.3395], device='cuda:3'), covar=tensor([0.1426, 0.3572, 0.3109, 0.0579, 0.4345, 0.2521, 0.3459, 0.3363], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0449, 0.0367, 0.0324, 0.0433, 0.0515, 0.0420, 0.0524], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:40:35,933 INFO [train.py:904] (3/8) Epoch 23, batch 7700, loss[loss=0.1806, simple_loss=0.2745, pruned_loss=0.04331, over 16718.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2878, pruned_loss=0.05864, over 3093543.87 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:40:43,302 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8448, 2.9324, 2.5202, 4.7540, 3.6469, 4.2603, 1.6268, 3.1927], device='cuda:3'), covar=tensor([0.1358, 0.0739, 0.1333, 0.0168, 0.0272, 0.0351, 0.1713, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0199, 0.0194, 0.0209, 0.0218, 0.0206, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:40:57,683 INFO [optim.py:368] (3/8) 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 16:41:53,915 INFO [train.py:904] (3/8) Epoch 23, batch 7750, loss[loss=0.2154, simple_loss=0.3067, pruned_loss=0.06208, over 16724.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2881, pruned_loss=0.05838, over 3090036.69 frames. ], batch size: 134, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:43:09,649 INFO [train.py:904] (3/8) Epoch 23, batch 7800, loss[loss=0.1899, simple_loss=0.282, pruned_loss=0.04888, over 16680.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2893, pruned_loss=0.05927, over 3087999.14 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:43:30,321 INFO [optim.py:368] (3/8) 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 16:43:41,684 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231124.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:43:57,926 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8717, 5.1865, 4.9353, 4.9577, 4.7089, 4.6532, 4.6034, 5.2601], device='cuda:3'), covar=tensor([0.1210, 0.0816, 0.0946, 0.0926, 0.0832, 0.1002, 0.1248, 0.0926], device='cuda:3'), in_proj_covar=tensor([0.0675, 0.0817, 0.0680, 0.0629, 0.0520, 0.0532, 0.0691, 0.0647], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:44:17,269 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:24,895 INFO [train.py:904] (3/8) Epoch 23, batch 7850, loss[loss=0.2515, simple_loss=0.3179, pruned_loss=0.09259, over 11669.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2897, pruned_loss=0.05905, over 3075254.87 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:44:52,237 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:52,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8568, 4.8714, 5.2947, 5.2360, 5.3217, 4.9390, 4.8649, 4.6355], device='cuda:3'), covar=tensor([0.0357, 0.0490, 0.0357, 0.0417, 0.0511, 0.0401, 0.1163, 0.0528], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0459, 0.0445, 0.0414, 0.0490, 0.0468, 0.0553, 0.0374], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 16:45:05,931 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:16,886 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231189.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:37,199 INFO [train.py:904] (3/8) Epoch 23, batch 7900, loss[loss=0.2409, simple_loss=0.3056, pruned_loss=0.08808, over 11408.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2888, pruned_loss=0.05843, over 3082244.43 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:45:57,446 INFO [optim.py:368] (3/8) 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 16:45:58,597 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-01 16:46:16,872 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:44,778 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:52,746 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:54,843 INFO [train.py:904] (3/8) Epoch 23, batch 7950, loss[loss=0.2377, simple_loss=0.3007, pruned_loss=0.08741, over 11551.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2896, pruned_loss=0.05894, over 3091780.57 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:11,518 INFO [train.py:904] (3/8) Epoch 23, batch 8000, loss[loss=0.2071, simple_loss=0.3009, pruned_loss=0.05671, over 16379.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2905, pruned_loss=0.0601, over 3070982.43 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:15,257 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8589, 2.9382, 2.3786, 2.8134, 3.2853, 2.9566, 3.4618, 3.4440], device='cuda:3'), covar=tensor([0.0108, 0.0361, 0.0515, 0.0380, 0.0264, 0.0334, 0.0267, 0.0210], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0233, 0.0224, 0.0227, 0.0234, 0.0231, 0.0232, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:48:26,429 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:48:50,536 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0743, 2.3464, 2.3165, 2.6981, 1.7793, 3.1806, 1.8259, 2.7157], device='cuda:3'), covar=tensor([0.1177, 0.0653, 0.1065, 0.0205, 0.0106, 0.0352, 0.1482, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0191, 0.0207, 0.0215, 0.0204, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:49:22,951 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 16:49:26,909 INFO [train.py:904] (3/8) Epoch 23, batch 8050, loss[loss=0.2012, simple_loss=0.2916, pruned_loss=0.05537, over 16877.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2907, pruned_loss=0.06033, over 3058717.68 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:49:38,366 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8876, 2.7639, 2.6439, 1.9536, 2.6286, 2.7435, 2.6164, 1.9090], device='cuda:3'), covar=tensor([0.0415, 0.0095, 0.0082, 0.0365, 0.0130, 0.0132, 0.0123, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0085, 0.0085, 0.0134, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 16:50:14,042 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 16:50:31,616 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 16:50:42,758 INFO [train.py:904] (3/8) Epoch 23, batch 8100, loss[loss=0.2044, simple_loss=0.2931, pruned_loss=0.05784, over 16713.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2901, pruned_loss=0.05978, over 3067357.28 frames. ], batch size: 83, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:51:04,319 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.502e+02 3.046e+02 3.676e+02 7.002e+02, threshold=6.092e+02, percent-clipped=1.0 2023-05-01 16:51:51,674 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:51:59,312 INFO [train.py:904] (3/8) Epoch 23, batch 8150, loss[loss=0.1629, simple_loss=0.2555, pruned_loss=0.03514, over 16882.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2874, pruned_loss=0.05853, over 3069815.98 frames. ], batch size: 102, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:05,171 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:53:14,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4066, 2.9656, 2.7071, 2.3041, 2.2880, 2.3086, 3.0182, 2.8760], device='cuda:3'), covar=tensor([0.2531, 0.0715, 0.1514, 0.2343, 0.2275, 0.2098, 0.0510, 0.1345], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0270, 0.0309, 0.0318, 0.0301, 0.0264, 0.0299, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 16:53:14,922 INFO [train.py:904] (3/8) Epoch 23, batch 8200, loss[loss=0.221, simple_loss=0.2833, pruned_loss=0.07939, over 11851.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2851, pruned_loss=0.05758, over 3093479.25 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:38,107 INFO [optim.py:368] (3/8) 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 16:54:26,346 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:54:37,501 INFO [train.py:904] (3/8) Epoch 23, batch 8250, loss[loss=0.182, simple_loss=0.2695, pruned_loss=0.0472, over 11861.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2841, pruned_loss=0.0556, over 3052152.76 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:55:42,940 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:55:57,538 INFO [train.py:904] (3/8) Epoch 23, batch 8300, loss[loss=0.1641, simple_loss=0.2501, pruned_loss=0.03903, over 11676.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2811, pruned_loss=0.05257, over 3031583.05 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:56:04,249 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231607.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:56:22,160 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.236e+02 2.512e+02 2.973e+02 5.364e+02, threshold=5.024e+02, percent-clipped=0.0 2023-05-01 16:56:31,486 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9898, 4.2705, 4.0943, 4.1416, 3.8187, 3.8288, 3.9115, 4.2634], device='cuda:3'), covar=tensor([0.1205, 0.0972, 0.1082, 0.0830, 0.0938, 0.1853, 0.1035, 0.1101], device='cuda:3'), in_proj_covar=tensor([0.0670, 0.0813, 0.0675, 0.0624, 0.0517, 0.0529, 0.0685, 0.0642], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:57:11,260 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 8350, loss[loss=0.2051, simple_loss=0.3005, pruned_loss=0.05486, over 15461.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.28, pruned_loss=0.0499, over 3056146.43 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:57:54,072 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5766, 4.5318, 4.2987, 3.4133, 4.4532, 1.5663, 4.1418, 4.0914], device='cuda:3'), covar=tensor([0.0124, 0.0125, 0.0245, 0.0495, 0.0131, 0.3313, 0.0172, 0.0298], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0158, 0.0200, 0.0177, 0.0175, 0.0208, 0.0188, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 16:57:54,122 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4113, 3.5307, 2.1939, 3.9021, 2.7134, 3.8504, 2.2843, 2.8528], device='cuda:3'), covar=tensor([0.0308, 0.0341, 0.1549, 0.0242, 0.0796, 0.0563, 0.1501, 0.0754], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0174, 0.0192, 0.0162, 0.0175, 0.0215, 0.0202, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:57:55,454 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7982, 2.9517, 2.5747, 4.2114, 2.7100, 4.1145, 1.5442, 3.0997], device='cuda:3'), covar=tensor([0.1413, 0.0697, 0.1143, 0.0184, 0.0135, 0.0353, 0.1749, 0.0671], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0174, 0.0196, 0.0190, 0.0206, 0.0214, 0.0203, 0.0193], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 16:58:41,947 INFO [train.py:904] (3/8) Epoch 23, batch 8400, loss[loss=0.171, simple_loss=0.2658, pruned_loss=0.03808, over 16288.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2772, pruned_loss=0.04761, over 3064828.72 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:58:46,530 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 16:58:50,850 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231708.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:59:06,587 INFO [optim.py:368] (3/8) 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 16:59:44,693 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7410, 4.6911, 4.5113, 3.9273, 4.5819, 1.9617, 4.3581, 4.4306], device='cuda:3'), covar=tensor([0.0118, 0.0134, 0.0230, 0.0423, 0.0135, 0.2752, 0.0162, 0.0233], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0159, 0.0200, 0.0177, 0.0176, 0.0208, 0.0188, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:00:04,852 INFO [train.py:904] (3/8) Epoch 23, batch 8450, loss[loss=0.1712, simple_loss=0.2587, pruned_loss=0.04191, over 12458.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2752, pruned_loss=0.04566, over 3066232.98 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:24,902 INFO [train.py:904] (3/8) Epoch 23, batch 8500, loss[loss=0.1694, simple_loss=0.2629, pruned_loss=0.03797, over 16715.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2716, pruned_loss=0.04367, over 3036740.13 frames. ], batch size: 124, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:48,516 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 2.174e+02 2.671e+02 3.276e+02 7.093e+02, threshold=5.341e+02, percent-clipped=2.0 2023-05-01 17:02:48,223 INFO [train.py:904] (3/8) Epoch 23, batch 8550, loss[loss=0.1719, simple_loss=0.2751, pruned_loss=0.03431, over 16822.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2694, pruned_loss=0.04275, over 3032364.00 frames. ], batch size: 102, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:02:58,123 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6571, 2.6078, 1.8426, 2.7954, 2.0835, 2.7947, 2.1614, 2.4323], device='cuda:3'), covar=tensor([0.0301, 0.0306, 0.1356, 0.0266, 0.0696, 0.0406, 0.1250, 0.0573], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0173, 0.0192, 0.0161, 0.0175, 0.0213, 0.0201, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 17:04:28,271 INFO [train.py:904] (3/8) Epoch 23, batch 8600, loss[loss=0.1713, simple_loss=0.2597, pruned_loss=0.04142, over 12382.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2692, pruned_loss=0.04179, over 3024687.33 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:04:36,647 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:05:00,379 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.309e+02 2.625e+02 3.302e+02 5.565e+02, threshold=5.250e+02, percent-clipped=1.0 2023-05-01 17:05:07,184 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 17:06:03,590 INFO [train.py:904] (3/8) Epoch 23, batch 8650, loss[loss=0.1883, simple_loss=0.2839, pruned_loss=0.04635, over 16858.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2677, pruned_loss=0.04071, over 3022690.61 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:06:10,288 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:07:40,220 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6386, 2.9564, 3.3222, 2.0358, 2.8431, 2.1799, 3.2217, 3.2530], device='cuda:3'), covar=tensor([0.0271, 0.0931, 0.0482, 0.2003, 0.0813, 0.1008, 0.0624, 0.0911], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 17:07:53,210 INFO [train.py:904] (3/8) Epoch 23, batch 8700, loss[loss=0.1654, simple_loss=0.2615, pruned_loss=0.03466, over 16513.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2651, pruned_loss=0.03958, over 3031393.18 frames. ], batch size: 68, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:07:54,713 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232003.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:08:23,061 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 2.142e+02 2.520e+02 3.226e+02 6.058e+02, threshold=5.039e+02, percent-clipped=2.0 2023-05-01 17:09:27,791 INFO [train.py:904] (3/8) Epoch 23, batch 8750, loss[loss=0.1722, simple_loss=0.272, pruned_loss=0.0362, over 16685.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2649, pruned_loss=0.03913, over 3028190.69 frames. ], batch size: 76, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:09:42,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9434, 4.4005, 3.2086, 2.4540, 2.8805, 2.6874, 4.7226, 3.6558], device='cuda:3'), covar=tensor([0.2962, 0.0543, 0.1881, 0.3244, 0.2854, 0.2158, 0.0390, 0.1422], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0263, 0.0302, 0.0311, 0.0292, 0.0259, 0.0293, 0.0333], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 17:11:15,140 INFO [train.py:904] (3/8) Epoch 23, batch 8800, loss[loss=0.1715, simple_loss=0.2634, pruned_loss=0.03982, over 12444.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2632, pruned_loss=0.03783, over 3041175.81 frames. ], batch size: 249, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:11:29,658 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2827, 5.5933, 5.3658, 5.3914, 5.0964, 4.9992, 5.0410, 5.6807], device='cuda:3'), covar=tensor([0.1087, 0.0859, 0.0865, 0.0741, 0.0756, 0.0794, 0.1141, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0666, 0.0810, 0.0670, 0.0620, 0.0513, 0.0527, 0.0681, 0.0636], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:11:34,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8402, 3.7520, 3.9907, 3.7898, 3.9355, 4.3081, 3.9025, 3.6122], device='cuda:3'), covar=tensor([0.2175, 0.2168, 0.1961, 0.2257, 0.2732, 0.1705, 0.1649, 0.2698], device='cuda:3'), in_proj_covar=tensor([0.0398, 0.0578, 0.0639, 0.0477, 0.0634, 0.0666, 0.0500, 0.0640], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 17:11:46,651 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.145e+02 2.530e+02 3.020e+02 7.216e+02, threshold=5.060e+02, percent-clipped=1.0 2023-05-01 17:12:57,919 INFO [train.py:904] (3/8) Epoch 23, batch 8850, loss[loss=0.1745, simple_loss=0.2866, pruned_loss=0.03116, over 16852.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2654, pruned_loss=0.03699, over 3035995.18 frames. ], batch size: 76, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:14:42,072 INFO [train.py:904] (3/8) Epoch 23, batch 8900, loss[loss=0.1754, simple_loss=0.273, pruned_loss=0.0389, over 16361.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2664, pruned_loss=0.03681, over 3041150.27 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:15:12,043 INFO [optim.py:368] (3/8) 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 17:16:33,765 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 8950, loss[loss=0.1661, simple_loss=0.2616, pruned_loss=0.03536, over 16718.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2663, pruned_loss=0.03728, over 3040660.99 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:17:01,547 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4003, 3.0507, 2.7112, 2.2527, 2.1887, 2.2921, 3.0613, 2.8430], device='cuda:3'), covar=tensor([0.2687, 0.0704, 0.1676, 0.2925, 0.2430, 0.2156, 0.0527, 0.1433], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0262, 0.0301, 0.0310, 0.0290, 0.0258, 0.0292, 0.0332], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 17:18:00,118 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0443, 4.7882, 4.6129, 5.1555, 5.2917, 4.8443, 5.3179, 5.2908], device='cuda:3'), covar=tensor([0.1801, 0.1469, 0.3228, 0.1185, 0.0900, 0.1102, 0.0919, 0.1478], device='cuda:3'), in_proj_covar=tensor([0.0617, 0.0763, 0.0877, 0.0773, 0.0587, 0.0615, 0.0641, 0.0747], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:18:31,849 INFO [train.py:904] (3/8) Epoch 23, batch 9000, loss[loss=0.1582, simple_loss=0.2486, pruned_loss=0.03389, over 11944.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2631, pruned_loss=0.0358, over 3058166.29 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:31,850 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 17:18:42,672 INFO [train.py:938] (3/8) Epoch 23, validation: loss=0.1452, simple_loss=0.249, pruned_loss=0.02066, over 944034.00 frames. 2023-05-01 17:18:42,673 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 17:18:44,246 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232303.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:18:54,222 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232308.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:19:18,069 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 1.963e+02 2.332e+02 2.758e+02 5.487e+02, threshold=4.665e+02, percent-clipped=1.0 2023-05-01 17:19:21,127 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 17:19:49,327 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 17:20:22,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4961, 4.4561, 4.3385, 3.6524, 4.3624, 1.7256, 4.1108, 4.1202], device='cuda:3'), covar=tensor([0.0153, 0.0140, 0.0230, 0.0423, 0.0160, 0.2968, 0.0209, 0.0306], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0157, 0.0197, 0.0174, 0.0174, 0.0206, 0.0186, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:20:24,509 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232351.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:20:28,118 INFO [train.py:904] (3/8) Epoch 23, batch 9050, loss[loss=0.1546, simple_loss=0.2473, pruned_loss=0.03091, over 16870.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2643, pruned_loss=0.03639, over 3079629.37 frames. ], batch size: 96, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:20:41,775 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 17:21:23,464 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:22:12,815 INFO [train.py:904] (3/8) Epoch 23, batch 9100, loss[loss=0.1661, simple_loss=0.2559, pruned_loss=0.03811, over 12668.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2638, pruned_loss=0.03696, over 3080795.91 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:22:23,264 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8599, 5.1179, 5.3112, 5.0898, 5.1455, 5.6810, 5.1729, 4.8635], device='cuda:3'), covar=tensor([0.1045, 0.1839, 0.1819, 0.1819, 0.2235, 0.0895, 0.1651, 0.2424], device='cuda:3'), in_proj_covar=tensor([0.0396, 0.0577, 0.0638, 0.0477, 0.0634, 0.0662, 0.0498, 0.0639], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 17:22:46,300 INFO [optim.py:368] (3/8) 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 17:23:46,401 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232442.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:24:09,371 INFO [train.py:904] (3/8) Epoch 23, batch 9150, loss[loss=0.1668, simple_loss=0.2577, pruned_loss=0.038, over 16202.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2645, pruned_loss=0.0366, over 3092713.92 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:25:00,449 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:25:09,189 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0883, 2.3081, 1.9931, 2.1677, 2.6461, 2.3270, 2.5276, 2.7991], device='cuda:3'), covar=tensor([0.0153, 0.0498, 0.0557, 0.0515, 0.0326, 0.0454, 0.0214, 0.0301], device='cuda:3'), in_proj_covar=tensor([0.0204, 0.0228, 0.0220, 0.0222, 0.0229, 0.0229, 0.0225, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:25:52,924 INFO [train.py:904] (3/8) Epoch 23, batch 9200, loss[loss=0.1504, simple_loss=0.233, pruned_loss=0.03387, over 12157.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2606, pruned_loss=0.03611, over 3081013.54 frames. ], batch size: 250, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:25:58,391 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 17:26:21,342 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:26:24,112 INFO [optim.py:368] (3/8) 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 17:26:54,747 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8721, 3.0597, 3.3197, 1.6842, 2.8309, 1.9900, 3.2854, 3.3745], device='cuda:3'), covar=tensor([0.0238, 0.0925, 0.0562, 0.2544, 0.0981, 0.1262, 0.0676, 0.0980], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0159, 0.0164, 0.0151, 0.0143, 0.0127, 0.0140, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 17:26:57,030 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:27:17,345 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 17:27:27,732 INFO [train.py:904] (3/8) Epoch 23, batch 9250, loss[loss=0.1617, simple_loss=0.2583, pruned_loss=0.03252, over 16344.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2604, pruned_loss=0.03593, over 3090928.73 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:28:23,257 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232579.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:17,461 INFO [train.py:904] (3/8) Epoch 23, batch 9300, loss[loss=0.1734, simple_loss=0.2673, pruned_loss=0.03979, over 16652.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2585, pruned_loss=0.03552, over 3068354.66 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:29:18,375 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232603.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:22,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0037, 3.8941, 4.0997, 4.1772, 4.2789, 3.8382, 4.2404, 4.3240], device='cuda:3'), covar=tensor([0.1677, 0.1251, 0.1193, 0.0677, 0.0534, 0.1474, 0.0744, 0.0662], device='cuda:3'), in_proj_covar=tensor([0.0612, 0.0759, 0.0870, 0.0770, 0.0584, 0.0611, 0.0637, 0.0743], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:29:58,634 INFO [optim.py:368] (3/8) 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 17:30:18,016 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4023, 2.2663, 2.2430, 4.0936, 2.1547, 2.6312, 2.3651, 2.4058], device='cuda:3'), covar=tensor([0.1226, 0.3751, 0.3373, 0.0497, 0.4484, 0.2742, 0.4035, 0.3654], device='cuda:3'), in_proj_covar=tensor([0.0394, 0.0443, 0.0365, 0.0318, 0.0429, 0.0507, 0.0414, 0.0515], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:30:44,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5670, 3.0384, 3.2603, 2.0042, 2.8265, 2.1779, 3.1118, 3.2812], device='cuda:3'), covar=tensor([0.0397, 0.0889, 0.0598, 0.2246, 0.0957, 0.1133, 0.0808, 0.1047], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0158, 0.0163, 0.0150, 0.0142, 0.0126, 0.0139, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 17:30:46,729 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-05-01 17:31:04,804 INFO [train.py:904] (3/8) Epoch 23, batch 9350, loss[loss=0.1729, simple_loss=0.2649, pruned_loss=0.04042, over 15454.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2581, pruned_loss=0.0354, over 3078214.44 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:32:12,836 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 17:32:47,940 INFO [train.py:904] (3/8) Epoch 23, batch 9400, loss[loss=0.1623, simple_loss=0.2711, pruned_loss=0.02674, over 16880.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2583, pruned_loss=0.03506, over 3074500.56 frames. ], batch size: 96, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:33:21,844 INFO [optim.py:368] (3/8) 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 17:33:50,161 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-01 17:33:57,944 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232737.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:34:29,649 INFO [train.py:904] (3/8) Epoch 23, batch 9450, loss[loss=0.1574, simple_loss=0.2586, pruned_loss=0.0281, over 16533.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2595, pruned_loss=0.03493, over 3069865.19 frames. ], batch size: 75, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:35:07,242 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 17:35:41,265 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5884, 4.8733, 4.6931, 4.6909, 4.4067, 4.3352, 4.3670, 4.9193], device='cuda:3'), covar=tensor([0.1140, 0.0926, 0.0871, 0.0797, 0.0730, 0.1365, 0.1041, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0802, 0.0664, 0.0616, 0.0509, 0.0523, 0.0675, 0.0635], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:36:09,229 INFO [train.py:904] (3/8) Epoch 23, batch 9500, loss[loss=0.1651, simple_loss=0.2573, pruned_loss=0.03643, over 16724.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.259, pruned_loss=0.03472, over 3070068.58 frames. ], batch size: 76, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:44,543 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.194e+02 2.525e+02 3.286e+02 6.197e+02, threshold=5.051e+02, percent-clipped=6.0 2023-05-01 17:37:08,415 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 23, batch 9550, loss[loss=0.1759, simple_loss=0.2608, pruned_loss=0.04554, over 12376.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2584, pruned_loss=0.03493, over 3050051.01 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:38:01,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0176, 3.8836, 4.0900, 4.1656, 4.2732, 3.8749, 4.2857, 4.3095], device='cuda:3'), covar=tensor([0.1531, 0.1133, 0.1213, 0.0685, 0.0526, 0.1521, 0.0629, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0612, 0.0756, 0.0868, 0.0766, 0.0583, 0.0608, 0.0635, 0.0741], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:38:09,354 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 17:38:37,958 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:38:59,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8662, 1.4627, 1.7579, 1.7353, 1.8808, 1.8962, 1.6639, 1.8169], device='cuda:3'), covar=tensor([0.0266, 0.0434, 0.0226, 0.0346, 0.0318, 0.0234, 0.0422, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0188, 0.0175, 0.0178, 0.0193, 0.0150, 0.0192, 0.0148], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:39:31,180 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2606, 3.4245, 2.0201, 3.7888, 2.4920, 3.7061, 2.2379, 2.7396], device='cuda:3'), covar=tensor([0.0370, 0.0435, 0.1807, 0.0310, 0.0941, 0.0768, 0.1627, 0.0880], device='cuda:3'), in_proj_covar=tensor([0.0164, 0.0169, 0.0187, 0.0157, 0.0170, 0.0206, 0.0198, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 17:39:33,897 INFO [train.py:904] (3/8) Epoch 23, batch 9600, loss[loss=0.1718, simple_loss=0.2746, pruned_loss=0.03447, over 16707.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2599, pruned_loss=0.03564, over 3045420.98 frames. ], batch size: 134, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:39:34,574 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:41:17,899 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 17:41:20,221 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-01 17:41:20,697 INFO [train.py:904] (3/8) Epoch 23, batch 9650, loss[loss=0.1682, simple_loss=0.2623, pruned_loss=0.03702, over 16684.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2623, pruned_loss=0.03609, over 3051648.38 frames. ], batch size: 134, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:42:00,627 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0178, 2.1541, 2.2989, 3.5021, 2.1105, 2.4306, 2.3146, 2.2554], device='cuda:3'), covar=tensor([0.1333, 0.3506, 0.3009, 0.0631, 0.4440, 0.2630, 0.3553, 0.3522], device='cuda:3'), in_proj_covar=tensor([0.0393, 0.0442, 0.0364, 0.0317, 0.0427, 0.0504, 0.0414, 0.0513], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:42:24,277 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9730, 2.6843, 2.8883, 2.0202, 2.6891, 2.1432, 2.6148, 2.8265], device='cuda:3'), covar=tensor([0.0353, 0.0946, 0.0490, 0.1957, 0.0824, 0.0957, 0.0683, 0.0929], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0157, 0.0163, 0.0150, 0.0141, 0.0126, 0.0139, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 17:42:41,934 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:09,180 INFO [train.py:904] (3/8) Epoch 23, batch 9700, loss[loss=0.1919, simple_loss=0.28, pruned_loss=0.05186, over 16939.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2617, pruned_loss=0.03622, over 3049626.88 frames. ], batch size: 109, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:43:40,163 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:43,780 INFO [optim.py:368] (3/8) 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 17:44:22,525 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:44:49,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8918, 5.0810, 5.2327, 5.0067, 5.0754, 5.6224, 5.1307, 4.8484], device='cuda:3'), covar=tensor([0.0969, 0.2039, 0.2436, 0.1901, 0.2484, 0.0909, 0.1776, 0.2576], device='cuda:3'), in_proj_covar=tensor([0.0388, 0.0570, 0.0630, 0.0470, 0.0627, 0.0656, 0.0492, 0.0626], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 17:44:50,793 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233051.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:44:53,543 INFO [train.py:904] (3/8) Epoch 23, batch 9750, loss[loss=0.1425, simple_loss=0.2451, pruned_loss=0.02001, over 16555.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2603, pruned_loss=0.0361, over 3057354.32 frames. ], batch size: 75, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:45:43,932 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:45:59,982 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233085.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:46:21,780 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1129, 5.4381, 5.2217, 5.2451, 4.9417, 4.9391, 4.7764, 5.5346], device='cuda:3'), covar=tensor([0.1225, 0.0833, 0.0937, 0.0756, 0.0718, 0.0862, 0.1159, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0662, 0.0801, 0.0661, 0.0613, 0.0509, 0.0520, 0.0675, 0.0632], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:46:31,034 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6384, 4.0444, 2.9173, 2.2004, 2.5277, 2.5679, 4.2550, 3.4704], device='cuda:3'), covar=tensor([0.3111, 0.0515, 0.1842, 0.3051, 0.2786, 0.2093, 0.0409, 0.1231], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0262, 0.0300, 0.0309, 0.0286, 0.0257, 0.0290, 0.0329], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 17:46:31,579 INFO [train.py:904] (3/8) Epoch 23, batch 9800, loss[loss=0.1669, simple_loss=0.2712, pruned_loss=0.03128, over 16751.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2612, pruned_loss=0.03548, over 3070048.84 frames. ], batch size: 83, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:47:03,692 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.119e+02 2.583e+02 3.356e+02 7.260e+02, threshold=5.167e+02, percent-clipped=1.0 2023-05-01 17:47:26,900 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:48:00,953 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 17:48:16,806 INFO [train.py:904] (3/8) Epoch 23, batch 9850, loss[loss=0.1387, simple_loss=0.2316, pruned_loss=0.02292, over 12403.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2618, pruned_loss=0.03525, over 3050226.75 frames. ], batch size: 247, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:49:01,106 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233174.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:13,851 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233180.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:53,828 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-01 17:50:08,889 INFO [train.py:904] (3/8) Epoch 23, batch 9900, loss[loss=0.1626, simple_loss=0.2523, pruned_loss=0.03641, over 12750.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2625, pruned_loss=0.03543, over 3040497.59 frames. ], batch size: 248, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:50:46,347 INFO [optim.py:368] (3/8) 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 17:50:52,784 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:52:06,030 INFO [train.py:904] (3/8) Epoch 23, batch 9950, loss[loss=0.161, simple_loss=0.263, pruned_loss=0.02953, over 16230.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2643, pruned_loss=0.03572, over 3051034.63 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:52:20,251 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:53:30,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2595, 3.0041, 3.1996, 1.7895, 3.2904, 3.3881, 2.8451, 2.6703], device='cuda:3'), covar=tensor([0.0775, 0.0269, 0.0169, 0.1255, 0.0089, 0.0181, 0.0384, 0.0482], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0103, 0.0093, 0.0134, 0.0078, 0.0119, 0.0122, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 17:53:38,941 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4116, 2.3960, 2.4321, 4.2550, 2.2736, 2.7281, 2.4437, 2.5338], device='cuda:3'), covar=tensor([0.1182, 0.3468, 0.2969, 0.0442, 0.3954, 0.2464, 0.3852, 0.3069], device='cuda:3'), in_proj_covar=tensor([0.0395, 0.0443, 0.0365, 0.0318, 0.0430, 0.0505, 0.0415, 0.0515], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:54:07,453 INFO [train.py:904] (3/8) Epoch 23, batch 10000, loss[loss=0.1649, simple_loss=0.2667, pruned_loss=0.03153, over 16288.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2626, pruned_loss=0.03519, over 3067112.41 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:54:40,344 INFO [optim.py:368] (3/8) 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 17:54:41,046 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:55:20,861 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5893, 2.6129, 2.3463, 2.3713, 2.9226, 2.6615, 3.0541, 3.1351], device='cuda:3'), covar=tensor([0.0156, 0.0444, 0.0501, 0.0500, 0.0328, 0.0406, 0.0308, 0.0306], device='cuda:3'), in_proj_covar=tensor([0.0204, 0.0229, 0.0221, 0.0223, 0.0229, 0.0230, 0.0223, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:55:32,358 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2368, 2.2731, 2.2491, 4.0095, 2.0874, 2.6058, 2.3721, 2.4567], device='cuda:3'), covar=tensor([0.1261, 0.3627, 0.3094, 0.0477, 0.4357, 0.2643, 0.3523, 0.3420], device='cuda:3'), in_proj_covar=tensor([0.0392, 0.0440, 0.0363, 0.0315, 0.0427, 0.0502, 0.0412, 0.0512], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:55:36,787 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233346.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:55:47,981 INFO [train.py:904] (3/8) Epoch 23, batch 10050, loss[loss=0.1689, simple_loss=0.2664, pruned_loss=0.03575, over 16499.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2626, pruned_loss=0.03514, over 3066836.80 frames. ], batch size: 68, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:56:31,051 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:57:20,832 INFO [train.py:904] (3/8) Epoch 23, batch 10100, loss[loss=0.1716, simple_loss=0.265, pruned_loss=0.03915, over 17040.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2628, pruned_loss=0.03524, over 3065157.37 frames. ], batch size: 53, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:57:41,850 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 17:57:53,979 INFO [optim.py:368] (3/8) 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 17:58:29,623 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3771, 2.4210, 2.0559, 2.1947, 2.8012, 2.4316, 2.9312, 2.9561], device='cuda:3'), covar=tensor([0.0161, 0.0495, 0.0572, 0.0538, 0.0339, 0.0478, 0.0262, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0204, 0.0230, 0.0221, 0.0223, 0.0229, 0.0230, 0.0224, 0.0223], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 17:59:06,335 INFO [train.py:904] (3/8) Epoch 24, batch 0, loss[loss=0.1619, simple_loss=0.2451, pruned_loss=0.03938, over 16762.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2451, pruned_loss=0.03938, over 16762.00 frames. ], batch size: 39, lr: 2.84e-03, grad_scale: 8.0 2023-05-01 17:59:06,335 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 17:59:14,248 INFO [train.py:938] (3/8) Epoch 24, validation: loss=0.145, simple_loss=0.2483, pruned_loss=0.02085, over 944034.00 frames. 2023-05-01 17:59:14,248 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 18:00:06,183 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6942, 2.6502, 2.4462, 2.5242, 3.0053, 2.7684, 3.3001, 3.1600], device='cuda:3'), covar=tensor([0.0178, 0.0527, 0.0557, 0.0536, 0.0349, 0.0464, 0.0320, 0.0368], device='cuda:3'), in_proj_covar=tensor([0.0207, 0.0231, 0.0222, 0.0224, 0.0230, 0.0231, 0.0225, 0.0224], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:00:23,695 INFO [train.py:904] (3/8) Epoch 24, batch 50, loss[loss=0.191, simple_loss=0.279, pruned_loss=0.05154, over 16887.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2689, pruned_loss=0.04858, over 746585.70 frames. ], batch size: 96, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:00:52,607 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 100, loss[loss=0.145, simple_loss=0.228, pruned_loss=0.03099, over 16808.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2634, pruned_loss=0.04573, over 1320670.91 frames. ], batch size: 39, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:02:40,657 INFO [train.py:904] (3/8) Epoch 24, batch 150, loss[loss=0.17, simple_loss=0.2468, pruned_loss=0.04665, over 16581.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2623, pruned_loss=0.04409, over 1767084.65 frames. ], batch size: 75, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:02:56,016 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:08,348 INFO [optim.py:368] (3/8) 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 18:03:30,589 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:32,743 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 18:03:39,295 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233646.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:03:40,853 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 18:03:48,630 INFO [train.py:904] (3/8) Epoch 24, batch 200, loss[loss=0.1488, simple_loss=0.2354, pruned_loss=0.03111, over 16969.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2636, pruned_loss=0.04436, over 2111773.36 frames. ], batch size: 41, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:03:56,693 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-05-01 18:04:11,936 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 18:04:15,564 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2042, 5.8240, 5.9396, 5.6205, 5.7875, 6.2595, 5.7874, 5.5436], device='cuda:3'), covar=tensor([0.0905, 0.1894, 0.2412, 0.2120, 0.2519, 0.0908, 0.1471, 0.2169], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0589, 0.0652, 0.0487, 0.0649, 0.0679, 0.0507, 0.0649], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 18:04:17,904 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:45,850 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:54,285 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233700.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 18:04:58,754 INFO [train.py:904] (3/8) Epoch 24, batch 250, loss[loss=0.1697, simple_loss=0.2611, pruned_loss=0.03918, over 17037.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2614, pruned_loss=0.04299, over 2378651.30 frames. ], batch size: 55, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:05:07,766 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233710.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:25,966 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.284e+02 2.752e+02 3.292e+02 5.130e+02, threshold=5.503e+02, percent-clipped=0.0 2023-05-01 18:06:08,092 INFO [train.py:904] (3/8) Epoch 24, batch 300, loss[loss=0.1701, simple_loss=0.249, pruned_loss=0.04562, over 12493.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2584, pruned_loss=0.0421, over 2583477.08 frames. ], batch size: 247, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:06:33,476 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233771.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:06:54,252 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7470, 4.5336, 4.8189, 4.9694, 5.1443, 4.6318, 5.1281, 5.1444], device='cuda:3'), covar=tensor([0.2017, 0.1265, 0.1638, 0.0794, 0.0573, 0.0867, 0.0812, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0638, 0.0785, 0.0905, 0.0795, 0.0604, 0.0627, 0.0662, 0.0765], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:07:16,411 INFO [train.py:904] (3/8) Epoch 24, batch 350, loss[loss=0.1391, simple_loss=0.2247, pruned_loss=0.02676, over 16963.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2551, pruned_loss=0.0409, over 2750242.15 frames. ], batch size: 41, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:07:43,747 INFO [optim.py:368] (3/8) 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 18:08:25,444 INFO [train.py:904] (3/8) Epoch 24, batch 400, loss[loss=0.1418, simple_loss=0.2284, pruned_loss=0.0276, over 16955.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.253, pruned_loss=0.04019, over 2880540.57 frames. ], batch size: 41, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:08,211 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4993, 5.9060, 5.6768, 5.6970, 5.2972, 5.3769, 5.3053, 6.0356], device='cuda:3'), covar=tensor([0.1542, 0.1083, 0.1077, 0.0917, 0.0963, 0.0724, 0.1354, 0.1019], device='cuda:3'), in_proj_covar=tensor([0.0681, 0.0825, 0.0680, 0.0633, 0.0524, 0.0532, 0.0698, 0.0650], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:09:32,974 INFO [train.py:904] (3/8) Epoch 24, batch 450, loss[loss=0.165, simple_loss=0.2461, pruned_loss=0.04194, over 16450.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.252, pruned_loss=0.0396, over 2985580.45 frames. ], batch size: 75, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:47,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2781, 4.0398, 4.5640, 2.5913, 4.7311, 4.8106, 3.5310, 3.6466], device='cuda:3'), covar=tensor([0.0705, 0.0235, 0.0192, 0.1062, 0.0082, 0.0171, 0.0405, 0.0419], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0107, 0.0097, 0.0139, 0.0081, 0.0125, 0.0126, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 18:09:50,191 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:09:59,638 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 500, loss[loss=0.1582, simple_loss=0.2357, pruned_loss=0.04031, over 15360.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2508, pruned_loss=0.03906, over 3059655.87 frames. ], batch size: 190, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:10:48,787 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 18:10:53,521 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233963.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:11:20,699 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 18:11:22,601 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5281, 4.0040, 3.6105, 3.9211, 3.4233, 3.6370, 3.6220, 4.0333], device='cuda:3'), covar=tensor([0.3510, 0.2073, 0.2806, 0.1710, 0.2260, 0.3374, 0.2717, 0.2266], device='cuda:3'), in_proj_covar=tensor([0.0688, 0.0833, 0.0687, 0.0639, 0.0529, 0.0537, 0.0705, 0.0657], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:11:37,608 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233995.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:11:50,840 INFO [train.py:904] (3/8) Epoch 24, batch 550, loss[loss=0.1384, simple_loss=0.2191, pruned_loss=0.02881, over 15919.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2502, pruned_loss=0.0389, over 3111883.34 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:12:13,000 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4959, 2.4022, 2.4088, 4.2897, 2.3474, 2.7692, 2.5149, 2.5292], device='cuda:3'), covar=tensor([0.1332, 0.3829, 0.3077, 0.0528, 0.4221, 0.2664, 0.3524, 0.3907], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0454, 0.0373, 0.0326, 0.0438, 0.0517, 0.0425, 0.0529], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:12:17,095 INFO [optim.py:368] (3/8) 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 18:12:20,445 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:12:57,925 INFO [train.py:904] (3/8) Epoch 24, batch 600, loss[loss=0.1851, simple_loss=0.2621, pruned_loss=0.05406, over 16385.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.25, pruned_loss=0.0396, over 3162299.59 frames. ], batch size: 145, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:13:17,435 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234066.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:13:34,156 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7093, 2.6290, 2.4127, 2.5377, 2.9729, 2.8005, 3.2479, 3.1543], device='cuda:3'), covar=tensor([0.0178, 0.0505, 0.0527, 0.0486, 0.0328, 0.0408, 0.0294, 0.0327], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0239, 0.0229, 0.0230, 0.0238, 0.0239, 0.0236, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:13:45,356 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234086.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:14:08,032 INFO [train.py:904] (3/8) Epoch 24, batch 650, loss[loss=0.156, simple_loss=0.2567, pruned_loss=0.02764, over 17266.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2493, pruned_loss=0.03929, over 3197275.06 frames. ], batch size: 52, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:14:14,523 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3169, 6.0194, 6.1040, 5.8539, 5.8987, 6.4100, 6.0046, 5.6524], device='cuda:3'), covar=tensor([0.0936, 0.1885, 0.2392, 0.1848, 0.2313, 0.0829, 0.1543, 0.2231], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0603, 0.0667, 0.0499, 0.0667, 0.0698, 0.0520, 0.0666], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 18:14:36,190 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 700, loss[loss=0.1686, simple_loss=0.2435, pruned_loss=0.04682, over 16776.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2497, pruned_loss=0.03931, over 3222211.97 frames. ], batch size: 83, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:00,838 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8452, 4.6132, 4.9289, 5.0970, 5.3063, 4.6573, 5.2746, 5.2872], device='cuda:3'), covar=tensor([0.2026, 0.1405, 0.1791, 0.0829, 0.0586, 0.0964, 0.0745, 0.0684], device='cuda:3'), in_proj_covar=tensor([0.0655, 0.0808, 0.0929, 0.0815, 0.0620, 0.0644, 0.0678, 0.0786], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:16:23,753 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:16:24,493 INFO [train.py:904] (3/8) Epoch 24, batch 750, loss[loss=0.1677, simple_loss=0.2505, pruned_loss=0.04249, over 16782.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2498, pruned_loss=0.03964, over 3250581.40 frames. ], batch size: 89, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:31,194 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 18:16:52,425 INFO [optim.py:368] (3/8) 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 18:17:04,990 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234231.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:17:33,979 INFO [train.py:904] (3/8) Epoch 24, batch 800, loss[loss=0.1505, simple_loss=0.2376, pruned_loss=0.03167, over 17221.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2495, pruned_loss=0.03918, over 3270301.39 frames. ], batch size: 46, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:17:48,300 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234263.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:18:29,129 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234292.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:18:33,116 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234295.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:18:43,353 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 18:18:43,642 INFO [train.py:904] (3/8) Epoch 24, batch 850, loss[loss=0.1544, simple_loss=0.2523, pruned_loss=0.0282, over 17131.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2498, pruned_loss=0.03874, over 3275380.24 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:18:54,133 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9867, 2.1523, 2.5205, 2.9504, 2.7466, 3.4145, 2.3147, 3.4063], device='cuda:3'), covar=tensor([0.0261, 0.0550, 0.0394, 0.0326, 0.0355, 0.0203, 0.0536, 0.0184], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0185, 0.0200, 0.0157, 0.0198, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:19:11,880 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.010e+02 2.479e+02 2.918e+02 4.237e+02, threshold=4.958e+02, percent-clipped=0.0 2023-05-01 18:19:19,301 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0045, 5.3392, 5.4966, 5.2776, 5.2845, 5.9302, 5.4457, 5.0452], device='cuda:3'), covar=tensor([0.1102, 0.2191, 0.2719, 0.2036, 0.2974, 0.1063, 0.1571, 0.2491], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0606, 0.0672, 0.0502, 0.0670, 0.0704, 0.0523, 0.0669], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 18:19:38,813 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2093, 5.1903, 5.0950, 4.5871, 4.7364, 5.1040, 5.0654, 4.7509], device='cuda:3'), covar=tensor([0.0642, 0.0541, 0.0324, 0.0359, 0.1120, 0.0485, 0.0369, 0.0757], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0444, 0.0351, 0.0350, 0.0353, 0.0406, 0.0238, 0.0420], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:19:40,584 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:19:50,936 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 18:19:52,507 INFO [train.py:904] (3/8) Epoch 24, batch 900, loss[loss=0.1651, simple_loss=0.256, pruned_loss=0.03712, over 17132.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.249, pruned_loss=0.03831, over 3282159.27 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:20:11,536 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:20:42,322 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 18:21:00,724 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.70 vs. limit=5.0 2023-05-01 18:21:02,818 INFO [train.py:904] (3/8) Epoch 24, batch 950, loss[loss=0.1549, simple_loss=0.2409, pruned_loss=0.03441, over 15894.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2494, pruned_loss=0.0389, over 3294634.29 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:21:17,243 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:27,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-05-01 18:21:30,309 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.144e+02 2.493e+02 3.109e+02 1.071e+03, threshold=4.986e+02, percent-clipped=5.0 2023-05-01 18:22:10,607 INFO [train.py:904] (3/8) Epoch 24, batch 1000, loss[loss=0.1674, simple_loss=0.2377, pruned_loss=0.04854, over 16759.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2481, pruned_loss=0.0383, over 3305366.65 frames. ], batch size: 124, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:22:42,800 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 18:23:05,891 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6144, 3.8212, 4.0059, 2.8259, 3.5990, 4.0427, 3.7052, 2.4509], device='cuda:3'), covar=tensor([0.0549, 0.0236, 0.0064, 0.0415, 0.0126, 0.0114, 0.0115, 0.0478], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0087, 0.0087, 0.0136, 0.0100, 0.0111, 0.0096, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 18:23:22,529 INFO [train.py:904] (3/8) Epoch 24, batch 1050, loss[loss=0.1449, simple_loss=0.2279, pruned_loss=0.03099, over 12553.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2475, pruned_loss=0.03839, over 3298303.07 frames. ], batch size: 247, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:23:50,904 INFO [optim.py:368] (3/8) 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 18:24:23,485 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5521, 4.8358, 4.6626, 4.6198, 4.3784, 4.3516, 4.3699, 4.8963], device='cuda:3'), covar=tensor([0.1218, 0.0931, 0.1040, 0.0919, 0.0891, 0.1360, 0.1145, 0.1062], device='cuda:3'), in_proj_covar=tensor([0.0701, 0.0853, 0.0701, 0.0655, 0.0540, 0.0546, 0.0720, 0.0670], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:24:30,872 INFO [train.py:904] (3/8) Epoch 24, batch 1100, loss[loss=0.1602, simple_loss=0.2584, pruned_loss=0.03101, over 17148.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2468, pruned_loss=0.03819, over 3298050.13 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:24:37,556 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234558.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:25:18,104 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:25:34,379 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 18:25:38,658 INFO [train.py:904] (3/8) Epoch 24, batch 1150, loss[loss=0.1674, simple_loss=0.2393, pruned_loss=0.04774, over 16793.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2468, pruned_loss=0.03781, over 3302954.53 frames. ], batch size: 124, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:25:52,434 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 18:25:55,638 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3985, 2.3709, 2.5488, 4.1328, 2.4059, 2.7349, 2.4870, 2.4882], device='cuda:3'), covar=tensor([0.1471, 0.3689, 0.3061, 0.0680, 0.4072, 0.2655, 0.3813, 0.3446], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0458, 0.0377, 0.0331, 0.0441, 0.0524, 0.0429, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:26:06,113 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.089e+02 2.435e+02 2.920e+02 5.927e+02, threshold=4.869e+02, percent-clipped=1.0 2023-05-01 18:26:33,581 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 2023-05-01 18:26:46,784 INFO [train.py:904] (3/8) Epoch 24, batch 1200, loss[loss=0.1653, simple_loss=0.2443, pruned_loss=0.04321, over 12158.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.246, pruned_loss=0.03736, over 3295125.24 frames. ], batch size: 247, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:27:09,082 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:21,668 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5641, 4.4069, 4.6065, 4.7488, 4.8512, 4.4252, 4.7683, 4.8665], device='cuda:3'), covar=tensor([0.1886, 0.1398, 0.1454, 0.0844, 0.0658, 0.1089, 0.1986, 0.0995], device='cuda:3'), in_proj_covar=tensor([0.0667, 0.0822, 0.0950, 0.0832, 0.0635, 0.0656, 0.0689, 0.0801], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:27:25,008 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:26,430 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8442, 2.7731, 2.4960, 2.7895, 3.1278, 2.9272, 3.4909, 3.3611], device='cuda:3'), covar=tensor([0.0168, 0.0460, 0.0528, 0.0421, 0.0310, 0.0407, 0.0242, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0244, 0.0233, 0.0234, 0.0244, 0.0243, 0.0242, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:27:38,845 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0503, 4.3863, 4.4087, 3.2740, 3.6077, 4.3666, 3.9020, 2.7206], device='cuda:3'), covar=tensor([0.0452, 0.0081, 0.0055, 0.0358, 0.0155, 0.0118, 0.0103, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0087, 0.0088, 0.0136, 0.0100, 0.0112, 0.0097, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 18:27:53,727 INFO [train.py:904] (3/8) Epoch 24, batch 1250, loss[loss=0.1444, simple_loss=0.2308, pruned_loss=0.029, over 16796.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2454, pruned_loss=0.03751, over 3295268.19 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:28:20,471 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9161, 2.0741, 2.4546, 2.8950, 2.6975, 3.3683, 2.2680, 3.3166], device='cuda:3'), covar=tensor([0.0305, 0.0585, 0.0393, 0.0372, 0.0410, 0.0213, 0.0574, 0.0229], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0184, 0.0189, 0.0204, 0.0160, 0.0201, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:28:21,122 INFO [optim.py:368] (3/8) 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 18:28:25,243 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:30,423 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234729.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:31,676 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234730.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:28:48,925 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9536, 2.9254, 2.9179, 5.1126, 4.1970, 4.4417, 1.7240, 3.3555], device='cuda:3'), covar=tensor([0.1297, 0.0811, 0.1133, 0.0201, 0.0267, 0.0438, 0.1612, 0.0745], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0175, 0.0195, 0.0193, 0.0203, 0.0216, 0.0205, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 18:29:02,408 INFO [train.py:904] (3/8) Epoch 24, batch 1300, loss[loss=0.1633, simple_loss=0.2699, pruned_loss=0.02839, over 17102.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2457, pruned_loss=0.03758, over 3286793.18 frames. ], batch size: 49, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:29:41,564 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:29:47,315 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:30:10,001 INFO [train.py:904] (3/8) Epoch 24, batch 1350, loss[loss=0.1589, simple_loss=0.2466, pruned_loss=0.0356, over 17030.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2465, pruned_loss=0.03726, over 3298136.27 frames. ], batch size: 55, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:30:38,932 INFO [optim.py:368] (3/8) 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 18:31:07,891 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234843.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:19,629 INFO [train.py:904] (3/8) Epoch 24, batch 1400, loss[loss=0.1595, simple_loss=0.2672, pruned_loss=0.02588, over 17116.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2463, pruned_loss=0.03734, over 3297483.28 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:31:27,951 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:00,187 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:07,291 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 1450, loss[loss=0.1527, simple_loss=0.2375, pruned_loss=0.03398, over 16840.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2458, pruned_loss=0.03717, over 3307493.55 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:32:34,693 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:41,835 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7423, 3.8679, 2.3570, 4.4682, 2.9991, 4.3631, 2.6873, 3.1495], device='cuda:3'), covar=tensor([0.0345, 0.0400, 0.1683, 0.0343, 0.0873, 0.0557, 0.1367, 0.0774], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0170, 0.0179, 0.0221, 0.0206, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 18:32:58,962 INFO [optim.py:368] (3/8) 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 18:33:14,527 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:26,759 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234943.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:38,961 INFO [train.py:904] (3/8) Epoch 24, batch 1500, loss[loss=0.1626, simple_loss=0.2421, pruned_loss=0.04155, over 16809.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2459, pruned_loss=0.03768, over 3305394.93 frames. ], batch size: 96, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:33:45,551 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:49,638 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234960.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:34:49,722 INFO [train.py:904] (3/8) Epoch 24, batch 1550, loss[loss=0.1576, simple_loss=0.2545, pruned_loss=0.03039, over 17135.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.247, pruned_loss=0.03867, over 3308623.13 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:35:12,980 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235019.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:15,775 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:16,770 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3722, 4.6623, 4.5134, 4.4659, 4.2382, 4.1809, 4.2268, 4.7418], device='cuda:3'), covar=tensor([0.1298, 0.0956, 0.1042, 0.0866, 0.0774, 0.1571, 0.1087, 0.0794], device='cuda:3'), in_proj_covar=tensor([0.0708, 0.0865, 0.0708, 0.0662, 0.0547, 0.0553, 0.0726, 0.0675], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:35:18,855 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.608e+02 2.203e+02 2.509e+02 3.045e+02 6.004e+02, threshold=5.019e+02, percent-clipped=5.0 2023-05-01 18:35:20,357 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235025.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:35:58,105 INFO [train.py:904] (3/8) Epoch 24, batch 1600, loss[loss=0.1954, simple_loss=0.2674, pruned_loss=0.0617, over 16756.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2484, pruned_loss=0.03893, over 3315104.44 frames. ], batch size: 124, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:36:37,993 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 1650, loss[loss=0.1456, simple_loss=0.2201, pruned_loss=0.03555, over 16686.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2509, pruned_loss=0.04012, over 3311808.67 frames. ], batch size: 83, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:37:23,209 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9276, 2.5811, 2.3004, 3.8607, 3.1534, 3.8102, 1.5827, 2.7567], device='cuda:3'), covar=tensor([0.1273, 0.0694, 0.1269, 0.0224, 0.0168, 0.0513, 0.1543, 0.0935], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0198, 0.0196, 0.0206, 0.0219, 0.0207, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 18:37:35,280 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.609e+02 2.234e+02 2.602e+02 3.440e+02 9.714e+02, threshold=5.204e+02, percent-clipped=4.0 2023-05-01 18:37:54,828 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235138.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:38:08,233 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8294, 4.2870, 4.3307, 2.9468, 3.7185, 4.2351, 3.8506, 2.4936], device='cuda:3'), covar=tensor([0.0548, 0.0094, 0.0057, 0.0451, 0.0131, 0.0116, 0.0109, 0.0527], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0087, 0.0087, 0.0135, 0.0100, 0.0112, 0.0096, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 18:38:16,050 INFO [train.py:904] (3/8) Epoch 24, batch 1700, loss[loss=0.1759, simple_loss=0.2663, pruned_loss=0.04272, over 16253.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2524, pruned_loss=0.03995, over 3317026.16 frames. ], batch size: 145, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:24,669 INFO [train.py:904] (3/8) Epoch 24, batch 1750, loss[loss=0.1932, simple_loss=0.2635, pruned_loss=0.06147, over 16905.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2534, pruned_loss=0.04052, over 3321487.93 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:52,421 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.256e+02 2.911e+02 3.628e+02 6.601e+02, threshold=5.823e+02, percent-clipped=7.0 2023-05-01 18:40:12,949 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235238.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:33,173 INFO [train.py:904] (3/8) Epoch 24, batch 1800, loss[loss=0.1863, simple_loss=0.265, pruned_loss=0.0538, over 16859.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2541, pruned_loss=0.04043, over 3320081.62 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:40:36,406 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:16,437 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4072, 5.4307, 5.2415, 4.7234, 5.3099, 2.4534, 5.0295, 5.1143], device='cuda:3'), covar=tensor([0.0109, 0.0105, 0.0212, 0.0414, 0.0108, 0.2477, 0.0149, 0.0225], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0167, 0.0209, 0.0184, 0.0184, 0.0216, 0.0197, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:41:40,805 INFO [train.py:904] (3/8) Epoch 24, batch 1850, loss[loss=0.1612, simple_loss=0.2565, pruned_loss=0.03293, over 17118.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2533, pruned_loss=0.03936, over 3328398.20 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:41:56,544 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:56,628 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4140, 4.4690, 4.8130, 4.7845, 4.8207, 4.5027, 4.4927, 4.3707], device='cuda:3'), covar=tensor([0.0376, 0.0743, 0.0382, 0.0421, 0.0540, 0.0461, 0.0878, 0.0632], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0474, 0.0463, 0.0426, 0.0506, 0.0483, 0.0567, 0.0387], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 18:41:59,454 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,660 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.161e+02 2.587e+02 3.187e+02 4.962e+02, threshold=5.175e+02, percent-clipped=0.0 2023-05-01 18:42:12,315 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:49,547 INFO [train.py:904] (3/8) Epoch 24, batch 1900, loss[loss=0.1711, simple_loss=0.2475, pruned_loss=0.04735, over 16773.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.253, pruned_loss=0.03911, over 3320588.51 frames. ], batch size: 124, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:43:01,090 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235373.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:43:31,552 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 1950, loss[loss=0.1715, simple_loss=0.2693, pruned_loss=0.0369, over 16731.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2527, pruned_loss=0.03866, over 3311334.81 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:44:26,867 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235421.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:44:28,208 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 18:44:31,116 INFO [optim.py:368] (3/8) 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 18:44:37,915 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:48,959 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 2000, loss[loss=0.1825, simple_loss=0.26, pruned_loss=0.05252, over 16378.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2523, pruned_loss=0.03881, over 3316509.04 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:45:46,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4504, 2.5982, 2.1502, 2.4241, 2.9233, 2.6617, 3.0996, 3.0990], device='cuda:3'), covar=tensor([0.0233, 0.0487, 0.0631, 0.0519, 0.0339, 0.0432, 0.0312, 0.0363], device='cuda:3'), in_proj_covar=tensor([0.0227, 0.0245, 0.0234, 0.0235, 0.0246, 0.0245, 0.0245, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:45:53,864 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235486.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:46:02,481 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0006, 5.0026, 4.8174, 4.2751, 4.9586, 2.0494, 4.6327, 4.6061], device='cuda:3'), covar=tensor([0.0109, 0.0104, 0.0208, 0.0404, 0.0097, 0.2840, 0.0147, 0.0255], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0166, 0.0208, 0.0183, 0.0183, 0.0214, 0.0196, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:46:16,400 INFO [train.py:904] (3/8) Epoch 24, batch 2050, loss[loss=0.1711, simple_loss=0.2664, pruned_loss=0.03792, over 17131.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2526, pruned_loss=0.03929, over 3312499.21 frames. ], batch size: 48, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:46:46,282 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.111e+02 2.349e+02 3.134e+02 8.501e+02, threshold=4.699e+02, percent-clipped=2.0 2023-05-01 18:46:58,309 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4954, 3.5854, 2.1871, 3.8308, 2.8106, 3.7573, 2.3842, 2.8561], device='cuda:3'), covar=tensor([0.0297, 0.0429, 0.1589, 0.0374, 0.0822, 0.0727, 0.1337, 0.0763], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0173, 0.0180, 0.0224, 0.0207, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 18:47:03,586 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:47:24,545 INFO [train.py:904] (3/8) Epoch 24, batch 2100, loss[loss=0.1922, simple_loss=0.281, pruned_loss=0.05173, over 12071.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2535, pruned_loss=0.03979, over 3306344.23 frames. ], batch size: 246, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:47:38,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2168, 3.6588, 3.7233, 1.9644, 2.9630, 2.5466, 3.7372, 3.8219], device='cuda:3'), covar=tensor([0.0355, 0.0919, 0.0624, 0.2363, 0.0967, 0.1024, 0.0687, 0.1099], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 18:48:10,074 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:31,076 INFO [train.py:904] (3/8) Epoch 24, batch 2150, loss[loss=0.158, simple_loss=0.2392, pruned_loss=0.03839, over 16745.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2539, pruned_loss=0.03987, over 3300761.09 frames. ], batch size: 83, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:48:43,103 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:46,873 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:49,845 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.182e+02 2.677e+02 3.005e+02 4.991e+02, threshold=5.354e+02, percent-clipped=2.0 2023-05-01 18:49:31,466 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7213, 2.4376, 1.9723, 2.2054, 2.7594, 2.4822, 2.7478, 2.8832], device='cuda:3'), covar=tensor([0.0256, 0.0416, 0.0550, 0.0487, 0.0305, 0.0375, 0.0236, 0.0314], device='cuda:3'), in_proj_covar=tensor([0.0227, 0.0245, 0.0234, 0.0235, 0.0245, 0.0245, 0.0245, 0.0241], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:49:41,310 INFO [train.py:904] (3/8) Epoch 24, batch 2200, loss[loss=0.1689, simple_loss=0.2663, pruned_loss=0.03581, over 17134.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2545, pruned_loss=0.0408, over 3301675.91 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:49:54,416 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235662.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:49:57,657 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:18,226 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235679.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:50,799 INFO [train.py:904] (3/8) Epoch 24, batch 2250, loss[loss=0.1754, simple_loss=0.2583, pruned_loss=0.04627, over 16483.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2556, pruned_loss=0.04095, over 3309738.83 frames. ], batch size: 75, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:51:08,553 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235716.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:51:20,156 INFO [optim.py:368] (3/8) 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 18:51:41,275 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:51:57,980 INFO [train.py:904] (3/8) Epoch 24, batch 2300, loss[loss=0.1653, simple_loss=0.2427, pruned_loss=0.04391, over 16749.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2563, pruned_loss=0.0411, over 3313022.54 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:52:50,290 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 2350, loss[loss=0.1848, simple_loss=0.2658, pruned_loss=0.05189, over 16887.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2564, pruned_loss=0.04132, over 3315505.85 frames. ], batch size: 96, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:53:13,824 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 18:53:37,790 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.657e+02 2.284e+02 2.753e+02 3.332e+02 6.201e+02, threshold=5.507e+02, percent-clipped=2.0 2023-05-01 18:54:06,770 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 18:54:14,689 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:54:15,315 INFO [train.py:904] (3/8) Epoch 24, batch 2400, loss[loss=0.1672, simple_loss=0.2593, pruned_loss=0.03759, over 17036.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2567, pruned_loss=0.04124, over 3322789.70 frames. ], batch size: 50, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:22,143 INFO [train.py:904] (3/8) Epoch 24, batch 2450, loss[loss=0.1426, simple_loss=0.2324, pruned_loss=0.02644, over 16963.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2574, pruned_loss=0.04098, over 3321853.07 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:32,524 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 18:55:33,241 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:55:51,880 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.168e+02 2.510e+02 2.858e+02 5.891e+02, threshold=5.019e+02, percent-clipped=1.0 2023-05-01 18:56:03,651 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 18:56:28,797 INFO [train.py:904] (3/8) Epoch 24, batch 2500, loss[loss=0.1593, simple_loss=0.2639, pruned_loss=0.02735, over 17041.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2574, pruned_loss=0.04062, over 3326381.49 frames. ], batch size: 50, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:56:37,230 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:57:05,135 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6717, 4.7587, 4.9291, 4.7212, 4.7380, 5.3937, 4.8666, 4.5068], device='cuda:3'), covar=tensor([0.1540, 0.2489, 0.2778, 0.2417, 0.3116, 0.1218, 0.1966, 0.3013], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0622, 0.0686, 0.0514, 0.0683, 0.0716, 0.0536, 0.0683], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 18:57:28,600 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4070, 2.3231, 2.3079, 4.1743, 2.2379, 2.7294, 2.3994, 2.4508], device='cuda:3'), covar=tensor([0.1328, 0.3714, 0.3249, 0.0577, 0.4209, 0.2604, 0.3618, 0.3813], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0461, 0.0378, 0.0334, 0.0442, 0.0529, 0.0432, 0.0539], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:57:38,832 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1227, 3.9572, 4.2165, 4.3151, 4.3768, 3.9705, 4.1786, 4.3855], device='cuda:3'), covar=tensor([0.1659, 0.1225, 0.1186, 0.0660, 0.0654, 0.1496, 0.2701, 0.0830], device='cuda:3'), in_proj_covar=tensor([0.0682, 0.0845, 0.0973, 0.0853, 0.0652, 0.0673, 0.0708, 0.0820], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 18:57:41,344 INFO [train.py:904] (3/8) Epoch 24, batch 2550, loss[loss=0.1288, simple_loss=0.215, pruned_loss=0.02133, over 16791.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2573, pruned_loss=0.04077, over 3318316.32 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:57:47,293 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-01 18:57:59,288 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236016.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:58:01,892 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.266e+02 2.526e+02 3.100e+02 4.908e+02, threshold=5.053e+02, percent-clipped=0.0 2023-05-01 18:58:24,269 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236035.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:49,127 INFO [train.py:904] (3/8) Epoch 24, batch 2600, loss[loss=0.1481, simple_loss=0.2504, pruned_loss=0.02297, over 17115.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2569, pruned_loss=0.04079, over 3317568.55 frames. ], batch size: 49, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:59:03,926 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236064.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:59:25,448 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:59:34,824 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5863, 3.7407, 2.3520, 4.1635, 2.8894, 4.1023, 2.3723, 3.0854], device='cuda:3'), covar=tensor([0.0347, 0.0419, 0.1562, 0.0376, 0.0849, 0.0649, 0.1546, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0182, 0.0197, 0.0173, 0.0179, 0.0223, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 18:59:58,182 INFO [train.py:904] (3/8) Epoch 24, batch 2650, loss[loss=0.1636, simple_loss=0.2651, pruned_loss=0.03109, over 16061.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2576, pruned_loss=0.04031, over 3324555.00 frames. ], batch size: 35, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:00:01,998 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 19:00:12,848 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 19:00:27,606 INFO [optim.py:368] (3/8) 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 19:00:58,590 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:01:05,908 INFO [train.py:904] (3/8) Epoch 24, batch 2700, loss[loss=0.1508, simple_loss=0.2457, pruned_loss=0.02789, over 17010.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2579, pruned_loss=0.03972, over 3333757.27 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:01:57,228 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6255, 3.7043, 2.3907, 4.1046, 2.9700, 4.0354, 2.4898, 3.0489], device='cuda:3'), covar=tensor([0.0313, 0.0405, 0.1496, 0.0380, 0.0764, 0.0664, 0.1383, 0.0727], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0174, 0.0180, 0.0224, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:02:15,469 INFO [train.py:904] (3/8) Epoch 24, batch 2750, loss[loss=0.1452, simple_loss=0.2444, pruned_loss=0.02299, over 17169.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2574, pruned_loss=0.03933, over 3326821.84 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:02:44,667 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.972e+02 2.253e+02 2.676e+02 4.118e+02, threshold=4.507e+02, percent-clipped=0.0 2023-05-01 19:03:22,962 INFO [train.py:904] (3/8) Epoch 24, batch 2800, loss[loss=0.1656, simple_loss=0.2628, pruned_loss=0.03415, over 16716.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2573, pruned_loss=0.03882, over 3329487.88 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:04:32,099 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5619, 4.5605, 4.7147, 4.5313, 4.5477, 5.1431, 4.6914, 4.3732], device='cuda:3'), covar=tensor([0.1593, 0.2254, 0.2705, 0.2448, 0.2979, 0.1132, 0.1807, 0.2737], device='cuda:3'), in_proj_covar=tensor([0.0420, 0.0615, 0.0680, 0.0510, 0.0677, 0.0710, 0.0531, 0.0676], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 19:04:32,915 INFO [train.py:904] (3/8) Epoch 24, batch 2850, loss[loss=0.1911, simple_loss=0.2614, pruned_loss=0.06042, over 16356.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2567, pruned_loss=0.0392, over 3324699.08 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:05:04,031 INFO [optim.py:368] (3/8) 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 19:05:17,546 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 2900, loss[loss=0.1496, simple_loss=0.2329, pruned_loss=0.03316, over 16965.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2558, pruned_loss=0.0397, over 3322006.49 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:05:50,124 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4179, 5.3099, 5.2814, 4.7624, 4.8919, 5.2934, 5.2538, 4.8534], device='cuda:3'), covar=tensor([0.0549, 0.0523, 0.0293, 0.0356, 0.1132, 0.0487, 0.0280, 0.0810], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0466, 0.0366, 0.0366, 0.0372, 0.0425, 0.0250, 0.0443], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 19:06:13,275 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:25,482 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236383.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:48,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6872, 1.7790, 2.2711, 2.5548, 2.6033, 2.6411, 1.9291, 2.8542], device='cuda:3'), covar=tensor([0.0195, 0.0538, 0.0381, 0.0328, 0.0323, 0.0353, 0.0540, 0.0188], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0197, 0.0185, 0.0190, 0.0205, 0.0163, 0.0201, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:06:53,578 INFO [train.py:904] (3/8) Epoch 24, batch 2950, loss[loss=0.157, simple_loss=0.2511, pruned_loss=0.0315, over 17110.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2556, pruned_loss=0.04073, over 3319097.47 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:07:24,081 INFO [optim.py:368] (3/8) 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 19:07:50,677 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3911, 4.1619, 4.5551, 2.5897, 4.7935, 4.7971, 3.6296, 3.8060], device='cuda:3'), covar=tensor([0.0628, 0.0237, 0.0192, 0.1022, 0.0071, 0.0180, 0.0366, 0.0367], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0108, 0.0099, 0.0138, 0.0082, 0.0128, 0.0128, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:07:54,791 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8448, 4.8428, 4.6907, 4.0220, 4.7992, 1.8627, 4.5335, 4.4161], device='cuda:3'), covar=tensor([0.0164, 0.0131, 0.0249, 0.0463, 0.0133, 0.2982, 0.0182, 0.0260], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0166, 0.0208, 0.0183, 0.0183, 0.0213, 0.0197, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:07:54,795 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:08:02,709 INFO [train.py:904] (3/8) Epoch 24, batch 3000, loss[loss=0.1901, simple_loss=0.2595, pruned_loss=0.06039, over 16394.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.256, pruned_loss=0.04148, over 3316039.08 frames. ], batch size: 146, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:08:02,710 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 19:08:12,054 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 19:08:46,513 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9322, 4.1366, 4.0128, 3.0814, 3.5290, 4.0405, 3.7225, 1.9509], device='cuda:3'), covar=tensor([0.0518, 0.0108, 0.0089, 0.0423, 0.0199, 0.0173, 0.0140, 0.0733], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0136, 0.0101, 0.0112, 0.0097, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 19:09:12,138 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236495.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:09:22,990 INFO [train.py:904] (3/8) Epoch 24, batch 3050, loss[loss=0.1555, simple_loss=0.2489, pruned_loss=0.03109, over 17207.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2561, pruned_loss=0.04185, over 3318999.59 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:09:53,422 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.094e+02 2.423e+02 2.785e+02 4.383e+02, threshold=4.846e+02, percent-clipped=1.0 2023-05-01 19:10:32,486 INFO [train.py:904] (3/8) Epoch 24, batch 3100, loss[loss=0.1386, simple_loss=0.2235, pruned_loss=0.02691, over 16800.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2554, pruned_loss=0.04123, over 3316396.78 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:10:51,040 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2034, 4.2361, 4.5580, 4.5475, 4.5929, 4.2973, 4.3277, 4.2304], device='cuda:3'), covar=tensor([0.0387, 0.0699, 0.0448, 0.0432, 0.0533, 0.0458, 0.0782, 0.0625], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0477, 0.0464, 0.0427, 0.0508, 0.0484, 0.0569, 0.0389], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 19:11:25,200 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8526, 2.7847, 2.6215, 4.2552, 3.4859, 4.1422, 1.7399, 3.0395], device='cuda:3'), covar=tensor([0.1297, 0.0651, 0.1098, 0.0165, 0.0160, 0.0381, 0.1473, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0176, 0.0195, 0.0195, 0.0205, 0.0216, 0.0204, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:11:43,886 INFO [train.py:904] (3/8) Epoch 24, batch 3150, loss[loss=0.1796, simple_loss=0.2632, pruned_loss=0.04797, over 16472.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.255, pruned_loss=0.0414, over 3314876.48 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:11:56,301 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0674, 4.0837, 4.3790, 4.3808, 4.4208, 4.1408, 4.1708, 4.1316], device='cuda:3'), covar=tensor([0.0393, 0.0695, 0.0459, 0.0460, 0.0521, 0.0490, 0.0823, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0479, 0.0467, 0.0430, 0.0511, 0.0486, 0.0572, 0.0391], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 19:12:13,744 INFO [optim.py:368] (3/8) 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 19:12:52,323 INFO [train.py:904] (3/8) Epoch 24, batch 3200, loss[loss=0.165, simple_loss=0.2431, pruned_loss=0.04343, over 16769.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.254, pruned_loss=0.04054, over 3315729.39 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:13:09,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6848, 5.0284, 5.3393, 5.2806, 5.3553, 4.9805, 4.7492, 4.7567], device='cuda:3'), covar=tensor([0.0641, 0.0650, 0.0589, 0.0675, 0.0667, 0.0576, 0.1444, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0480, 0.0467, 0.0430, 0.0511, 0.0487, 0.0573, 0.0391], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 19:13:16,509 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 19:13:21,962 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:13:28,455 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5159, 3.6081, 2.3462, 3.8569, 2.8746, 3.8002, 2.2818, 2.9444], device='cuda:3'), covar=tensor([0.0287, 0.0380, 0.1357, 0.0305, 0.0725, 0.0696, 0.1455, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0182, 0.0198, 0.0174, 0.0179, 0.0224, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:14:01,576 INFO [train.py:904] (3/8) Epoch 24, batch 3250, loss[loss=0.1645, simple_loss=0.2555, pruned_loss=0.03671, over 16452.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2538, pruned_loss=0.04025, over 3323147.08 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:14:14,117 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-01 19:14:27,660 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 19:15:11,539 INFO [train.py:904] (3/8) Epoch 24, batch 3300, loss[loss=0.168, simple_loss=0.2672, pruned_loss=0.03439, over 16710.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2548, pruned_loss=0.04081, over 3321708.21 frames. ], batch size: 62, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:21,167 INFO [train.py:904] (3/8) Epoch 24, batch 3350, loss[loss=0.1584, simple_loss=0.2587, pruned_loss=0.02905, over 17115.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2548, pruned_loss=0.04017, over 3324699.44 frames. ], batch size: 49, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:51,748 INFO [optim.py:368] (3/8) 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 19:17:33,207 INFO [train.py:904] (3/8) Epoch 24, batch 3400, loss[loss=0.1463, simple_loss=0.2342, pruned_loss=0.02919, over 16832.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2551, pruned_loss=0.04023, over 3311543.94 frames. ], batch size: 42, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:18:31,120 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236892.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:18:34,762 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 19:18:45,596 INFO [train.py:904] (3/8) Epoch 24, batch 3450, loss[loss=0.1644, simple_loss=0.2399, pruned_loss=0.04443, over 16772.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2538, pruned_loss=0.0398, over 3315220.26 frames. ], batch size: 102, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:15,849 INFO [optim.py:368] (3/8) 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 19:19:45,816 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 19:19:55,267 INFO [train.py:904] (3/8) Epoch 24, batch 3500, loss[loss=0.1656, simple_loss=0.2633, pruned_loss=0.03392, over 17248.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2523, pruned_loss=0.03908, over 3320994.92 frames. ], batch size: 52, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:56,834 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236953.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 19:20:40,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2191, 2.1673, 2.3067, 3.8466, 2.2445, 2.5594, 2.3112, 2.3745], device='cuda:3'), covar=tensor([0.1519, 0.3964, 0.3100, 0.0638, 0.3815, 0.2804, 0.3955, 0.3391], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0461, 0.0380, 0.0335, 0.0441, 0.0531, 0.0432, 0.0540], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:20:41,679 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0842, 2.3528, 2.6282, 3.0728, 2.9222, 3.5585, 2.4814, 3.4866], device='cuda:3'), covar=tensor([0.0251, 0.0464, 0.0368, 0.0332, 0.0323, 0.0199, 0.0481, 0.0202], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0196, 0.0185, 0.0189, 0.0204, 0.0163, 0.0200, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:21:06,713 INFO [train.py:904] (3/8) Epoch 24, batch 3550, loss[loss=0.152, simple_loss=0.2526, pruned_loss=0.02569, over 17123.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2515, pruned_loss=0.03862, over 3320210.96 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:21:35,731 INFO [optim.py:368] (3/8) 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 19:22:09,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7425, 2.8830, 2.7187, 4.7275, 3.8654, 4.2773, 1.6315, 3.1635], device='cuda:3'), covar=tensor([0.1608, 0.0855, 0.1276, 0.0279, 0.0275, 0.0444, 0.1864, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0177, 0.0197, 0.0197, 0.0206, 0.0218, 0.0205, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:22:15,059 INFO [train.py:904] (3/8) Epoch 24, batch 3600, loss[loss=0.1793, simple_loss=0.2509, pruned_loss=0.0539, over 16934.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2507, pruned_loss=0.03871, over 3324758.81 frames. ], batch size: 109, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:22:42,798 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 19:23:26,125 INFO [train.py:904] (3/8) Epoch 24, batch 3650, loss[loss=0.1936, simple_loss=0.2616, pruned_loss=0.06281, over 11473.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2499, pruned_loss=0.03904, over 3312383.07 frames. ], batch size: 248, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:58,725 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.153e+02 2.502e+02 3.114e+02 9.969e+02, threshold=5.004e+02, percent-clipped=5.0 2023-05-01 19:24:39,858 INFO [train.py:904] (3/8) Epoch 24, batch 3700, loss[loss=0.1755, simple_loss=0.2563, pruned_loss=0.04734, over 11219.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2493, pruned_loss=0.04079, over 3301768.94 frames. ], batch size: 247, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:25:53,324 INFO [train.py:904] (3/8) Epoch 24, batch 3750, loss[loss=0.1904, simple_loss=0.2618, pruned_loss=0.05954, over 16746.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2505, pruned_loss=0.04242, over 3288058.34 frames. ], batch size: 134, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:26:25,688 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.223e+02 2.609e+02 3.048e+02 5.921e+02, threshold=5.219e+02, percent-clipped=3.0 2023-05-01 19:26:57,489 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237248.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:27:03,613 INFO [train.py:904] (3/8) Epoch 24, batch 3800, loss[loss=0.1908, simple_loss=0.2763, pruned_loss=0.05263, over 17006.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2518, pruned_loss=0.04347, over 3286919.27 frames. ], batch size: 55, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:28:20,063 INFO [train.py:904] (3/8) Epoch 24, batch 3850, loss[loss=0.1615, simple_loss=0.245, pruned_loss=0.03899, over 16461.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2514, pruned_loss=0.04402, over 3291125.53 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:28:45,547 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7364, 2.9328, 3.1867, 2.0658, 2.8534, 2.2218, 3.3899, 3.3201], device='cuda:3'), covar=tensor([0.0239, 0.0967, 0.0618, 0.1962, 0.0853, 0.1020, 0.0496, 0.0846], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-01 19:28:52,885 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.317e+02 2.601e+02 3.052e+02 4.766e+02, threshold=5.202e+02, percent-clipped=0.0 2023-05-01 19:29:30,850 INFO [train.py:904] (3/8) Epoch 24, batch 3900, loss[loss=0.168, simple_loss=0.247, pruned_loss=0.04448, over 16613.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2506, pruned_loss=0.04404, over 3297758.54 frames. ], batch size: 76, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:29:57,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0627, 5.0633, 4.8781, 3.5138, 5.0329, 1.7712, 4.6532, 4.3745], device='cuda:3'), covar=tensor([0.0106, 0.0087, 0.0225, 0.0755, 0.0106, 0.3720, 0.0174, 0.0422], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0168, 0.0210, 0.0186, 0.0186, 0.0215, 0.0198, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:30:01,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7498, 4.7111, 4.6453, 4.0953, 4.7005, 1.7819, 4.4503, 4.2144], device='cuda:3'), covar=tensor([0.0114, 0.0099, 0.0190, 0.0336, 0.0108, 0.2940, 0.0133, 0.0255], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0168, 0.0210, 0.0185, 0.0185, 0.0215, 0.0198, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:30:43,384 INFO [train.py:904] (3/8) Epoch 24, batch 3950, loss[loss=0.1769, simple_loss=0.2524, pruned_loss=0.05071, over 16505.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2503, pruned_loss=0.04463, over 3304715.17 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:30:57,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3071, 2.7976, 3.1398, 1.8213, 3.2600, 3.2025, 2.7386, 2.5014], device='cuda:3'), covar=tensor([0.0755, 0.0321, 0.0221, 0.1142, 0.0134, 0.0245, 0.0448, 0.0464], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0140, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:31:17,988 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.223e+02 2.645e+02 3.283e+02 6.262e+02, threshold=5.289e+02, percent-clipped=1.0 2023-05-01 19:31:57,060 INFO [train.py:904] (3/8) Epoch 24, batch 4000, loss[loss=0.171, simple_loss=0.2439, pruned_loss=0.04904, over 16915.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2505, pruned_loss=0.04496, over 3295441.58 frames. ], batch size: 109, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:32:53,238 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237491.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:33:06,366 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5616, 3.6721, 2.3489, 4.1679, 2.8087, 4.2302, 2.3676, 2.9020], device='cuda:3'), covar=tensor([0.0313, 0.0400, 0.1702, 0.0148, 0.0871, 0.0321, 0.1717, 0.0839], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0174, 0.0179, 0.0223, 0.0205, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:33:10,918 INFO [train.py:904] (3/8) Epoch 24, batch 4050, loss[loss=0.1703, simple_loss=0.2544, pruned_loss=0.0431, over 16487.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2512, pruned_loss=0.04438, over 3291579.21 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:33:32,940 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5340, 2.2713, 1.8440, 2.0678, 2.5615, 2.2635, 2.2894, 2.7260], device='cuda:3'), covar=tensor([0.0192, 0.0464, 0.0567, 0.0493, 0.0291, 0.0403, 0.0226, 0.0279], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0244, 0.0234, 0.0235, 0.0245, 0.0245, 0.0246, 0.0243], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:33:43,734 INFO [optim.py:368] (3/8) 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 19:34:01,573 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 19:34:17,613 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=237548.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:34:23,854 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237552.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:34:24,477 INFO [train.py:904] (3/8) Epoch 24, batch 4100, loss[loss=0.1639, simple_loss=0.2507, pruned_loss=0.03859, over 16513.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2529, pruned_loss=0.04389, over 3291680.44 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:34:38,912 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237563.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:35:09,050 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2074, 4.2812, 4.5383, 4.4933, 4.5144, 4.2690, 4.2509, 4.2122], device='cuda:3'), covar=tensor([0.0324, 0.0477, 0.0329, 0.0376, 0.0504, 0.0371, 0.0861, 0.0496], device='cuda:3'), in_proj_covar=tensor([0.0426, 0.0476, 0.0460, 0.0423, 0.0506, 0.0481, 0.0565, 0.0386], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 19:35:26,282 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7783, 1.3915, 1.6726, 1.6880, 1.7158, 1.9006, 1.5815, 1.7872], device='cuda:3'), covar=tensor([0.0274, 0.0386, 0.0216, 0.0315, 0.0271, 0.0204, 0.0426, 0.0144], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0195, 0.0184, 0.0189, 0.0204, 0.0163, 0.0201, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:35:30,063 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=237596.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:35:37,575 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6847, 5.9585, 5.6508, 5.8394, 5.4190, 5.3319, 5.3867, 6.1549], device='cuda:3'), covar=tensor([0.1353, 0.0901, 0.1392, 0.0852, 0.0860, 0.0648, 0.1195, 0.0928], device='cuda:3'), in_proj_covar=tensor([0.0707, 0.0858, 0.0709, 0.0664, 0.0547, 0.0551, 0.0724, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:35:40,165 INFO [train.py:904] (3/8) Epoch 24, batch 4150, loss[loss=0.2394, simple_loss=0.3148, pruned_loss=0.08199, over 11414.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2606, pruned_loss=0.04675, over 3257301.79 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:36:14,072 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237624.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:36:16,368 INFO [optim.py:368] (3/8) 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 19:36:22,637 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 19:36:56,393 INFO [train.py:904] (3/8) Epoch 24, batch 4200, loss[loss=0.2154, simple_loss=0.3091, pruned_loss=0.06091, over 16241.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2672, pruned_loss=0.04833, over 3233587.95 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:37:14,895 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 19:37:46,526 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4704, 4.2437, 4.1571, 4.6256, 4.8201, 4.2822, 4.7957, 4.8281], device='cuda:3'), covar=tensor([0.1797, 0.1637, 0.2668, 0.0947, 0.0696, 0.1807, 0.0942, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0670, 0.0829, 0.0953, 0.0834, 0.0637, 0.0661, 0.0692, 0.0805], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:37:53,912 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0664, 3.3082, 3.2826, 2.2467, 3.0672, 3.3692, 3.1803, 2.0983], device='cuda:3'), covar=tensor([0.0561, 0.0076, 0.0089, 0.0424, 0.0116, 0.0129, 0.0105, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0087, 0.0088, 0.0136, 0.0101, 0.0113, 0.0097, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 19:38:10,756 INFO [train.py:904] (3/8) Epoch 24, batch 4250, loss[loss=0.1839, simple_loss=0.2763, pruned_loss=0.04579, over 16713.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2697, pruned_loss=0.0475, over 3224693.21 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:45,324 INFO [optim.py:368] (3/8) 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 19:39:26,152 INFO [train.py:904] (3/8) Epoch 24, batch 4300, loss[loss=0.188, simple_loss=0.2808, pruned_loss=0.04755, over 16272.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2711, pruned_loss=0.04671, over 3224198.19 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:40:00,920 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0291, 2.8887, 2.8296, 5.1249, 4.0430, 4.2924, 1.9358, 3.1366], device='cuda:3'), covar=tensor([0.1176, 0.0776, 0.1175, 0.0150, 0.0394, 0.0410, 0.1470, 0.0881], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0194, 0.0205, 0.0215, 0.0203, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:40:23,430 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237790.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:40:41,737 INFO [train.py:904] (3/8) Epoch 24, batch 4350, loss[loss=0.1897, simple_loss=0.2782, pruned_loss=0.05057, over 16689.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.274, pruned_loss=0.04764, over 3203401.10 frames. ], batch size: 62, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:41:03,102 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1265, 2.5047, 2.6198, 1.9138, 2.7484, 2.7763, 2.3700, 2.3920], device='cuda:3'), covar=tensor([0.0687, 0.0270, 0.0222, 0.0926, 0.0118, 0.0222, 0.0501, 0.0426], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0111, 0.0101, 0.0141, 0.0083, 0.0129, 0.0130, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:41:14,404 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.331e+02 2.533e+02 3.016e+02 1.010e+03, threshold=5.065e+02, percent-clipped=1.0 2023-05-01 19:41:45,911 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:49,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9448, 4.1835, 4.0093, 4.0547, 3.7610, 3.8116, 3.8029, 4.1881], device='cuda:3'), covar=tensor([0.1008, 0.0871, 0.0973, 0.0752, 0.0733, 0.1562, 0.0877, 0.0898], device='cuda:3'), in_proj_covar=tensor([0.0696, 0.0842, 0.0697, 0.0654, 0.0537, 0.0542, 0.0711, 0.0661], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:41:52,145 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237851.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:54,783 INFO [train.py:904] (3/8) Epoch 24, batch 4400, loss[loss=0.1997, simple_loss=0.2841, pruned_loss=0.05769, over 16606.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2768, pruned_loss=0.04977, over 3165413.81 frames. ], batch size: 57, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:05,639 INFO [train.py:904] (3/8) Epoch 24, batch 4450, loss[loss=0.191, simple_loss=0.2914, pruned_loss=0.04529, over 15376.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2799, pruned_loss=0.05085, over 3181893.41 frames. ], batch size: 191, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:18,904 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237912.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:43:28,554 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237919.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:43:38,899 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.917e+02 2.137e+02 2.680e+02 5.470e+02, threshold=4.273e+02, percent-clipped=1.0 2023-05-01 19:44:16,768 INFO [train.py:904] (3/8) Epoch 24, batch 4500, loss[loss=0.2013, simple_loss=0.2917, pruned_loss=0.05542, over 16716.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2812, pruned_loss=0.05182, over 3200138.55 frames. ], batch size: 124, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:44:46,725 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237973.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:44:59,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9451, 4.1958, 4.0193, 4.0719, 3.7882, 3.8378, 3.8190, 4.1881], device='cuda:3'), covar=tensor([0.1119, 0.0891, 0.1065, 0.0809, 0.0798, 0.1634, 0.0952, 0.1033], device='cuda:3'), in_proj_covar=tensor([0.0694, 0.0840, 0.0696, 0.0652, 0.0537, 0.0541, 0.0709, 0.0661], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:45:32,218 INFO [train.py:904] (3/8) Epoch 24, batch 4550, loss[loss=0.1962, simple_loss=0.2886, pruned_loss=0.0519, over 16998.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2823, pruned_loss=0.0529, over 3204713.42 frames. ], batch size: 50, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:45:49,358 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-05-01 19:46:04,568 INFO [optim.py:368] (3/8) 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 19:46:20,515 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-05-01 19:46:44,383 INFO [train.py:904] (3/8) Epoch 24, batch 4600, loss[loss=0.2019, simple_loss=0.2921, pruned_loss=0.05587, over 16250.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2829, pruned_loss=0.05289, over 3217579.85 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:46:49,557 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0077, 4.3760, 3.0248, 2.6834, 3.1589, 2.5939, 4.8443, 3.7685], device='cuda:3'), covar=tensor([0.2885, 0.0592, 0.1951, 0.2429, 0.2496, 0.2178, 0.0415, 0.1223], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0272, 0.0308, 0.0319, 0.0302, 0.0266, 0.0298, 0.0342], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 19:47:00,824 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:15,183 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238074.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:18,985 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 19:47:35,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6648, 5.9228, 5.6849, 5.7618, 5.4178, 5.2458, 5.3264, 6.0669], device='cuda:3'), covar=tensor([0.1123, 0.0700, 0.0950, 0.0738, 0.0797, 0.0666, 0.1095, 0.0743], device='cuda:3'), in_proj_covar=tensor([0.0691, 0.0837, 0.0694, 0.0650, 0.0535, 0.0539, 0.0707, 0.0659], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:47:43,921 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7759, 4.8308, 4.6617, 4.3196, 4.3744, 4.7612, 4.4793, 4.4739], device='cuda:3'), covar=tensor([0.0468, 0.0335, 0.0219, 0.0242, 0.0676, 0.0277, 0.0407, 0.0529], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0444, 0.0349, 0.0349, 0.0352, 0.0402, 0.0239, 0.0420], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:47:50,311 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3045, 3.2291, 1.6602, 3.6865, 2.3564, 3.6067, 1.9922, 2.5856], device='cuda:3'), covar=tensor([0.0299, 0.0442, 0.2268, 0.0199, 0.1047, 0.0520, 0.1929, 0.0910], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0194, 0.0169, 0.0177, 0.0220, 0.0203, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:47:53,659 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 19:47:56,653 INFO [train.py:904] (3/8) Epoch 24, batch 4650, loss[loss=0.2078, simple_loss=0.2832, pruned_loss=0.06621, over 16387.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2818, pruned_loss=0.05307, over 3223460.58 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:48:29,618 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:30,376 INFO [optim.py:368] (3/8) 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 19:48:30,790 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238126.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:43,747 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:59,278 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:00,460 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 4700, loss[loss=0.1841, simple_loss=0.2731, pruned_loss=0.04758, over 16385.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2793, pruned_loss=0.05214, over 3222976.42 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:49:58,163 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:06,544 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 19:50:09,159 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238195.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:20,662 INFO [train.py:904] (3/8) Epoch 24, batch 4750, loss[loss=0.162, simple_loss=0.2454, pruned_loss=0.03926, over 17119.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2755, pruned_loss=0.05023, over 3213494.53 frames. ], batch size: 49, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:50:41,740 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-01 19:50:43,582 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238219.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:50:53,741 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 1.823e+02 2.174e+02 2.542e+02 5.265e+02, threshold=4.348e+02, percent-clipped=2.0 2023-05-01 19:51:31,485 INFO [train.py:904] (3/8) Epoch 24, batch 4800, loss[loss=0.1755, simple_loss=0.2701, pruned_loss=0.04046, over 16697.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2721, pruned_loss=0.04822, over 3222722.18 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:51:49,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0309, 2.9954, 1.8561, 3.2425, 2.2777, 3.2849, 2.1094, 2.5046], device='cuda:3'), covar=tensor([0.0300, 0.0401, 0.1687, 0.0190, 0.0929, 0.0524, 0.1512, 0.0782], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0177, 0.0220, 0.0203, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 19:51:52,869 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238267.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:51:54,668 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238268.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:52:25,785 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3841, 2.5121, 2.4824, 4.2422, 2.4609, 2.8769, 2.5689, 2.6389], device='cuda:3'), covar=tensor([0.1390, 0.3487, 0.2801, 0.0496, 0.3699, 0.2369, 0.3344, 0.2906], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0460, 0.0375, 0.0331, 0.0440, 0.0528, 0.0429, 0.0537], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:52:46,485 INFO [train.py:904] (3/8) Epoch 24, batch 4850, loss[loss=0.1791, simple_loss=0.2755, pruned_loss=0.04135, over 16434.00 frames. ], tot_loss[loss=0.184, simple_loss=0.273, pruned_loss=0.04749, over 3200622.52 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:52:53,317 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7872, 4.0561, 4.2284, 4.1559, 4.2014, 3.9874, 3.7411, 3.9554], device='cuda:3'), covar=tensor([0.0557, 0.0667, 0.0541, 0.0680, 0.0589, 0.0557, 0.1358, 0.0580], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0460, 0.0449, 0.0413, 0.0495, 0.0468, 0.0552, 0.0376], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 19:53:22,735 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 4900, loss[loss=0.1561, simple_loss=0.255, pruned_loss=0.02857, over 16358.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2725, pruned_loss=0.0462, over 3179987.43 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:17,229 INFO [train.py:904] (3/8) Epoch 24, batch 4950, loss[loss=0.1937, simple_loss=0.2855, pruned_loss=0.05091, over 16873.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2715, pruned_loss=0.04516, over 3198847.32 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:41,592 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:41,744 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238420.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:55:49,724 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.002e+02 2.407e+02 2.827e+02 4.284e+02, threshold=4.815e+02, percent-clipped=0.0 2023-05-01 19:55:57,440 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:59,983 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9787, 4.9372, 4.8348, 3.7558, 4.9159, 1.7329, 4.5986, 4.5238], device='cuda:3'), covar=tensor([0.0138, 0.0116, 0.0211, 0.0681, 0.0152, 0.3216, 0.0169, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0163, 0.0204, 0.0181, 0.0180, 0.0209, 0.0192, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:56:20,218 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238446.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:29,585 INFO [train.py:904] (3/8) Epoch 24, batch 5000, loss[loss=0.1826, simple_loss=0.2799, pruned_loss=0.04269, over 16645.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2733, pruned_loss=0.04566, over 3209061.48 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:56:35,183 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4689, 4.3152, 4.5060, 4.6606, 4.8407, 4.4252, 4.8297, 4.8618], device='cuda:3'), covar=tensor([0.1711, 0.1288, 0.1641, 0.0781, 0.0499, 0.1024, 0.0616, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0644, 0.0797, 0.0918, 0.0806, 0.0615, 0.0636, 0.0665, 0.0775], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 19:57:11,133 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238481.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:57:11,981 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238482.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:30,535 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:42,731 INFO [train.py:904] (3/8) Epoch 24, batch 5050, loss[loss=0.1795, simple_loss=0.2717, pruned_loss=0.04365, over 16526.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.274, pruned_loss=0.04533, over 3211565.84 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:58:15,454 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 2.006e+02 2.316e+02 2.692e+02 7.292e+02, threshold=4.632e+02, percent-clipped=1.0 2023-05-01 19:58:53,425 INFO [train.py:904] (3/8) Epoch 24, batch 5100, loss[loss=0.1675, simple_loss=0.2601, pruned_loss=0.03746, over 16694.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2722, pruned_loss=0.04466, over 3212863.22 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:59:16,757 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:07,249 INFO [train.py:904] (3/8) Epoch 24, batch 5150, loss[loss=0.1824, simple_loss=0.2851, pruned_loss=0.03983, over 15496.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2724, pruned_loss=0.04431, over 3198551.24 frames. ], batch size: 191, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:00:26,118 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:36,902 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6798, 2.5539, 2.5218, 3.8188, 2.5765, 3.8227, 1.5546, 2.8573], device='cuda:3'), covar=tensor([0.1341, 0.0799, 0.1106, 0.0139, 0.0178, 0.0354, 0.1652, 0.0770], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0193, 0.0204, 0.0215, 0.0204, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:00:38,785 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.999e+02 2.272e+02 2.615e+02 4.924e+02, threshold=4.544e+02, percent-clipped=1.0 2023-05-01 20:01:17,675 INFO [train.py:904] (3/8) Epoch 24, batch 5200, loss[loss=0.1679, simple_loss=0.2576, pruned_loss=0.03907, over 16402.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2711, pruned_loss=0.0438, over 3186967.66 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:01:19,622 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 20:01:57,535 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:02:18,717 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 20:02:19,359 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4989, 3.4157, 3.9344, 1.8916, 4.0603, 4.0754, 3.0170, 3.0012], device='cuda:3'), covar=tensor([0.0796, 0.0259, 0.0156, 0.1205, 0.0064, 0.0107, 0.0396, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0138, 0.0082, 0.0126, 0.0129, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:02:28,268 INFO [train.py:904] (3/8) Epoch 24, batch 5250, loss[loss=0.1703, simple_loss=0.258, pruned_loss=0.04137, over 16620.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2687, pruned_loss=0.04317, over 3197958.46 frames. ], batch size: 62, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:02:52,124 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 20:03:07,595 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238730.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:25,260 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:39,614 INFO [train.py:904] (3/8) Epoch 24, batch 5300, loss[loss=0.1472, simple_loss=0.2258, pruned_loss=0.03426, over 16316.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2651, pruned_loss=0.04216, over 3198253.33 frames. ], batch size: 35, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:01,108 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:12,364 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238776.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:04:15,072 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238778.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:20,319 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:43,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6623, 3.0014, 3.2265, 1.9427, 2.7827, 2.2150, 3.1837, 3.2504], device='cuda:3'), covar=tensor([0.0373, 0.0883, 0.0609, 0.2119, 0.0954, 0.0982, 0.0817, 0.0988], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0166, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:04:50,564 INFO [train.py:904] (3/8) Epoch 24, batch 5350, loss[loss=0.2161, simple_loss=0.2889, pruned_loss=0.07164, over 12244.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2632, pruned_loss=0.04171, over 3209081.83 frames. ], batch size: 246, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:05:21,926 INFO [optim.py:368] (3/8) 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 20:05:27,688 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238830.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:06:01,798 INFO [train.py:904] (3/8) Epoch 24, batch 5400, loss[loss=0.1593, simple_loss=0.2592, pruned_loss=0.02968, over 16853.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.266, pruned_loss=0.04251, over 3198497.95 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:04,581 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8000, 4.0327, 3.0246, 2.4822, 2.7615, 2.6136, 4.4243, 3.5753], device='cuda:3'), covar=tensor([0.2736, 0.0589, 0.1698, 0.2506, 0.2493, 0.1889, 0.0365, 0.1166], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0271, 0.0307, 0.0318, 0.0300, 0.0265, 0.0298, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 20:07:11,346 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2831, 4.1627, 4.3644, 4.4942, 4.6480, 4.2663, 4.6194, 4.6594], device='cuda:3'), covar=tensor([0.1762, 0.1275, 0.1619, 0.0794, 0.0576, 0.1159, 0.0823, 0.0774], device='cuda:3'), in_proj_covar=tensor([0.0649, 0.0799, 0.0921, 0.0810, 0.0617, 0.0639, 0.0667, 0.0779], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:07:18,015 INFO [train.py:904] (3/8) Epoch 24, batch 5450, loss[loss=0.2248, simple_loss=0.3121, pruned_loss=0.06875, over 16398.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2691, pruned_loss=0.04417, over 3183468.44 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:54,282 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.137e+02 2.656e+02 3.537e+02 7.468e+02, threshold=5.312e+02, percent-clipped=13.0 2023-05-01 20:08:11,052 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238936.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:08:17,831 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6894, 4.7682, 4.5687, 4.2760, 4.2603, 4.6663, 4.4967, 4.3996], device='cuda:3'), covar=tensor([0.0676, 0.0657, 0.0318, 0.0322, 0.0900, 0.0613, 0.0439, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0298, 0.0448, 0.0351, 0.0351, 0.0354, 0.0408, 0.0238, 0.0423], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:08:28,479 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2340, 3.3178, 1.9953, 3.6045, 2.4855, 3.6380, 2.2271, 2.7408], device='cuda:3'), covar=tensor([0.0292, 0.0389, 0.1555, 0.0182, 0.0789, 0.0490, 0.1362, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0166, 0.0176, 0.0216, 0.0201, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:08:35,279 INFO [train.py:904] (3/8) Epoch 24, batch 5500, loss[loss=0.2349, simple_loss=0.3178, pruned_loss=0.076, over 15247.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2754, pruned_loss=0.04777, over 3170524.96 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:09:43,408 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238997.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:09:53,111 INFO [train.py:904] (3/8) Epoch 24, batch 5550, loss[loss=0.2543, simple_loss=0.3254, pruned_loss=0.09159, over 11112.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2821, pruned_loss=0.0526, over 3141727.13 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:10:06,790 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5950, 2.7719, 2.8532, 4.5633, 2.5196, 3.0425, 2.7417, 2.9200], device='cuda:3'), covar=tensor([0.1227, 0.2918, 0.2423, 0.0414, 0.3644, 0.2027, 0.3010, 0.2739], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0456, 0.0373, 0.0329, 0.0436, 0.0523, 0.0427, 0.0532], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:10:27,346 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3058, 3.7313, 3.7082, 2.3083, 3.1479, 2.5348, 3.7182, 4.0491], device='cuda:3'), covar=tensor([0.0350, 0.0820, 0.0606, 0.1998, 0.0911, 0.0950, 0.0763, 0.0938], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0154, 0.0146, 0.0130, 0.0143, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:10:30,506 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.032e+02 3.628e+02 4.510e+02 7.943e+02, threshold=7.255e+02, percent-clipped=11.0 2023-05-01 20:10:48,344 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 5600, loss[loss=0.1944, simple_loss=0.2819, pruned_loss=0.05342, over 17028.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2868, pruned_loss=0.05695, over 3095402.51 frames. ], batch size: 55, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:11:52,285 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239076.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 20:12:26,318 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 20:12:36,910 INFO [train.py:904] (3/8) Epoch 24, batch 5650, loss[loss=0.2301, simple_loss=0.3097, pruned_loss=0.07519, over 16851.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2911, pruned_loss=0.06024, over 3084543.27 frames. ], batch size: 116, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:12:54,594 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-05-01 20:13:11,111 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239124.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:13:14,764 INFO [optim.py:368] (3/8) 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 20:13:46,692 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7938, 5.2034, 5.3721, 5.0948, 5.1584, 5.7138, 5.1866, 4.9593], device='cuda:3'), covar=tensor([0.1060, 0.1796, 0.2044, 0.1906, 0.2269, 0.0984, 0.1652, 0.2410], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0602, 0.0661, 0.0496, 0.0660, 0.0692, 0.0519, 0.0659], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 20:13:55,700 INFO [train.py:904] (3/8) Epoch 24, batch 5700, loss[loss=0.202, simple_loss=0.2917, pruned_loss=0.0561, over 16668.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2927, pruned_loss=0.06142, over 3091475.80 frames. ], batch size: 57, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:14,247 INFO [train.py:904] (3/8) Epoch 24, batch 5750, loss[loss=0.2522, simple_loss=0.3157, pruned_loss=0.09429, over 11121.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2955, pruned_loss=0.0635, over 3057167.55 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:53,493 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 5800, loss[loss=0.1776, simple_loss=0.2726, pruned_loss=0.04132, over 16484.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2948, pruned_loss=0.06188, over 3051148.59 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:17:40,219 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239292.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:17:56,787 INFO [train.py:904] (3/8) Epoch 24, batch 5850, loss[loss=0.2199, simple_loss=0.3018, pruned_loss=0.06896, over 15417.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2924, pruned_loss=0.0601, over 3065841.51 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:18:33,897 INFO [optim.py:368] (3/8) 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 20:18:34,714 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-05-01 20:18:52,941 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239337.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:19:19,099 INFO [train.py:904] (3/8) Epoch 24, batch 5900, loss[loss=0.195, simple_loss=0.2915, pruned_loss=0.04924, over 16408.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2925, pruned_loss=0.06006, over 3068203.09 frames. ], batch size: 35, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:19:41,076 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.36 vs. limit=5.0 2023-05-01 20:20:14,708 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239385.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:20:42,777 INFO [train.py:904] (3/8) Epoch 24, batch 5950, loss[loss=0.2047, simple_loss=0.2966, pruned_loss=0.05643, over 16416.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2932, pruned_loss=0.05876, over 3084833.14 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:21:21,253 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 6000, loss[loss=0.19, simple_loss=0.2698, pruned_loss=0.05504, over 16782.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2922, pruned_loss=0.05827, over 3096232.36 frames. ], batch size: 62, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:22:03,333 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 20:22:14,270 INFO [train.py:938] (3/8) Epoch 24, validation: loss=0.1493, simple_loss=0.2618, pruned_loss=0.01837, over 944034.00 frames. 2023-05-01 20:22:14,271 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 20:22:36,173 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 20:23:32,268 INFO [train.py:904] (3/8) Epoch 24, batch 6050, loss[loss=0.1838, simple_loss=0.2828, pruned_loss=0.04239, over 16552.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2911, pruned_loss=0.05872, over 3070400.40 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:24:09,878 INFO [optim.py:368] (3/8) 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 20:24:22,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5324, 5.4882, 5.2831, 4.5105, 5.4309, 1.8533, 5.0963, 4.9287], device='cuda:3'), covar=tensor([0.0112, 0.0108, 0.0232, 0.0432, 0.0117, 0.2835, 0.0178, 0.0249], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0164, 0.0205, 0.0181, 0.0179, 0.0210, 0.0192, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:24:51,433 INFO [train.py:904] (3/8) Epoch 24, batch 6100, loss[loss=0.1883, simple_loss=0.2773, pruned_loss=0.04968, over 17034.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.29, pruned_loss=0.05744, over 3083781.12 frames. ], batch size: 55, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:24:52,555 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1206, 3.3703, 3.5790, 2.1274, 3.0760, 2.3215, 3.6373, 3.6757], device='cuda:3'), covar=tensor([0.0238, 0.0752, 0.0598, 0.2094, 0.0820, 0.1017, 0.0568, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0155, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:25:08,527 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 20:25:56,122 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239592.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:26:14,091 INFO [train.py:904] (3/8) Epoch 24, batch 6150, loss[loss=0.1777, simple_loss=0.2617, pruned_loss=0.04688, over 16996.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2878, pruned_loss=0.05643, over 3096375.04 frames. ], batch size: 55, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:26:24,853 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8394, 1.4272, 1.7395, 1.7143, 1.7996, 1.9206, 1.6339, 1.7998], device='cuda:3'), covar=tensor([0.0267, 0.0383, 0.0227, 0.0309, 0.0261, 0.0183, 0.0441, 0.0137], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0185, 0.0201, 0.0160, 0.0199, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:26:41,479 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.8184, 6.1854, 5.6404, 6.0693, 5.5977, 5.1770, 5.7280, 6.1768], device='cuda:3'), covar=tensor([0.1956, 0.1173, 0.2206, 0.1287, 0.1360, 0.1176, 0.1985, 0.1571], device='cuda:3'), in_proj_covar=tensor([0.0689, 0.0829, 0.0688, 0.0641, 0.0530, 0.0531, 0.0697, 0.0650], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:26:45,910 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=239622.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:26:53,237 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239640.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:27:34,956 INFO [train.py:904] (3/8) Epoch 24, batch 6200, loss[loss=0.2034, simple_loss=0.2892, pruned_loss=0.05881, over 16442.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2857, pruned_loss=0.05636, over 3068571.92 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:28:13,177 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 20:28:22,547 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=239683.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:28:52,026 INFO [train.py:904] (3/8) Epoch 24, batch 6250, loss[loss=0.231, simple_loss=0.3013, pruned_loss=0.08038, over 11622.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2856, pruned_loss=0.05616, over 3077319.01 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:29:29,985 INFO [optim.py:368] (3/8) 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 20:29:43,151 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0670, 2.5183, 2.5862, 1.9091, 2.7401, 2.7978, 2.4335, 2.3690], device='cuda:3'), covar=tensor([0.0744, 0.0277, 0.0244, 0.1039, 0.0129, 0.0279, 0.0508, 0.0506], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0109, 0.0100, 0.0139, 0.0083, 0.0129, 0.0129, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:30:05,921 INFO [train.py:904] (3/8) Epoch 24, batch 6300, loss[loss=0.1748, simple_loss=0.2652, pruned_loss=0.04224, over 16694.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2857, pruned_loss=0.05582, over 3081060.10 frames. ], batch size: 62, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:30:20,201 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8139, 1.4176, 1.6889, 1.7165, 1.7633, 1.8824, 1.6058, 1.7861], device='cuda:3'), covar=tensor([0.0273, 0.0407, 0.0233, 0.0324, 0.0279, 0.0184, 0.0474, 0.0153], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0194, 0.0182, 0.0185, 0.0202, 0.0160, 0.0200, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:30:47,816 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3258, 2.9382, 3.0152, 2.0081, 2.7494, 2.1769, 3.0332, 3.1999], device='cuda:3'), covar=tensor([0.0307, 0.0805, 0.0680, 0.2149, 0.0931, 0.1132, 0.0718, 0.0834], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0166, 0.0169, 0.0155, 0.0147, 0.0132, 0.0144, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:31:13,288 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 20:31:24,178 INFO [train.py:904] (3/8) Epoch 24, batch 6350, loss[loss=0.1845, simple_loss=0.2645, pruned_loss=0.05222, over 16515.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2876, pruned_loss=0.05768, over 3056822.93 frames. ], batch size: 62, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:25,553 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 20:32:03,925 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.790e+02 3.396e+02 4.270e+02 1.026e+03, threshold=6.791e+02, percent-clipped=1.0 2023-05-01 20:32:15,582 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5770, 4.5199, 4.4524, 3.6842, 4.4944, 1.6592, 4.2529, 4.0151], device='cuda:3'), covar=tensor([0.0108, 0.0082, 0.0189, 0.0358, 0.0092, 0.3064, 0.0134, 0.0283], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0165, 0.0206, 0.0182, 0.0181, 0.0212, 0.0193, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:32:17,534 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3101, 2.3508, 2.3779, 3.9491, 2.1958, 2.6882, 2.3972, 2.4910], device='cuda:3'), covar=tensor([0.1240, 0.3268, 0.2843, 0.0527, 0.3967, 0.2304, 0.3336, 0.3052], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0456, 0.0373, 0.0329, 0.0437, 0.0522, 0.0427, 0.0533], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:32:42,131 INFO [train.py:904] (3/8) Epoch 24, batch 6400, loss[loss=0.2077, simple_loss=0.3003, pruned_loss=0.05756, over 15577.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2879, pruned_loss=0.05844, over 3063629.32 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:32:44,339 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 20:33:58,179 INFO [train.py:904] (3/8) Epoch 24, batch 6450, loss[loss=0.1784, simple_loss=0.2758, pruned_loss=0.04055, over 16876.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2871, pruned_loss=0.05697, over 3085095.64 frames. ], batch size: 96, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:34:37,433 INFO [optim.py:368] (3/8) 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 20:35:16,108 INFO [train.py:904] (3/8) Epoch 24, batch 6500, loss[loss=0.2164, simple_loss=0.3051, pruned_loss=0.06379, over 16188.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2852, pruned_loss=0.05639, over 3082492.33 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:35:55,174 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:36:27,575 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 20:36:39,172 INFO [train.py:904] (3/8) Epoch 24, batch 6550, loss[loss=0.1929, simple_loss=0.2914, pruned_loss=0.04719, over 16903.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.287, pruned_loss=0.05665, over 3102533.34 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:37:16,973 INFO [optim.py:368] (3/8) 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 20:37:36,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8287, 3.9032, 2.5656, 4.5457, 2.9967, 4.4164, 2.4182, 3.1362], device='cuda:3'), covar=tensor([0.0259, 0.0332, 0.1478, 0.0196, 0.0782, 0.0555, 0.1552, 0.0776], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0167, 0.0177, 0.0217, 0.0203, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:37:47,958 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 20:37:54,308 INFO [train.py:904] (3/8) Epoch 24, batch 6600, loss[loss=0.2037, simple_loss=0.2976, pruned_loss=0.05489, over 16414.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2903, pruned_loss=0.05774, over 3085361.10 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:11,701 INFO [train.py:904] (3/8) Epoch 24, batch 6650, loss[loss=0.225, simple_loss=0.2957, pruned_loss=0.07719, over 11352.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2904, pruned_loss=0.05872, over 3074290.69 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:50,364 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 6700, loss[loss=0.1884, simple_loss=0.2786, pruned_loss=0.04908, over 16890.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.289, pruned_loss=0.05818, over 3088592.78 frames. ], batch size: 90, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:41:45,713 INFO [train.py:904] (3/8) Epoch 24, batch 6750, loss[loss=0.1928, simple_loss=0.2806, pruned_loss=0.05248, over 16479.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2879, pruned_loss=0.05824, over 3088685.52 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:42:23,539 INFO [optim.py:368] (3/8) 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 20:43:01,406 INFO [train.py:904] (3/8) Epoch 24, batch 6800, loss[loss=0.2157, simple_loss=0.2941, pruned_loss=0.06862, over 11391.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2887, pruned_loss=0.05855, over 3080740.88 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:43:33,035 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-01 20:43:42,194 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:44:06,655 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1191, 5.0694, 4.9280, 4.0640, 4.9870, 1.7802, 4.6816, 4.5937], device='cuda:3'), covar=tensor([0.0174, 0.0156, 0.0230, 0.0565, 0.0146, 0.3030, 0.0259, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0181, 0.0212, 0.0194, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:44:07,916 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9459, 3.0350, 2.7666, 5.0462, 3.7939, 4.2902, 1.8272, 3.0657], device='cuda:3'), covar=tensor([0.1314, 0.0719, 0.1164, 0.0152, 0.0377, 0.0444, 0.1549, 0.0856], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0193, 0.0205, 0.0216, 0.0203, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:44:21,118 INFO [train.py:904] (3/8) Epoch 24, batch 6850, loss[loss=0.224, simple_loss=0.297, pruned_loss=0.07545, over 11550.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2898, pruned_loss=0.05913, over 3083634.75 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:44:56,065 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 20:45:17,476 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:35,444 INFO [train.py:904] (3/8) Epoch 24, batch 6900, loss[loss=0.2338, simple_loss=0.327, pruned_loss=0.07024, over 16795.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.292, pruned_loss=0.05815, over 3103132.05 frames. ], batch size: 124, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:46:26,526 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 20:46:33,699 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 20:46:50,858 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240401.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:46:52,761 INFO [train.py:904] (3/8) Epoch 24, batch 6950, loss[loss=0.1877, simple_loss=0.2782, pruned_loss=0.04863, over 16449.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.294, pruned_loss=0.06007, over 3103024.92 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:47:33,339 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 2.974e+02 3.564e+02 4.384e+02 8.202e+02, threshold=7.128e+02, percent-clipped=2.0 2023-05-01 20:48:07,677 INFO [train.py:904] (3/8) Epoch 24, batch 7000, loss[loss=0.226, simple_loss=0.3132, pruned_loss=0.0694, over 16455.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2942, pruned_loss=0.05938, over 3098540.73 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:48:31,510 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2222, 4.2722, 4.5855, 4.5377, 4.5608, 4.2993, 4.2624, 4.1776], device='cuda:3'), covar=tensor([0.0364, 0.0593, 0.0465, 0.0499, 0.0564, 0.0507, 0.1040, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0415, 0.0464, 0.0451, 0.0416, 0.0497, 0.0472, 0.0555, 0.0377], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 20:48:49,496 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1019, 2.4090, 2.5896, 1.9145, 2.7288, 2.7724, 2.4253, 2.3622], device='cuda:3'), covar=tensor([0.0688, 0.0252, 0.0242, 0.0985, 0.0127, 0.0265, 0.0470, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0138, 0.0082, 0.0128, 0.0129, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:49:13,686 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8724, 2.1331, 2.2324, 3.4526, 2.0091, 2.3733, 2.2281, 2.2370], device='cuda:3'), covar=tensor([0.1576, 0.3432, 0.3116, 0.0705, 0.4494, 0.2593, 0.3504, 0.3571], device='cuda:3'), in_proj_covar=tensor([0.0408, 0.0457, 0.0375, 0.0330, 0.0440, 0.0523, 0.0429, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:49:21,372 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0478, 4.0256, 3.9637, 3.1940, 3.9966, 1.7768, 3.8001, 3.5054], device='cuda:3'), covar=tensor([0.0134, 0.0116, 0.0205, 0.0334, 0.0105, 0.3012, 0.0144, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0181, 0.0213, 0.0194, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:49:23,902 INFO [train.py:904] (3/8) Epoch 24, batch 7050, loss[loss=0.2172, simple_loss=0.2883, pruned_loss=0.07304, over 11420.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2943, pruned_loss=0.05951, over 3078089.63 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:49:24,945 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3175, 4.3510, 4.6660, 4.6185, 4.6489, 4.3579, 4.3378, 4.2713], device='cuda:3'), covar=tensor([0.0333, 0.0529, 0.0399, 0.0417, 0.0498, 0.0402, 0.0936, 0.0546], device='cuda:3'), in_proj_covar=tensor([0.0416, 0.0465, 0.0451, 0.0416, 0.0498, 0.0473, 0.0555, 0.0377], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 20:50:06,655 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.536e+02 3.115e+02 4.008e+02 6.614e+02, threshold=6.229e+02, percent-clipped=0.0 2023-05-01 20:50:42,265 INFO [train.py:904] (3/8) Epoch 24, batch 7100, loss[loss=0.2255, simple_loss=0.3032, pruned_loss=0.0739, over 15448.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2923, pruned_loss=0.05882, over 3087803.96 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:51:59,029 INFO [train.py:904] (3/8) Epoch 24, batch 7150, loss[loss=0.1982, simple_loss=0.2856, pruned_loss=0.0554, over 15328.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2905, pruned_loss=0.05882, over 3079594.39 frames. ], batch size: 190, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:52:39,164 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.962e+02 2.941e+02 3.518e+02 4.081e+02 6.999e+02, threshold=7.036e+02, percent-clipped=1.0 2023-05-01 20:52:49,513 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 20:53:12,643 INFO [train.py:904] (3/8) Epoch 24, batch 7200, loss[loss=0.1935, simple_loss=0.2774, pruned_loss=0.05476, over 11994.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2886, pruned_loss=0.05752, over 3061809.67 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:54:20,531 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=240696.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:54:32,008 INFO [train.py:904] (3/8) Epoch 24, batch 7250, loss[loss=0.1954, simple_loss=0.2905, pruned_loss=0.05017, over 16177.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2865, pruned_loss=0.0566, over 3063058.15 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:54:47,628 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 20:55:12,201 INFO [optim.py:368] (3/8) 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 20:55:44,706 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 20:55:45,057 INFO [train.py:904] (3/8) Epoch 24, batch 7300, loss[loss=0.1939, simple_loss=0.2822, pruned_loss=0.05282, over 16624.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2861, pruned_loss=0.05658, over 3072243.26 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:56:49,452 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 20:56:58,429 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5915, 5.9234, 5.6397, 5.7125, 5.2913, 5.2027, 5.3370, 6.0287], device='cuda:3'), covar=tensor([0.1208, 0.0718, 0.1033, 0.0804, 0.0880, 0.0778, 0.1090, 0.0771], device='cuda:3'), in_proj_covar=tensor([0.0689, 0.0827, 0.0686, 0.0641, 0.0528, 0.0533, 0.0698, 0.0651], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:57:02,841 INFO [train.py:904] (3/8) Epoch 24, batch 7350, loss[loss=0.2031, simple_loss=0.2882, pruned_loss=0.05902, over 15288.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.287, pruned_loss=0.05739, over 3059589.44 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:09,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1349, 3.6416, 3.5831, 2.2881, 3.3129, 3.5989, 3.2627, 2.0381], device='cuda:3'), covar=tensor([0.0551, 0.0056, 0.0067, 0.0441, 0.0119, 0.0132, 0.0125, 0.0468], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0086, 0.0086, 0.0134, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 20:57:44,658 INFO [optim.py:368] (3/8) 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 20:58:04,287 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3234, 2.1034, 1.7260, 1.9205, 2.3531, 2.0441, 2.0502, 2.4844], device='cuda:3'), covar=tensor([0.0221, 0.0462, 0.0595, 0.0494, 0.0280, 0.0408, 0.0212, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0211, 0.0232, 0.0225, 0.0225, 0.0234, 0.0231, 0.0231, 0.0230], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 20:58:18,665 INFO [train.py:904] (3/8) Epoch 24, batch 7400, loss[loss=0.214, simple_loss=0.3051, pruned_loss=0.06145, over 16732.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.288, pruned_loss=0.05785, over 3068745.40 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:58:40,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7115, 3.0965, 2.9714, 5.1803, 3.9132, 4.3326, 1.8201, 3.1709], device='cuda:3'), covar=tensor([0.1464, 0.0716, 0.1078, 0.0139, 0.0349, 0.0398, 0.1613, 0.0804], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0194, 0.0206, 0.0217, 0.0206, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 20:59:01,100 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 20:59:34,863 INFO [train.py:904] (3/8) Epoch 24, batch 7450, loss[loss=0.1842, simple_loss=0.27, pruned_loss=0.04922, over 16303.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2889, pruned_loss=0.05882, over 3063506.16 frames. ], batch size: 35, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:00:19,560 INFO [optim.py:368] (3/8) 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 21:00:22,564 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240932.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:00:55,587 INFO [train.py:904] (3/8) Epoch 24, batch 7500, loss[loss=0.2071, simple_loss=0.2968, pruned_loss=0.05876, over 16409.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2891, pruned_loss=0.05795, over 3064325.93 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:01:03,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0607, 4.0496, 3.9639, 3.2201, 4.0151, 1.7496, 3.8238, 3.5446], device='cuda:3'), covar=tensor([0.0133, 0.0115, 0.0210, 0.0323, 0.0098, 0.2846, 0.0133, 0.0269], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0163, 0.0205, 0.0181, 0.0179, 0.0211, 0.0192, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:01:45,350 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240986.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:01:57,387 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240993.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:02:00,888 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240996.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:02:11,648 INFO [train.py:904] (3/8) Epoch 24, batch 7550, loss[loss=0.1674, simple_loss=0.2654, pruned_loss=0.03471, over 16924.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.288, pruned_loss=0.05807, over 3066805.59 frames. ], batch size: 90, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:02:53,682 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:20,237 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241047.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:29,245 INFO [train.py:904] (3/8) Epoch 24, batch 7600, loss[loss=0.2412, simple_loss=0.3079, pruned_loss=0.08723, over 11217.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2872, pruned_loss=0.05791, over 3078101.42 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:04:47,281 INFO [train.py:904] (3/8) Epoch 24, batch 7650, loss[loss=0.2494, simple_loss=0.3136, pruned_loss=0.09262, over 11147.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2877, pruned_loss=0.05823, over 3087036.72 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:05:30,616 INFO [optim.py:368] (3/8) 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 21:06:04,717 INFO [train.py:904] (3/8) Epoch 24, batch 7700, loss[loss=0.1959, simple_loss=0.297, pruned_loss=0.04738, over 16878.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2877, pruned_loss=0.05844, over 3089469.41 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:06:08,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-01 21:06:28,216 INFO [scaling.py:679] (3/8) 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] (3/8) Epoch 24, batch 7750, loss[loss=0.206, simple_loss=0.3008, pruned_loss=0.05562, over 16678.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2878, pruned_loss=0.05798, over 3093417.51 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:06,386 INFO [optim.py:368] (3/8) 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 21:08:11,957 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3104, 3.4559, 3.6113, 3.5801, 3.5924, 3.4322, 3.4500, 3.4866], device='cuda:3'), covar=tensor([0.0422, 0.0731, 0.0508, 0.0455, 0.0574, 0.0546, 0.0853, 0.0580], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0467, 0.0452, 0.0418, 0.0498, 0.0473, 0.0556, 0.0380], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 21:08:19,167 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-05-01 21:08:27,258 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 21:08:38,514 INFO [train.py:904] (3/8) Epoch 24, batch 7800, loss[loss=0.1851, simple_loss=0.267, pruned_loss=0.05156, over 17266.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2885, pruned_loss=0.05836, over 3104576.93 frames. ], batch size: 52, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:55,431 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8378, 4.6545, 4.8414, 5.0336, 5.2468, 4.6985, 5.2234, 5.2161], device='cuda:3'), covar=tensor([0.2031, 0.1332, 0.1801, 0.0853, 0.0735, 0.0994, 0.0796, 0.0764], device='cuda:3'), in_proj_covar=tensor([0.0635, 0.0787, 0.0901, 0.0790, 0.0607, 0.0626, 0.0658, 0.0763], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:09:34,093 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 7850, loss[loss=0.2284, simple_loss=0.3062, pruned_loss=0.07534, over 11551.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2898, pruned_loss=0.05862, over 3092498.16 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:10:30,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9196, 2.6804, 2.6241, 1.9978, 2.5775, 2.7162, 2.5670, 1.8886], device='cuda:3'), covar=tensor([0.0482, 0.0111, 0.0102, 0.0395, 0.0149, 0.0148, 0.0132, 0.0456], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0086, 0.0086, 0.0135, 0.0099, 0.0110, 0.0095, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-01 21:10:38,476 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) Epoch 24, batch 7900, loss[loss=0.1849, simple_loss=0.2773, pruned_loss=0.04618, over 17117.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2891, pruned_loss=0.05823, over 3093982.93 frames. ], batch size: 49, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:11:19,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8803, 3.5266, 4.0889, 2.1627, 4.2897, 4.3004, 3.0804, 3.2384], device='cuda:3'), covar=tensor([0.0714, 0.0287, 0.0189, 0.1118, 0.0068, 0.0178, 0.0445, 0.0429], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0109, 0.0100, 0.0139, 0.0083, 0.0128, 0.0129, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 21:12:29,004 INFO [train.py:904] (3/8) Epoch 24, batch 7950, loss[loss=0.2038, simple_loss=0.2889, pruned_loss=0.05929, over 16463.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.289, pruned_loss=0.05809, over 3111372.20 frames. ], batch size: 75, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:35,108 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:13:13,216 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4300, 3.4215, 2.6171, 2.1236, 2.2282, 2.2352, 3.6240, 3.1186], device='cuda:3'), covar=tensor([0.3294, 0.0710, 0.2112, 0.3003, 0.3059, 0.2439, 0.0531, 0.1376], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0272, 0.0310, 0.0321, 0.0302, 0.0268, 0.0301, 0.0344], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:13:22,211 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-05-01 21:13:40,317 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9353, 2.7880, 2.8155, 2.2177, 2.7028, 2.1698, 2.7395, 2.9622], device='cuda:3'), covar=tensor([0.0259, 0.0754, 0.0472, 0.1602, 0.0762, 0.0932, 0.0533, 0.0654], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0165, 0.0167, 0.0154, 0.0145, 0.0130, 0.0143, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 21:13:46,678 INFO [train.py:904] (3/8) Epoch 24, batch 8000, loss[loss=0.1764, simple_loss=0.2718, pruned_loss=0.04049, over 16899.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2896, pruned_loss=0.05896, over 3088019.04 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:14:09,536 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 8050, loss[loss=0.1924, simple_loss=0.2838, pruned_loss=0.05044, over 16393.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2891, pruned_loss=0.05824, over 3091775.67 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:15:47,626 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 8100, loss[loss=0.1774, simple_loss=0.2628, pruned_loss=0.04606, over 16868.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2888, pruned_loss=0.05803, over 3095783.00 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:16:50,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7144, 2.6173, 2.5614, 4.0344, 2.9248, 3.9448, 1.5589, 2.8788], device='cuda:3'), covar=tensor([0.1368, 0.0795, 0.1214, 0.0190, 0.0230, 0.0417, 0.1736, 0.0864], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0193, 0.0205, 0.0216, 0.0204, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 21:17:14,418 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241588.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:17:38,324 INFO [train.py:904] (3/8) Epoch 24, batch 8150, loss[loss=0.1724, simple_loss=0.2605, pruned_loss=0.04211, over 16964.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2864, pruned_loss=0.05665, over 3104083.60 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:17:39,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 21:18:06,798 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:22,058 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241636.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:18:39,461 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241642.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:55,841 INFO [train.py:904] (3/8) Epoch 24, batch 8200, loss[loss=0.182, simple_loss=0.2766, pruned_loss=0.04366, over 15319.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2829, pruned_loss=0.05522, over 3116501.54 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:19:20,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2273, 4.3214, 4.4757, 4.1947, 4.3380, 4.8306, 4.3972, 4.0694], device='cuda:3'), covar=tensor([0.1788, 0.2117, 0.2322, 0.2279, 0.2750, 0.1102, 0.1643, 0.2754], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0612, 0.0676, 0.0503, 0.0663, 0.0698, 0.0525, 0.0669], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:19:35,520 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 21:19:42,939 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:19:44,575 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2252, 5.7372, 5.9728, 5.5664, 5.7685, 6.2528, 5.7516, 5.5212], device='cuda:3'), covar=tensor([0.0834, 0.1728, 0.2665, 0.1934, 0.2128, 0.0811, 0.1576, 0.2329], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0611, 0.0675, 0.0502, 0.0663, 0.0697, 0.0524, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:19:56,098 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241690.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:20:16,207 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 21:20:16,378 INFO [train.py:904] (3/8) Epoch 24, batch 8250, loss[loss=0.1811, simple_loss=0.2766, pruned_loss=0.04278, over 16858.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.282, pruned_loss=0.05317, over 3099352.01 frames. ], batch size: 116, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:03,231 INFO [optim.py:368] (3/8) 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 21:21:39,002 INFO [train.py:904] (3/8) Epoch 24, batch 8300, loss[loss=0.1931, simple_loss=0.2899, pruned_loss=0.04811, over 16682.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2797, pruned_loss=0.05061, over 3086608.28 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:54,684 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241762.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:22:29,314 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 21:22:52,817 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9609, 4.2470, 4.0886, 4.1140, 3.7574, 3.8979, 3.8698, 4.2526], device='cuda:3'), covar=tensor([0.1197, 0.1005, 0.1027, 0.0883, 0.0911, 0.1523, 0.1024, 0.1048], device='cuda:3'), in_proj_covar=tensor([0.0688, 0.0822, 0.0686, 0.0640, 0.0524, 0.0530, 0.0695, 0.0646], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:23:01,943 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2711, 4.3324, 4.4649, 4.2288, 4.3366, 4.8368, 4.3922, 4.0525], device='cuda:3'), covar=tensor([0.1577, 0.1984, 0.2232, 0.2254, 0.2689, 0.1100, 0.1665, 0.2646], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0604, 0.0668, 0.0496, 0.0655, 0.0691, 0.0519, 0.0661], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:23:02,690 INFO [train.py:904] (3/8) Epoch 24, batch 8350, loss[loss=0.1842, simple_loss=0.2802, pruned_loss=0.04416, over 15367.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2793, pruned_loss=0.04902, over 3069602.80 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:23:03,883 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:23:48,823 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241831.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:24:19,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8288, 3.9307, 4.0289, 3.7711, 3.9218, 4.3403, 3.9725, 3.6326], device='cuda:3'), covar=tensor([0.2270, 0.2206, 0.2395, 0.2629, 0.2942, 0.1660, 0.1565, 0.2874], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0599, 0.0662, 0.0492, 0.0649, 0.0686, 0.0515, 0.0656], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:24:23,521 INFO [train.py:904] (3/8) Epoch 24, batch 8400, loss[loss=0.1843, simple_loss=0.2734, pruned_loss=0.04756, over 16571.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2765, pruned_loss=0.04729, over 3044017.00 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:24:42,782 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241864.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:25:24,484 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0963, 5.0598, 4.8110, 4.2504, 4.9133, 1.9522, 4.6432, 4.6803], device='cuda:3'), covar=tensor([0.0111, 0.0134, 0.0249, 0.0443, 0.0129, 0.2787, 0.0158, 0.0253], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0162, 0.0202, 0.0178, 0.0177, 0.0208, 0.0190, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:25:28,799 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241892.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:25:45,953 INFO [train.py:904] (3/8) Epoch 24, batch 8450, loss[loss=0.1866, simple_loss=0.2802, pruned_loss=0.0465, over 15383.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2749, pruned_loss=0.04609, over 3032609.05 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:26:31,270 INFO [optim.py:368] (3/8) 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 21:26:44,746 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0162, 2.2506, 2.3386, 2.9647, 1.7478, 3.2809, 1.7325, 2.7828], device='cuda:3'), covar=tensor([0.1159, 0.0671, 0.0952, 0.0173, 0.0083, 0.0360, 0.1510, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0174, 0.0194, 0.0190, 0.0202, 0.0213, 0.0203, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 21:27:07,284 INFO [train.py:904] (3/8) Epoch 24, batch 8500, loss[loss=0.1556, simple_loss=0.2557, pruned_loss=0.02772, over 16478.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2714, pruned_loss=0.04378, over 3038621.53 frames. ], batch size: 68, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:27:47,067 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:28:05,357 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 21:28:33,821 INFO [train.py:904] (3/8) Epoch 24, batch 8550, loss[loss=0.1746, simple_loss=0.2729, pruned_loss=0.03818, over 15385.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2695, pruned_loss=0.04286, over 3022169.97 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:29:02,024 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 21:29:26,915 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.227e+02 2.590e+02 3.113e+02 4.507e+02, threshold=5.180e+02, percent-clipped=0.0 2023-05-01 21:30:13,213 INFO [train.py:904] (3/8) Epoch 24, batch 8600, loss[loss=0.1521, simple_loss=0.2532, pruned_loss=0.02556, over 16442.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2698, pruned_loss=0.04216, over 3013269.44 frames. ], batch size: 75, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:30:31,472 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:31:51,604 INFO [train.py:904] (3/8) Epoch 24, batch 8650, loss[loss=0.1719, simple_loss=0.2677, pruned_loss=0.03807, over 16778.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2681, pruned_loss=0.04067, over 3018598.72 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:32:10,578 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:33:05,317 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-05-01 21:33:37,183 INFO [train.py:904] (3/8) Epoch 24, batch 8700, loss[loss=0.1664, simple_loss=0.2517, pruned_loss=0.04058, over 12650.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2656, pruned_loss=0.03966, over 3040313.03 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:33:38,218 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242153.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:33:49,873 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:33:57,019 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 21:34:42,451 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242187.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:35:13,443 INFO [train.py:904] (3/8) Epoch 24, batch 8750, loss[loss=0.1856, simple_loss=0.2798, pruned_loss=0.04573, over 15207.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.266, pruned_loss=0.03919, over 3054590.25 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:35:42,294 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:10,037 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:21,484 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.077e+02 2.590e+02 3.155e+02 6.830e+02, threshold=5.181e+02, percent-clipped=2.0 2023-05-01 21:37:05,080 INFO [train.py:904] (3/8) Epoch 24, batch 8800, loss[loss=0.1622, simple_loss=0.2679, pruned_loss=0.02827, over 16908.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2644, pruned_loss=0.03793, over 3058231.46 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:37:56,098 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242277.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:17,986 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242287.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:18,285 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 21:38:47,695 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 21:38:50,077 INFO [train.py:904] (3/8) Epoch 24, batch 8850, loss[loss=0.1532, simple_loss=0.2529, pruned_loss=0.02676, over 16891.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2661, pruned_loss=0.03714, over 3041603.07 frames. ], batch size: 42, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:39:03,869 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242309.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:38,582 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:52,554 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 24, batch 8900, loss[loss=0.1552, simple_loss=0.2518, pruned_loss=0.02932, over 16727.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2666, pruned_loss=0.03673, over 3048387.29 frames. ], batch size: 76, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:40:53,119 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2436, 4.5958, 4.2009, 4.4747, 4.1771, 4.0810, 4.1427, 4.6017], device='cuda:3'), covar=tensor([0.2354, 0.1612, 0.2198, 0.1397, 0.1498, 0.2563, 0.2284, 0.1944], device='cuda:3'), in_proj_covar=tensor([0.0675, 0.0808, 0.0672, 0.0628, 0.0515, 0.0522, 0.0682, 0.0636], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:41:12,532 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:42:44,441 INFO [train.py:904] (3/8) Epoch 24, batch 8950, loss[loss=0.1379, simple_loss=0.2383, pruned_loss=0.01875, over 16860.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2666, pruned_loss=0.03735, over 3046596.54 frames. ], batch size: 90, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:43:49,726 INFO [optim.py:368] (3/8) 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 21:43:53,569 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4613, 3.3892, 3.5225, 3.5851, 3.6331, 3.3329, 3.5707, 3.6762], device='cuda:3'), covar=tensor([0.1319, 0.1107, 0.1024, 0.0623, 0.0594, 0.2255, 0.0868, 0.0700], device='cuda:3'), in_proj_covar=tensor([0.0623, 0.0771, 0.0883, 0.0778, 0.0595, 0.0616, 0.0645, 0.0749], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:44:03,060 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6174, 3.6653, 3.5131, 3.1972, 3.3050, 3.5864, 3.3616, 3.4370], device='cuda:3'), covar=tensor([0.0534, 0.0525, 0.0296, 0.0257, 0.0501, 0.0454, 0.1304, 0.0483], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0426, 0.0333, 0.0334, 0.0334, 0.0386, 0.0230, 0.0400], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:44:35,817 INFO [train.py:904] (3/8) Epoch 24, batch 9000, loss[loss=0.1645, simple_loss=0.2518, pruned_loss=0.03866, over 12151.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2636, pruned_loss=0.0361, over 3040875.59 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:44:35,818 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 21:44:42,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7201, 4.7012, 4.6107, 4.1853, 4.3264, 4.5936, 4.8212, 4.3451], device='cuda:3'), covar=tensor([0.0464, 0.0365, 0.0265, 0.0338, 0.0682, 0.0459, 0.0133, 0.0686], device='cuda:3'), in_proj_covar=tensor([0.0286, 0.0427, 0.0333, 0.0334, 0.0334, 0.0386, 0.0231, 0.0400], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:44:45,533 INFO [train.py:938] (3/8) Epoch 24, validation: loss=0.1445, simple_loss=0.2484, pruned_loss=0.02026, over 944034.00 frames. 2023-05-01 21:44:45,533 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 21:44:58,950 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:48,512 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:58,851 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242487.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:46:30,342 INFO [train.py:904] (3/8) Epoch 24, batch 9050, loss[loss=0.1662, simple_loss=0.2579, pruned_loss=0.03723, over 16378.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2648, pruned_loss=0.03658, over 3061656.08 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:46:40,246 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:40,473 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0811, 2.3922, 2.3186, 3.0158, 1.7612, 3.3016, 1.8338, 2.8281], device='cuda:3'), covar=tensor([0.1040, 0.0590, 0.1010, 0.0164, 0.0082, 0.0456, 0.1365, 0.0630], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0173, 0.0193, 0.0188, 0.0200, 0.0212, 0.0202, 0.0192], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 21:46:45,332 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:47:30,075 INFO [optim.py:368] (3/8) 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 21:47:34,254 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4175, 4.5134, 4.6849, 4.4421, 4.6086, 5.0611, 4.5984, 4.2612], device='cuda:3'), covar=tensor([0.1377, 0.1950, 0.2131, 0.2136, 0.2186, 0.0915, 0.1612, 0.2559], device='cuda:3'), in_proj_covar=tensor([0.0400, 0.0592, 0.0652, 0.0485, 0.0638, 0.0677, 0.0506, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:47:35,729 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:48:00,769 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242544.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:15,297 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 21:48:17,898 INFO [train.py:904] (3/8) Epoch 24, batch 9100, loss[loss=0.1621, simple_loss=0.2617, pruned_loss=0.03125, over 16919.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2637, pruned_loss=0.03693, over 3046629.28 frames. ], batch size: 116, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:48:23,294 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:56,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2953, 2.2637, 2.1502, 3.9309, 2.1476, 2.5805, 2.3013, 2.3671], device='cuda:3'), covar=tensor([0.1229, 0.3782, 0.3436, 0.0564, 0.4404, 0.2720, 0.3905, 0.3506], device='cuda:3'), in_proj_covar=tensor([0.0399, 0.0447, 0.0368, 0.0322, 0.0432, 0.0510, 0.0419, 0.0521], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:49:27,710 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242582.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,299 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242602.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,967 INFO [train.py:904] (3/8) Epoch 24, batch 9150, loss[loss=0.1632, simple_loss=0.2482, pruned_loss=0.03911, over 12367.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2643, pruned_loss=0.03668, over 3045867.78 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:50:36,923 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 21:50:47,225 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:51:21,272 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.227e+02 2.578e+02 3.148e+02 6.071e+02, threshold=5.155e+02, percent-clipped=3.0 2023-05-01 21:52:01,740 INFO [train.py:904] (3/8) Epoch 24, batch 9200, loss[loss=0.1592, simple_loss=0.2493, pruned_loss=0.03456, over 16794.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2597, pruned_loss=0.03561, over 3044765.61 frames. ], batch size: 124, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:52:22,364 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:24,774 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:53:09,846 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5287, 3.4945, 2.8146, 2.1663, 2.2140, 2.3723, 3.7574, 3.0939], device='cuda:3'), covar=tensor([0.3076, 0.0799, 0.1853, 0.3268, 0.3397, 0.2268, 0.0411, 0.1608], device='cuda:3'), in_proj_covar=tensor([0.0323, 0.0265, 0.0302, 0.0313, 0.0291, 0.0262, 0.0292, 0.0334], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 21:53:38,106 INFO [train.py:904] (3/8) Epoch 24, batch 9250, loss[loss=0.1593, simple_loss=0.2503, pruned_loss=0.03413, over 16453.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.26, pruned_loss=0.03568, over 3070992.07 frames. ], batch size: 62, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:54:40,869 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.148e+02 2.469e+02 2.996e+02 5.681e+02, threshold=4.938e+02, percent-clipped=3.0 2023-05-01 21:54:58,105 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8908, 1.4375, 1.7862, 1.7918, 1.9107, 1.9681, 1.7474, 1.8699], device='cuda:3'), covar=tensor([0.0339, 0.0453, 0.0265, 0.0344, 0.0348, 0.0248, 0.0492, 0.0159], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0190, 0.0178, 0.0180, 0.0195, 0.0155, 0.0194, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:55:16,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5758, 1.8373, 2.1555, 2.5488, 2.4743, 2.7947, 1.9589, 2.7697], device='cuda:3'), covar=tensor([0.0256, 0.0602, 0.0425, 0.0389, 0.0392, 0.0251, 0.0638, 0.0201], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0189, 0.0178, 0.0179, 0.0195, 0.0155, 0.0194, 0.0152], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:55:27,519 INFO [train.py:904] (3/8) Epoch 24, batch 9300, loss[loss=0.1553, simple_loss=0.2432, pruned_loss=0.03372, over 12108.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2584, pruned_loss=0.03508, over 3060673.93 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:56:13,793 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:56:52,244 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4895, 4.2929, 4.5487, 4.6766, 4.8354, 4.3808, 4.8295, 4.8795], device='cuda:3'), covar=tensor([0.1951, 0.1306, 0.1702, 0.0838, 0.0566, 0.1011, 0.0620, 0.0668], device='cuda:3'), in_proj_covar=tensor([0.0622, 0.0767, 0.0882, 0.0776, 0.0596, 0.0612, 0.0645, 0.0746], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:57:14,573 INFO [train.py:904] (3/8) Epoch 24, batch 9350, loss[loss=0.1545, simple_loss=0.2523, pruned_loss=0.02837, over 16840.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2587, pruned_loss=0.03518, over 3070629.11 frames. ], batch size: 102, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:57:27,040 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242809.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:13,360 INFO [optim.py:368] (3/8) 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 21:58:14,174 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242839.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:34,215 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0709, 4.0449, 3.9595, 3.2660, 3.9836, 1.8509, 3.7436, 3.5287], device='cuda:3'), covar=tensor([0.0115, 0.0117, 0.0194, 0.0243, 0.0096, 0.2775, 0.0125, 0.0254], device='cuda:3'), in_proj_covar=tensor([0.0165, 0.0158, 0.0196, 0.0172, 0.0173, 0.0203, 0.0185, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 21:58:54,546 INFO [train.py:904] (3/8) Epoch 24, batch 9400, loss[loss=0.134, simple_loss=0.2235, pruned_loss=0.02226, over 12615.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2583, pruned_loss=0.03488, over 3075716.75 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:59:03,676 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:59:55,055 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 9450, loss[loss=0.168, simple_loss=0.2667, pruned_loss=0.03465, over 16242.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2602, pruned_loss=0.03515, over 3067900.19 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:00:51,497 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:13,393 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:32,437 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 24, batch 9500, loss[loss=0.16, simple_loss=0.2669, pruned_loss=0.02654, over 16921.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2596, pruned_loss=0.03487, over 3074529.10 frames. ], batch size: 102, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:02:18,370 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-05-01 22:02:30,099 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:39,676 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 22:02:43,723 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:03:15,715 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242982.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 22:04:01,164 INFO [train.py:904] (3/8) Epoch 24, batch 9550, loss[loss=0.1418, simple_loss=0.2389, pruned_loss=0.02238, over 16517.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2589, pruned_loss=0.03465, over 3076531.31 frames. ], batch size: 68, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:04:24,732 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:05:43,248 INFO [train.py:904] (3/8) Epoch 24, batch 9600, loss[loss=0.1735, simple_loss=0.2623, pruned_loss=0.0423, over 12304.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2605, pruned_loss=0.03586, over 3065541.07 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:06:40,966 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2583, 3.2730, 1.9950, 3.6512, 2.4149, 3.5853, 2.2328, 2.7293], device='cuda:3'), covar=tensor([0.0349, 0.0448, 0.1776, 0.0279, 0.0999, 0.0607, 0.1647, 0.0861], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0171, 0.0188, 0.0159, 0.0172, 0.0207, 0.0197, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 22:06:50,220 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-05-01 22:07:32,090 INFO [train.py:904] (3/8) Epoch 24, batch 9650, loss[loss=0.1604, simple_loss=0.2583, pruned_loss=0.03122, over 15456.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2627, pruned_loss=0.03595, over 3090176.62 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:08:29,798 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243128.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:08:29,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8594, 3.3386, 3.4764, 2.0160, 2.9618, 2.3002, 3.2124, 3.4279], device='cuda:3'), covar=tensor([0.0406, 0.0842, 0.0514, 0.2160, 0.0825, 0.0980, 0.0849, 0.0953], device='cuda:3'), in_proj_covar=tensor([0.0152, 0.0158, 0.0163, 0.0150, 0.0141, 0.0127, 0.0139, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 22:08:37,734 INFO [optim.py:368] (3/8) 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 22:08:48,832 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-01 22:08:50,109 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243139.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:09:15,520 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-05-01 22:09:19,548 INFO [train.py:904] (3/8) Epoch 24, batch 9700, loss[loss=0.1652, simple_loss=0.2516, pruned_loss=0.03935, over 12098.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2618, pruned_loss=0.03594, over 3094766.53 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:09:24,894 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 22:10:30,926 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:01,230 INFO [train.py:904] (3/8) Epoch 24, batch 9750, loss[loss=0.1734, simple_loss=0.2707, pruned_loss=0.03804, over 16424.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.26, pruned_loss=0.03587, over 3075748.70 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:11:01,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4404, 3.0391, 2.7521, 2.2762, 2.2230, 2.3018, 2.9701, 2.8773], device='cuda:3'), covar=tensor([0.2526, 0.0721, 0.1607, 0.2806, 0.2584, 0.2214, 0.0462, 0.1409], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0263, 0.0301, 0.0311, 0.0288, 0.0260, 0.0291, 0.0332], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 22:11:16,971 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8080, 3.6146, 3.7543, 3.9374, 3.9978, 3.6534, 4.0203, 4.0417], device='cuda:3'), covar=tensor([0.1501, 0.1323, 0.1670, 0.0855, 0.0804, 0.2121, 0.0913, 0.0942], device='cuda:3'), in_proj_covar=tensor([0.0619, 0.0763, 0.0879, 0.0773, 0.0592, 0.0609, 0.0641, 0.0743], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:11:16,987 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:03,357 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.977e+02 2.392e+02 3.038e+02 5.323e+02, threshold=4.783e+02, percent-clipped=0.0 2023-05-01 22:12:37,393 INFO [train.py:904] (3/8) Epoch 24, batch 9800, loss[loss=0.1798, simple_loss=0.2847, pruned_loss=0.03743, over 16803.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2603, pruned_loss=0.03506, over 3090288.58 frames. ], batch size: 124, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:12:49,024 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:50,825 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:13:22,144 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243277.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:13:58,472 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=243293.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:21,694 INFO [train.py:904] (3/8) Epoch 24, batch 9850, loss[loss=0.1756, simple_loss=0.2671, pruned_loss=0.04208, over 16950.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2614, pruned_loss=0.03491, over 3093826.37 frames. ], batch size: 109, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:14:28,946 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243306.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:15:00,643 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1770, 1.5690, 1.9354, 2.1440, 2.2030, 2.3572, 1.8686, 2.2581], device='cuda:3'), covar=tensor([0.0256, 0.0545, 0.0321, 0.0352, 0.0374, 0.0224, 0.0515, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0188, 0.0176, 0.0177, 0.0193, 0.0153, 0.0192, 0.0151], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:15:22,938 INFO [optim.py:368] (3/8) 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 22:16:01,382 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-05-01 22:16:11,622 INFO [train.py:904] (3/8) Epoch 24, batch 9900, loss[loss=0.156, simple_loss=0.2625, pruned_loss=0.02473, over 15288.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2613, pruned_loss=0.03473, over 3064969.21 frames. ], batch size: 191, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:16:15,167 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 24, batch 9950, loss[loss=0.1908, simple_loss=0.2748, pruned_loss=0.05343, over 12274.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2638, pruned_loss=0.03523, over 3054694.35 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:19:12,023 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:19:24,674 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.110e+02 2.470e+02 3.179e+02 5.008e+02, threshold=4.941e+02, percent-clipped=4.0 2023-05-01 22:20:10,933 INFO [train.py:904] (3/8) Epoch 24, batch 10000, loss[loss=0.1664, simple_loss=0.2524, pruned_loss=0.04016, over 12727.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.262, pruned_loss=0.03434, over 3090600.64 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:20:55,587 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:21:51,122 INFO [train.py:904] (3/8) Epoch 24, batch 10050, loss[loss=0.1701, simple_loss=0.2577, pruned_loss=0.04128, over 11776.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2622, pruned_loss=0.03426, over 3099533.57 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:22:52,099 INFO [optim.py:368] (3/8) 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 22:23:15,543 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7550, 3.6143, 3.7462, 3.9269, 3.9850, 3.6737, 4.0223, 4.0404], device='cuda:3'), covar=tensor([0.1810, 0.1622, 0.1834, 0.1010, 0.0988, 0.2268, 0.1180, 0.1133], device='cuda:3'), in_proj_covar=tensor([0.0615, 0.0757, 0.0871, 0.0768, 0.0588, 0.0605, 0.0635, 0.0737], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:23:25,157 INFO [train.py:904] (3/8) Epoch 24, batch 10100, loss[loss=0.1584, simple_loss=0.2544, pruned_loss=0.03119, over 16263.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2619, pruned_loss=0.03431, over 3085999.19 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:24:16,191 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243577.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 22:24:28,542 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9936, 4.2314, 4.0828, 4.0840, 3.7946, 3.8527, 3.8255, 4.2174], device='cuda:3'), covar=tensor([0.1036, 0.0850, 0.0856, 0.0755, 0.0735, 0.1603, 0.1000, 0.0965], device='cuda:3'), in_proj_covar=tensor([0.0666, 0.0802, 0.0660, 0.0620, 0.0508, 0.0516, 0.0672, 0.0628], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:25:09,202 INFO [train.py:904] (3/8) Epoch 25, batch 0, loss[loss=0.1662, simple_loss=0.2602, pruned_loss=0.03616, over 17223.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2602, pruned_loss=0.03616, over 17223.00 frames. ], batch size: 52, lr: 2.72e-03, grad_scale: 8.0 2023-05-01 22:25:09,202 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 22:25:16,828 INFO [train.py:938] (3/8) Epoch 25, validation: loss=0.1443, simple_loss=0.2477, pruned_loss=0.02048, over 944034.00 frames. 2023-05-01 22:25:16,828 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 22:25:48,588 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:25:52,539 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8584, 4.8293, 5.2241, 5.1825, 5.2764, 4.9468, 4.8874, 4.6412], device='cuda:3'), covar=tensor([0.0382, 0.0617, 0.0464, 0.0513, 0.0573, 0.0481, 0.1115, 0.0553], device='cuda:3'), in_proj_covar=tensor([0.0397, 0.0443, 0.0435, 0.0400, 0.0476, 0.0454, 0.0528, 0.0364], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 22:26:03,178 INFO [optim.py:368] (3/8) 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 22:26:21,433 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243649.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:26,914 INFO [train.py:904] (3/8) Epoch 25, batch 50, loss[loss=0.1894, simple_loss=0.271, pruned_loss=0.05383, over 16815.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2654, pruned_loss=0.04412, over 755061.30 frames. ], batch size: 102, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:26:30,713 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 22:26:42,164 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8422, 4.8268, 4.7211, 4.3718, 4.7685, 1.9844, 4.5444, 4.4936], device='cuda:3'), covar=tensor([0.0148, 0.0124, 0.0234, 0.0316, 0.0141, 0.2695, 0.0168, 0.0237], device='cuda:3'), in_proj_covar=tensor([0.0166, 0.0158, 0.0196, 0.0171, 0.0174, 0.0205, 0.0186, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:27:04,763 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 22:27:35,801 INFO [train.py:904] (3/8) Epoch 25, batch 100, loss[loss=0.1792, simple_loss=0.2544, pruned_loss=0.05203, over 16832.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2652, pruned_loss=0.04497, over 1322759.33 frames. ], batch size: 102, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:28:22,078 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.266e+02 2.864e+02 3.484e+02 1.313e+03, threshold=5.728e+02, percent-clipped=5.0 2023-05-01 22:28:35,898 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6955, 3.7129, 4.3394, 2.5248, 3.5588, 2.6276, 4.2515, 4.0141], device='cuda:3'), covar=tensor([0.0245, 0.0930, 0.0468, 0.1912, 0.0748, 0.1015, 0.0466, 0.1046], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0160, 0.0165, 0.0152, 0.0142, 0.0128, 0.0141, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-01 22:28:41,002 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2004, 2.1115, 1.6956, 1.7676, 2.3528, 2.0661, 2.1580, 2.3919], device='cuda:3'), covar=tensor([0.0271, 0.0404, 0.0577, 0.0514, 0.0265, 0.0370, 0.0230, 0.0328], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0238, 0.0228, 0.0228, 0.0238, 0.0236, 0.0234, 0.0232], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:28:45,144 INFO [train.py:904] (3/8) Epoch 25, batch 150, loss[loss=0.1806, simple_loss=0.2785, pruned_loss=0.04135, over 17102.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2627, pruned_loss=0.04321, over 1765534.65 frames. ], batch size: 53, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:29:26,432 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 22:29:53,831 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9683, 1.8229, 2.4889, 2.8538, 2.7338, 3.3877, 1.9704, 3.4940], device='cuda:3'), covar=tensor([0.0269, 0.0696, 0.0374, 0.0348, 0.0374, 0.0236, 0.0767, 0.0185], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0190, 0.0178, 0.0181, 0.0196, 0.0155, 0.0194, 0.0153], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:29:55,569 INFO [train.py:904] (3/8) Epoch 25, batch 200, loss[loss=0.1826, simple_loss=0.2643, pruned_loss=0.0505, over 16620.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2632, pruned_loss=0.04352, over 2111212.10 frames. ], batch size: 134, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:30:25,886 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5433, 5.9684, 5.7229, 5.7327, 5.3771, 5.3551, 5.3611, 6.1111], device='cuda:3'), covar=tensor([0.1750, 0.1118, 0.1102, 0.0975, 0.0862, 0.0700, 0.1442, 0.0991], device='cuda:3'), in_proj_covar=tensor([0.0685, 0.0824, 0.0679, 0.0638, 0.0522, 0.0529, 0.0691, 0.0646], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:30:40,581 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.064e+02 2.455e+02 3.172e+02 7.590e+02, threshold=4.909e+02, percent-clipped=1.0 2023-05-01 22:30:42,797 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8830, 3.7423, 3.9392, 4.0256, 4.0848, 3.7315, 3.9132, 4.1054], device='cuda:3'), covar=tensor([0.1553, 0.1133, 0.1132, 0.0673, 0.0611, 0.1641, 0.2130, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0633, 0.0781, 0.0895, 0.0789, 0.0605, 0.0621, 0.0655, 0.0759], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:31:04,165 INFO [train.py:904] (3/8) Epoch 25, batch 250, loss[loss=0.1727, simple_loss=0.2609, pruned_loss=0.04223, over 16649.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2613, pruned_loss=0.04353, over 2373285.19 frames. ], batch size: 62, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:32:14,684 INFO [train.py:904] (3/8) Epoch 25, batch 300, loss[loss=0.1657, simple_loss=0.2596, pruned_loss=0.03585, over 16525.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2582, pruned_loss=0.04212, over 2580305.61 frames. ], batch size: 68, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:32:15,454 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 22:32:27,365 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6763, 4.6963, 4.6163, 4.2048, 4.6451, 1.8693, 4.4408, 4.3687], device='cuda:3'), covar=tensor([0.0140, 0.0106, 0.0203, 0.0298, 0.0112, 0.2609, 0.0157, 0.0232], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0162, 0.0200, 0.0175, 0.0177, 0.0209, 0.0189, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:32:56,296 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 22:33:00,865 INFO [optim.py:368] (3/8) 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 22:33:18,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2649, 3.2163, 2.1110, 3.4140, 2.5703, 3.4486, 2.1807, 2.6674], device='cuda:3'), covar=tensor([0.0315, 0.0507, 0.1560, 0.0392, 0.0836, 0.0784, 0.1454, 0.0793], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0166, 0.0177, 0.0215, 0.0203, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 22:33:19,568 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 350, loss[loss=0.1552, simple_loss=0.2534, pruned_loss=0.02851, over 17116.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2556, pruned_loss=0.0412, over 2746777.71 frames. ], batch size: 47, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:34:10,505 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 22:34:25,843 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 400, loss[loss=0.1807, simple_loss=0.2568, pruned_loss=0.05226, over 16718.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2544, pruned_loss=0.04089, over 2878975.10 frames. ], batch size: 134, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:35:22,991 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.193e+02 2.643e+02 3.297e+02 6.329e+02, threshold=5.285e+02, percent-clipped=2.0 2023-05-01 22:35:47,033 INFO [train.py:904] (3/8) Epoch 25, batch 450, loss[loss=0.1393, simple_loss=0.2363, pruned_loss=0.02111, over 17181.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2533, pruned_loss=0.04003, over 2968409.66 frames. ], batch size: 46, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:35:53,947 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9808, 4.9730, 4.7852, 4.2502, 4.8448, 2.0054, 4.6184, 4.5216], device='cuda:3'), covar=tensor([0.0144, 0.0114, 0.0243, 0.0393, 0.0144, 0.2903, 0.0159, 0.0258], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0164, 0.0202, 0.0178, 0.0180, 0.0212, 0.0192, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:36:50,450 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1926, 4.1368, 4.5152, 4.5014, 4.5390, 4.2848, 4.2799, 4.1822], device='cuda:3'), covar=tensor([0.0422, 0.0839, 0.0419, 0.0414, 0.0504, 0.0460, 0.0770, 0.0623], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0459, 0.0447, 0.0412, 0.0491, 0.0470, 0.0546, 0.0375], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 22:36:55,204 INFO [train.py:904] (3/8) Epoch 25, batch 500, loss[loss=0.1524, simple_loss=0.2514, pruned_loss=0.02673, over 16642.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2514, pruned_loss=0.03953, over 3045003.39 frames. ], batch size: 57, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:37:18,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4150, 3.2650, 3.4307, 2.5048, 3.2810, 3.5836, 3.3007, 2.1512], device='cuda:3'), covar=tensor([0.0511, 0.0160, 0.0070, 0.0408, 0.0124, 0.0118, 0.0112, 0.0479], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0088, 0.0087, 0.0136, 0.0100, 0.0111, 0.0096, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 22:37:37,518 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2188, 5.7099, 5.8615, 5.5451, 5.6430, 6.2586, 5.7232, 5.5015], device='cuda:3'), covar=tensor([0.0913, 0.1841, 0.2571, 0.2209, 0.2548, 0.0881, 0.1710, 0.2307], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0610, 0.0672, 0.0499, 0.0659, 0.0693, 0.0520, 0.0659], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 22:37:42,042 INFO [optim.py:368] (3/8) 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 22:38:05,720 INFO [train.py:904] (3/8) Epoch 25, batch 550, loss[loss=0.1714, simple_loss=0.2438, pruned_loss=0.04956, over 16468.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2497, pruned_loss=0.03913, over 3103211.95 frames. ], batch size: 146, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:38:50,295 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7636, 4.3252, 3.0002, 2.2978, 2.6744, 2.6371, 4.5896, 3.5608], device='cuda:3'), covar=tensor([0.2882, 0.0506, 0.1862, 0.2839, 0.2845, 0.2032, 0.0344, 0.1391], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0268, 0.0306, 0.0316, 0.0295, 0.0266, 0.0296, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 22:39:05,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8953, 2.9675, 2.7727, 5.1024, 4.1019, 4.4537, 1.7288, 3.2417], device='cuda:3'), covar=tensor([0.1374, 0.0807, 0.1236, 0.0241, 0.0237, 0.0406, 0.1650, 0.0786], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0192, 0.0200, 0.0214, 0.0204, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 22:39:15,773 INFO [train.py:904] (3/8) Epoch 25, batch 600, loss[loss=0.1604, simple_loss=0.2389, pruned_loss=0.04096, over 16865.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2488, pruned_loss=0.03895, over 3150619.81 frames. ], batch size: 109, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:39:32,775 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5366, 3.6133, 2.7416, 2.1731, 2.2469, 2.2336, 3.6644, 3.0812], device='cuda:3'), covar=tensor([0.2891, 0.0603, 0.1860, 0.3035, 0.2948, 0.2421, 0.0524, 0.1686], device='cuda:3'), in_proj_covar=tensor([0.0328, 0.0268, 0.0306, 0.0317, 0.0296, 0.0266, 0.0297, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 22:39:57,014 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.8365, 6.2222, 5.9705, 6.0672, 5.5406, 5.6087, 5.7482, 6.3038], device='cuda:3'), covar=tensor([0.1226, 0.0938, 0.1005, 0.0850, 0.0887, 0.0624, 0.1006, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0698, 0.0841, 0.0691, 0.0650, 0.0533, 0.0539, 0.0707, 0.0660], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:40:02,986 INFO [optim.py:368] (3/8) 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 22:40:08,209 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2841, 5.2858, 5.1232, 4.5498, 4.7636, 5.1861, 5.0890, 4.7659], device='cuda:3'), covar=tensor([0.0619, 0.0553, 0.0320, 0.0391, 0.1092, 0.0480, 0.0319, 0.0849], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0447, 0.0347, 0.0350, 0.0351, 0.0402, 0.0240, 0.0419], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:40:25,349 INFO [train.py:904] (3/8) Epoch 25, batch 650, loss[loss=0.1665, simple_loss=0.2609, pruned_loss=0.03611, over 17042.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2472, pruned_loss=0.03855, over 3191178.94 frames. ], batch size: 55, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:40:51,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7207, 4.0271, 3.0170, 2.2971, 2.5450, 2.5212, 4.1022, 3.3755], device='cuda:3'), covar=tensor([0.2870, 0.0531, 0.1791, 0.3169, 0.2860, 0.2137, 0.0490, 0.1526], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0268, 0.0307, 0.0317, 0.0296, 0.0267, 0.0297, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 22:41:33,626 INFO [train.py:904] (3/8) Epoch 25, batch 700, loss[loss=0.1969, simple_loss=0.2655, pruned_loss=0.0641, over 16900.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2471, pruned_loss=0.03859, over 3221351.05 frames. ], batch size: 109, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:42:20,955 INFO [optim.py:368] (3/8) 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 22:42:21,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3821, 5.2883, 5.1817, 4.6236, 4.7897, 5.2397, 5.1985, 4.8000], device='cuda:3'), covar=tensor([0.0594, 0.0543, 0.0367, 0.0382, 0.1250, 0.0517, 0.0321, 0.0918], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0453, 0.0351, 0.0354, 0.0356, 0.0408, 0.0242, 0.0425], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:42:42,412 INFO [train.py:904] (3/8) Epoch 25, batch 750, loss[loss=0.1571, simple_loss=0.2398, pruned_loss=0.03715, over 16435.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2477, pruned_loss=0.03836, over 3251030.60 frames. ], batch size: 75, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:42:42,783 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:42:58,790 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7600, 1.9158, 2.3013, 2.6455, 2.6271, 2.6072, 1.9500, 2.8191], device='cuda:3'), covar=tensor([0.0190, 0.0506, 0.0382, 0.0311, 0.0340, 0.0339, 0.0566, 0.0202], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0185, 0.0200, 0.0159, 0.0196, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:43:52,159 INFO [train.py:904] (3/8) Epoch 25, batch 800, loss[loss=0.1634, simple_loss=0.2471, pruned_loss=0.0399, over 16540.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2482, pruned_loss=0.03896, over 3270915.68 frames. ], batch size: 68, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:44:08,218 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:44:39,329 INFO [optim.py:368] (3/8) 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 22:44:44,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1411, 3.0970, 3.1531, 2.2595, 2.9896, 3.2433, 3.0172, 2.0612], device='cuda:3'), covar=tensor([0.0543, 0.0134, 0.0087, 0.0445, 0.0136, 0.0166, 0.0131, 0.0486], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0088, 0.0088, 0.0136, 0.0101, 0.0111, 0.0097, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 22:45:03,354 INFO [train.py:904] (3/8) Epoch 25, batch 850, loss[loss=0.1541, simple_loss=0.256, pruned_loss=0.02605, over 17122.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2476, pruned_loss=0.0385, over 3282585.02 frames. ], batch size: 49, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:12,665 INFO [train.py:904] (3/8) Epoch 25, batch 900, loss[loss=0.1603, simple_loss=0.2535, pruned_loss=0.03351, over 17080.00 frames. ], tot_loss[loss=0.161, simple_loss=0.247, pruned_loss=0.03747, over 3300440.87 frames. ], batch size: 53, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:47:00,724 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 1.999e+02 2.403e+02 2.874e+02 4.869e+02, threshold=4.806e+02, percent-clipped=0.0 2023-05-01 22:47:23,398 INFO [train.py:904] (3/8) Epoch 25, batch 950, loss[loss=0.1712, simple_loss=0.2436, pruned_loss=0.04937, over 16352.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2481, pruned_loss=0.0379, over 3309547.42 frames. ], batch size: 146, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:48:33,896 INFO [train.py:904] (3/8) Epoch 25, batch 1000, loss[loss=0.1816, simple_loss=0.2504, pruned_loss=0.05643, over 16752.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2469, pruned_loss=0.03803, over 3313571.52 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:49:21,001 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 1050, loss[loss=0.1531, simple_loss=0.2513, pruned_loss=0.0275, over 17130.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2466, pruned_loss=0.03778, over 3318048.30 frames. ], batch size: 48, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:50:53,098 INFO [train.py:904] (3/8) Epoch 25, batch 1100, loss[loss=0.141, simple_loss=0.2211, pruned_loss=0.03049, over 16868.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2461, pruned_loss=0.03703, over 3312902.40 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:51:01,221 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=244709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:51:01,409 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8416, 2.7778, 2.6501, 4.8080, 3.6615, 4.2652, 1.6753, 3.0414], device='cuda:3'), covar=tensor([0.1508, 0.0954, 0.1390, 0.0229, 0.0282, 0.0514, 0.1843, 0.0920], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0177, 0.0196, 0.0194, 0.0201, 0.0215, 0.0205, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 22:51:13,228 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244717.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:51:30,242 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1902, 2.2672, 2.3742, 3.8912, 2.3350, 2.6095, 2.3338, 2.4570], device='cuda:3'), covar=tensor([0.1613, 0.3947, 0.3251, 0.0738, 0.3976, 0.2711, 0.4045, 0.3306], device='cuda:3'), in_proj_covar=tensor([0.0415, 0.0463, 0.0382, 0.0336, 0.0445, 0.0530, 0.0435, 0.0542], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:51:40,297 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.414e+02 2.068e+02 2.379e+02 2.713e+02 6.650e+02, threshold=4.759e+02, percent-clipped=1.0 2023-05-01 22:52:02,015 INFO [train.py:904] (3/8) Epoch 25, batch 1150, loss[loss=0.1636, simple_loss=0.2417, pruned_loss=0.04277, over 16812.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2462, pruned_loss=0.03707, over 3318872.63 frames. ], batch size: 102, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:52:37,202 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244778.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:53:11,758 INFO [train.py:904] (3/8) Epoch 25, batch 1200, loss[loss=0.1871, simple_loss=0.2651, pruned_loss=0.05459, over 16817.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2452, pruned_loss=0.03698, over 3304531.54 frames. ], batch size: 102, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:53:30,760 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 22:53:39,451 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8092, 4.2582, 3.0004, 2.3343, 2.6732, 2.6249, 4.5926, 3.4969], device='cuda:3'), covar=tensor([0.2897, 0.0578, 0.1880, 0.3013, 0.3021, 0.2146, 0.0355, 0.1487], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0271, 0.0309, 0.0318, 0.0299, 0.0269, 0.0299, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 22:53:57,424 INFO [optim.py:368] (3/8) 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 22:54:19,236 INFO [train.py:904] (3/8) Epoch 25, batch 1250, loss[loss=0.143, simple_loss=0.2334, pruned_loss=0.02626, over 17206.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2452, pruned_loss=0.03769, over 3314143.41 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:55:06,283 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:55:21,550 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0890, 4.8169, 5.0854, 5.2690, 5.4770, 4.8118, 5.4696, 5.4769], device='cuda:3'), covar=tensor([0.1932, 0.1389, 0.1785, 0.0829, 0.0562, 0.0875, 0.0509, 0.0584], device='cuda:3'), in_proj_covar=tensor([0.0669, 0.0826, 0.0951, 0.0836, 0.0638, 0.0657, 0.0687, 0.0801], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:55:26,316 INFO [train.py:904] (3/8) Epoch 25, batch 1300, loss[loss=0.1833, simple_loss=0.2639, pruned_loss=0.05141, over 16895.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2455, pruned_loss=0.03735, over 3319625.87 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:12,138 INFO [optim.py:368] (3/8) 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 22:56:12,811 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 22:56:24,547 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 22:56:28,180 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244948.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:56:31,305 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4511, 1.7140, 2.1514, 2.3249, 2.4617, 2.4302, 1.8255, 2.5387], device='cuda:3'), covar=tensor([0.0225, 0.0562, 0.0338, 0.0350, 0.0345, 0.0341, 0.0578, 0.0211], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0187, 0.0203, 0.0161, 0.0199, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:56:35,067 INFO [train.py:904] (3/8) Epoch 25, batch 1350, loss[loss=0.1629, simple_loss=0.2517, pruned_loss=0.0371, over 17108.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2457, pruned_loss=0.03732, over 3320049.44 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:36,679 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:01,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1595, 4.0037, 4.2243, 4.3336, 4.3881, 3.9864, 4.1994, 4.4128], device='cuda:3'), covar=tensor([0.1504, 0.1141, 0.1207, 0.0699, 0.0618, 0.1418, 0.2420, 0.0811], device='cuda:3'), in_proj_covar=tensor([0.0672, 0.0830, 0.0956, 0.0839, 0.0641, 0.0660, 0.0691, 0.0806], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:57:43,144 INFO [train.py:904] (3/8) Epoch 25, batch 1400, loss[loss=0.1474, simple_loss=0.2465, pruned_loss=0.02421, over 17141.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2457, pruned_loss=0.03723, over 3326287.26 frames. ], batch size: 48, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:57:48,568 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-05-01 22:57:51,808 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:00,016 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245015.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:28,649 INFO [optim.py:368] (3/8) 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 22:58:51,282 INFO [train.py:904] (3/8) Epoch 25, batch 1450, loss[loss=0.1507, simple_loss=0.2379, pruned_loss=0.03172, over 17186.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2445, pruned_loss=0.03701, over 3325284.38 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:58:55,879 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6540, 4.6621, 4.5587, 4.0861, 4.6236, 1.9049, 4.3610, 4.1593], device='cuda:3'), covar=tensor([0.0151, 0.0130, 0.0196, 0.0275, 0.0100, 0.2742, 0.0143, 0.0219], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0169, 0.0208, 0.0183, 0.0186, 0.0215, 0.0197, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:58:56,914 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:59:00,753 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1774, 2.1934, 2.3721, 3.8568, 2.2373, 2.5535, 2.2870, 2.3852], device='cuda:3'), covar=tensor([0.1586, 0.3701, 0.3196, 0.0747, 0.3923, 0.2663, 0.3913, 0.3168], device='cuda:3'), in_proj_covar=tensor([0.0416, 0.0465, 0.0383, 0.0336, 0.0446, 0.0532, 0.0437, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:59:12,866 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2872, 4.0863, 4.3387, 4.4711, 4.5467, 4.1397, 4.3261, 4.5432], device='cuda:3'), covar=tensor([0.1479, 0.1154, 0.1253, 0.0644, 0.0547, 0.1226, 0.2363, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0677, 0.0837, 0.0965, 0.0846, 0.0645, 0.0665, 0.0695, 0.0813], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 22:59:18,775 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245073.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 22:59:40,910 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0011, 2.2183, 2.5531, 2.9220, 2.8569, 3.0775, 2.2273, 3.1398], device='cuda:3'), covar=tensor([0.0207, 0.0492, 0.0373, 0.0314, 0.0318, 0.0268, 0.0530, 0.0190], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0197, 0.0184, 0.0187, 0.0203, 0.0161, 0.0199, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:00:00,085 INFO [train.py:904] (3/8) Epoch 25, batch 1500, loss[loss=0.1976, simple_loss=0.2648, pruned_loss=0.06522, over 16733.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2446, pruned_loss=0.03733, over 3328275.74 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:00:12,898 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-01 23:00:46,574 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 1550, loss[loss=0.1676, simple_loss=0.259, pruned_loss=0.0381, over 17118.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2464, pruned_loss=0.03772, over 3321210.88 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:01:36,845 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 23:02:19,884 INFO [train.py:904] (3/8) Epoch 25, batch 1600, loss[loss=0.1588, simple_loss=0.2537, pruned_loss=0.03197, over 17259.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2489, pruned_loss=0.03835, over 3322942.11 frames. ], batch size: 52, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:06,950 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.282e+02 2.634e+02 3.263e+02 7.681e+02, threshold=5.268e+02, percent-clipped=4.0 2023-05-01 23:03:16,822 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:03:29,745 INFO [train.py:904] (3/8) Epoch 25, batch 1650, loss[loss=0.1453, simple_loss=0.2309, pruned_loss=0.02984, over 16970.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2498, pruned_loss=0.03818, over 3329428.21 frames. ], batch size: 41, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:32,309 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 23:04:41,927 INFO [train.py:904] (3/8) Epoch 25, batch 1700, loss[loss=0.1546, simple_loss=0.2569, pruned_loss=0.02617, over 17126.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2514, pruned_loss=0.03897, over 3326358.72 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:04:42,720 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 23:04:51,856 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:05:30,603 INFO [optim.py:368] (3/8) 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 23:05:52,660 INFO [train.py:904] (3/8) Epoch 25, batch 1750, loss[loss=0.1947, simple_loss=0.2803, pruned_loss=0.05454, over 12271.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2519, pruned_loss=0.03927, over 3325112.67 frames. ], batch size: 247, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:06:20,884 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 1800, loss[loss=0.1914, simple_loss=0.2787, pruned_loss=0.05201, over 16169.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2534, pruned_loss=0.03961, over 3311116.72 frames. ], batch size: 164, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:07:10,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5022, 3.6083, 3.3129, 2.9570, 3.0436, 3.4949, 3.2568, 3.2393], device='cuda:3'), covar=tensor([0.0670, 0.0726, 0.0361, 0.0357, 0.0627, 0.0559, 0.1929, 0.0635], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0461, 0.0357, 0.0361, 0.0363, 0.0416, 0.0246, 0.0433], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:07:29,357 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245421.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:07:51,079 INFO [optim.py:368] (3/8) 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 23:08:12,923 INFO [train.py:904] (3/8) Epoch 25, batch 1850, loss[loss=0.1501, simple_loss=0.2334, pruned_loss=0.03344, over 15881.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2542, pruned_loss=0.0396, over 3307957.06 frames. ], batch size: 35, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:08:23,441 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:08:40,660 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 23:09:23,472 INFO [train.py:904] (3/8) Epoch 25, batch 1900, loss[loss=0.1561, simple_loss=0.2511, pruned_loss=0.03052, over 17154.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2534, pruned_loss=0.03916, over 3312068.60 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:09:49,812 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245521.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:11,813 INFO [optim.py:368] (3/8) 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 23:10:15,719 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 23:10:20,953 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245543.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:25,075 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 1950, loss[loss=0.1594, simple_loss=0.2573, pruned_loss=0.03077, over 17144.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2536, pruned_loss=0.03892, over 3322874.75 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:26,898 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:11:41,387 INFO [train.py:904] (3/8) Epoch 25, batch 2000, loss[loss=0.1735, simple_loss=0.255, pruned_loss=0.046, over 16667.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2526, pruned_loss=0.03898, over 3317494.06 frames. ], batch size: 134, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:49,129 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245607.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:11:53,234 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245610.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:12:05,609 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9326, 2.0846, 2.2810, 3.3977, 2.0907, 2.3206, 2.2393, 2.2195], device='cuda:3'), covar=tensor([0.1582, 0.3845, 0.3202, 0.0829, 0.4429, 0.2744, 0.3913, 0.3638], device='cuda:3'), in_proj_covar=tensor([0.0416, 0.0465, 0.0383, 0.0337, 0.0445, 0.0532, 0.0437, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:12:31,548 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 2050, loss[loss=0.1761, simple_loss=0.2692, pruned_loss=0.04153, over 17197.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2528, pruned_loss=0.03918, over 3317360.92 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:12:57,322 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245658.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:13:34,853 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245686.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:13:58,098 INFO [train.py:904] (3/8) Epoch 25, batch 2100, loss[loss=0.2046, simple_loss=0.2768, pruned_loss=0.06618, over 16689.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2541, pruned_loss=0.04022, over 3308985.83 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:14:12,452 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6632, 3.6536, 2.3859, 4.0841, 2.8981, 3.9873, 2.4107, 3.0084], device='cuda:3'), covar=tensor([0.0330, 0.0431, 0.1569, 0.0350, 0.0881, 0.0778, 0.1584, 0.0751], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0182, 0.0197, 0.0173, 0.0181, 0.0224, 0.0207, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 23:14:49,017 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.201e+02 2.533e+02 2.997e+02 6.005e+02, threshold=5.066e+02, percent-clipped=2.0 2023-05-01 23:14:59,050 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245747.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:15:08,426 INFO [train.py:904] (3/8) Epoch 25, batch 2150, loss[loss=0.1286, simple_loss=0.2157, pruned_loss=0.02081, over 17210.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2549, pruned_loss=0.04065, over 3304847.30 frames. ], batch size: 45, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:00,572 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 2200, loss[loss=0.1928, simple_loss=0.2697, pruned_loss=0.058, over 16747.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2558, pruned_loss=0.0413, over 3303045.28 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:34,701 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245816.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:16:36,665 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9334, 4.9530, 4.7630, 4.1689, 4.8598, 1.9159, 4.6418, 4.4495], device='cuda:3'), covar=tensor([0.0137, 0.0116, 0.0251, 0.0415, 0.0126, 0.2924, 0.0151, 0.0264], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0171, 0.0210, 0.0185, 0.0188, 0.0217, 0.0200, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:17:06,934 INFO [optim.py:368] (3/8) 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 23:17:23,656 INFO [train.py:904] (3/8) Epoch 25, batch 2250, loss[loss=0.1671, simple_loss=0.2624, pruned_loss=0.03594, over 17098.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2562, pruned_loss=0.04123, over 3310460.79 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:17:24,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7357, 2.7032, 2.3106, 2.5096, 3.0142, 2.7560, 3.2737, 3.2000], device='cuda:3'), covar=tensor([0.0170, 0.0495, 0.0601, 0.0530, 0.0310, 0.0459, 0.0296, 0.0312], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0248, 0.0235, 0.0237, 0.0248, 0.0247, 0.0248, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:17:24,812 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:16,591 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245891.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:24,698 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 23:18:32,173 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245902.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:18:32,938 INFO [train.py:904] (3/8) Epoch 25, batch 2300, loss[loss=0.1939, simple_loss=0.2686, pruned_loss=0.05963, over 16372.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2565, pruned_loss=0.04111, over 3313251.94 frames. ], batch size: 146, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:18:36,232 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2305, 3.2296, 2.1479, 3.4171, 2.6013, 3.4249, 2.2180, 2.6818], device='cuda:3'), covar=tensor([0.0310, 0.0419, 0.1414, 0.0359, 0.0782, 0.0710, 0.1423, 0.0742], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0182, 0.0197, 0.0173, 0.0181, 0.0223, 0.0206, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 23:18:36,318 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9952, 2.1089, 2.3911, 3.5175, 2.1504, 2.3757, 2.2566, 2.2653], device='cuda:3'), covar=tensor([0.1651, 0.3787, 0.2995, 0.0798, 0.4070, 0.2684, 0.3716, 0.3505], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0466, 0.0384, 0.0338, 0.0445, 0.0533, 0.0438, 0.0546], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:18:48,712 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9685, 4.1968, 4.0557, 4.1107, 3.7681, 3.8125, 3.8778, 4.2016], device='cuda:3'), covar=tensor([0.1063, 0.0930, 0.0999, 0.0805, 0.0760, 0.1864, 0.0887, 0.1030], device='cuda:3'), in_proj_covar=tensor([0.0718, 0.0871, 0.0711, 0.0672, 0.0552, 0.0553, 0.0730, 0.0680], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:19:24,138 INFO [optim.py:368] (3/8) 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 23:19:42,144 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 2350, loss[loss=0.1748, simple_loss=0.2539, pruned_loss=0.04779, over 16724.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2569, pruned_loss=0.04138, over 3310646.81 frames. ], batch size: 134, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:20:32,094 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3549, 3.5353, 3.6692, 3.6439, 3.6604, 3.4936, 3.5239, 3.5550], device='cuda:3'), covar=tensor([0.0437, 0.0652, 0.0528, 0.0462, 0.0549, 0.0536, 0.0844, 0.0579], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0483, 0.0472, 0.0433, 0.0517, 0.0497, 0.0575, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 23:20:54,784 INFO [train.py:904] (3/8) Epoch 25, batch 2400, loss[loss=0.1822, simple_loss=0.2626, pruned_loss=0.05089, over 16929.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2573, pruned_loss=0.04134, over 3314725.16 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:21:46,453 INFO [optim.py:368] (3/8) 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 23:21:49,227 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246042.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:22:04,193 INFO [train.py:904] (3/8) Epoch 25, batch 2450, loss[loss=0.1756, simple_loss=0.2527, pruned_loss=0.04925, over 16894.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2572, pruned_loss=0.0407, over 3321690.12 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:22:46,270 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 23:23:10,958 INFO [train.py:904] (3/8) Epoch 25, batch 2500, loss[loss=0.1514, simple_loss=0.2444, pruned_loss=0.02921, over 17146.00 frames. ], tot_loss[loss=0.169, simple_loss=0.257, pruned_loss=0.04047, over 3319724.39 frames. ], batch size: 48, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:21,526 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246111.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:28,083 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:01,581 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.119e+02 2.579e+02 3.197e+02 6.008e+02, threshold=5.158e+02, percent-clipped=3.0 2023-05-01 23:24:12,431 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:18,466 INFO [train.py:904] (3/8) Epoch 25, batch 2550, loss[loss=0.1586, simple_loss=0.2539, pruned_loss=0.03166, over 17207.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2571, pruned_loss=0.04089, over 3316719.69 frames. ], batch size: 44, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:24:35,188 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:38,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9317, 4.9829, 5.4085, 5.3622, 5.3819, 5.0606, 5.0224, 4.8849], device='cuda:3'), covar=tensor([0.0356, 0.0573, 0.0377, 0.0414, 0.0535, 0.0412, 0.0989, 0.0432], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0482, 0.0472, 0.0432, 0.0516, 0.0496, 0.0574, 0.0391], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 23:24:45,529 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:25:17,167 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246196.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:25:26,268 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246202.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:25:26,997 INFO [train.py:904] (3/8) Epoch 25, batch 2600, loss[loss=0.1809, simple_loss=0.2636, pruned_loss=0.04912, over 16888.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.256, pruned_loss=0.04031, over 3327053.91 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:25:52,182 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:18,784 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.097e+02 2.510e+02 3.006e+02 6.140e+02, threshold=5.021e+02, percent-clipped=2.0 2023-05-01 23:26:28,308 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:33,160 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:37,199 INFO [train.py:904] (3/8) Epoch 25, batch 2650, loss[loss=0.1603, simple_loss=0.2576, pruned_loss=0.03154, over 17119.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2564, pruned_loss=0.0401, over 3334525.48 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:26:42,358 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246257.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:27:17,745 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:27:31,067 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 23:27:44,603 INFO [train.py:904] (3/8) Epoch 25, batch 2700, loss[loss=0.1842, simple_loss=0.2749, pruned_loss=0.04671, over 17037.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2571, pruned_loss=0.03989, over 3335768.41 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:28:19,344 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 23:28:34,095 INFO [optim.py:368] (3/8) 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 23:28:36,905 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 2750, loss[loss=0.1663, simple_loss=0.2615, pruned_loss=0.03559, over 16742.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2571, pruned_loss=0.03914, over 3340449.57 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:29:44,294 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246390.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:30:01,769 INFO [train.py:904] (3/8) Epoch 25, batch 2800, loss[loss=0.1519, simple_loss=0.2464, pruned_loss=0.02874, over 15588.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2576, pruned_loss=0.03944, over 3339416.27 frames. ], batch size: 190, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:30:16,920 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9071, 4.3738, 3.0365, 2.3872, 2.7017, 2.6621, 4.7112, 3.5750], device='cuda:3'), covar=tensor([0.2900, 0.0587, 0.1985, 0.3189, 0.3090, 0.2187, 0.0360, 0.1514], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0275, 0.0312, 0.0323, 0.0304, 0.0273, 0.0303, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-01 23:30:54,341 INFO [optim.py:368] (3/8) 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 23:31:04,195 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:31:10,739 INFO [train.py:904] (3/8) Epoch 25, batch 2850, loss[loss=0.1641, simple_loss=0.2677, pruned_loss=0.03026, over 17258.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2565, pruned_loss=0.03919, over 3337758.29 frames. ], batch size: 52, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:31:25,871 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6198, 3.3684, 3.7588, 2.0417, 3.8088, 3.8381, 3.2042, 2.9169], device='cuda:3'), covar=tensor([0.0746, 0.0256, 0.0177, 0.1118, 0.0115, 0.0208, 0.0402, 0.0438], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0101, 0.0141, 0.0084, 0.0132, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-01 23:31:29,079 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9963, 2.1001, 2.5964, 2.9530, 2.8145, 3.4445, 2.4900, 3.4390], device='cuda:3'), covar=tensor([0.0300, 0.0596, 0.0393, 0.0369, 0.0386, 0.0202, 0.0524, 0.0200], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0196, 0.0185, 0.0189, 0.0204, 0.0163, 0.0201, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:31:31,873 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:31:50,629 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5228, 4.6784, 4.6902, 3.5880, 3.9069, 4.5890, 4.1120, 3.0139], device='cuda:3'), covar=tensor([0.0362, 0.0068, 0.0044, 0.0317, 0.0147, 0.0109, 0.0103, 0.0415], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0135, 0.0101, 0.0112, 0.0098, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-01 23:32:10,939 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:20,917 INFO [train.py:904] (3/8) Epoch 25, batch 2900, loss[loss=0.1774, simple_loss=0.249, pruned_loss=0.05288, over 16415.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2557, pruned_loss=0.03919, over 3335077.44 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:33:14,957 INFO [optim.py:368] (3/8) 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 23:33:25,241 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246547.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:33:32,610 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246552.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 23:33:33,513 INFO [train.py:904] (3/8) Epoch 25, batch 2950, loss[loss=0.1632, simple_loss=0.2573, pruned_loss=0.03459, over 17262.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2548, pruned_loss=0.03931, over 3334849.29 frames. ], batch size: 52, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:06,222 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:06,394 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7836, 2.4832, 2.0843, 2.3047, 2.8593, 2.6357, 2.8230, 2.9464], device='cuda:3'), covar=tensor([0.0242, 0.0427, 0.0543, 0.0507, 0.0246, 0.0348, 0.0261, 0.0299], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0246, 0.0233, 0.0236, 0.0248, 0.0245, 0.0249, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:34:32,303 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 3000, loss[loss=0.1574, simple_loss=0.2486, pruned_loss=0.0331, over 17220.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2556, pruned_loss=0.0402, over 3331704.10 frames. ], batch size: 44, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:42,716 INFO [train.py:929] (3/8) Computing validation loss 2023-05-01 23:34:52,581 INFO [train.py:938] (3/8) Epoch 25, validation: loss=0.1341, simple_loss=0.239, pruned_loss=0.01457, over 944034.00 frames. 2023-05-01 23:34:52,582 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-01 23:35:46,635 INFO [optim.py:368] (3/8) 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 23:36:05,630 INFO [train.py:904] (3/8) Epoch 25, batch 3050, loss[loss=0.1724, simple_loss=0.2656, pruned_loss=0.03958, over 17114.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2552, pruned_loss=0.04041, over 3329116.59 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:36:38,404 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:37:13,554 INFO [train.py:904] (3/8) Epoch 25, batch 3100, loss[loss=0.1707, simple_loss=0.2486, pruned_loss=0.04645, over 15669.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2553, pruned_loss=0.04012, over 3320625.17 frames. ], batch size: 191, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:37:47,025 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2557, 5.8115, 5.9680, 5.6359, 5.7758, 6.2893, 5.8053, 5.4869], device='cuda:3'), covar=tensor([0.0931, 0.1778, 0.2060, 0.1963, 0.2432, 0.0908, 0.1509, 0.2380], device='cuda:3'), in_proj_covar=tensor([0.0430, 0.0638, 0.0698, 0.0518, 0.0694, 0.0723, 0.0541, 0.0691], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 23:38:04,979 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.159e+02 2.563e+02 3.074e+02 4.728e+02, threshold=5.126e+02, percent-clipped=0.0 2023-05-01 23:38:25,015 INFO [train.py:904] (3/8) Epoch 25, batch 3150, loss[loss=0.1573, simple_loss=0.2477, pruned_loss=0.03348, over 16778.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2546, pruned_loss=0.04003, over 3326219.50 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:38:44,276 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:39:34,230 INFO [train.py:904] (3/8) Epoch 25, batch 3200, loss[loss=0.1407, simple_loss=0.2266, pruned_loss=0.02743, over 16993.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2535, pruned_loss=0.03994, over 3322109.71 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:39:51,750 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246815.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:39:56,135 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3706, 3.5423, 3.8546, 2.0630, 3.0562, 2.3947, 3.7384, 3.7414], device='cuda:3'), covar=tensor([0.0279, 0.0910, 0.0488, 0.2131, 0.0876, 0.1007, 0.0618, 0.1143], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0156, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-01 23:39:56,353 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 23:40:00,744 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 23:40:22,902 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6842, 2.7132, 2.5145, 2.5545, 3.0455, 2.7264, 3.2758, 3.1944], device='cuda:3'), covar=tensor([0.0191, 0.0453, 0.0499, 0.0491, 0.0285, 0.0449, 0.0285, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0248, 0.0234, 0.0237, 0.0249, 0.0247, 0.0250, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:40:25,966 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 23:40:27,530 INFO [optim.py:368] (3/8) 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 23:40:42,124 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 3250, loss[loss=0.1775, simple_loss=0.2761, pruned_loss=0.03944, over 17077.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.254, pruned_loss=0.04072, over 3329014.65 frames. ], batch size: 55, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:41:16,132 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:41:28,841 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3047, 2.1301, 2.3383, 4.0674, 2.1305, 2.3806, 2.2609, 2.2922], device='cuda:3'), covar=tensor([0.1752, 0.4774, 0.3604, 0.0702, 0.5559, 0.3625, 0.4312, 0.4702], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0466, 0.0384, 0.0338, 0.0445, 0.0534, 0.0439, 0.0546], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:41:49,850 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246900.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:41:53,758 INFO [train.py:904] (3/8) Epoch 25, batch 3300, loss[loss=0.1744, simple_loss=0.2778, pruned_loss=0.0355, over 17263.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2547, pruned_loss=0.04065, over 3325865.03 frames. ], batch size: 52, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:42:24,636 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:36,863 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 3350, loss[loss=0.1665, simple_loss=0.2493, pruned_loss=0.04186, over 16780.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2558, pruned_loss=0.0406, over 3323572.92 frames. ], batch size: 102, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:01,679 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246995.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:44:13,545 INFO [train.py:904] (3/8) Epoch 25, batch 3400, loss[loss=0.1869, simple_loss=0.2613, pruned_loss=0.05621, over 16886.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2558, pruned_loss=0.04015, over 3328838.40 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:56,883 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.118e+02 2.456e+02 2.904e+02 5.980e+02, threshold=4.912e+02, percent-clipped=3.0 2023-05-01 23:45:26,146 INFO [train.py:904] (3/8) Epoch 25, batch 3450, loss[loss=0.1894, simple_loss=0.2677, pruned_loss=0.05555, over 16720.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2537, pruned_loss=0.03944, over 3340825.79 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:45:39,985 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 23:46:33,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9868, 5.0551, 5.4647, 5.4688, 5.4657, 5.1456, 5.0992, 4.9165], device='cuda:3'), covar=tensor([0.0397, 0.0705, 0.0375, 0.0349, 0.0455, 0.0373, 0.0926, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0437, 0.0491, 0.0477, 0.0438, 0.0522, 0.0502, 0.0582, 0.0398], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-01 23:46:35,594 INFO [train.py:904] (3/8) Epoch 25, batch 3500, loss[loss=0.1538, simple_loss=0.2449, pruned_loss=0.03138, over 17119.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2525, pruned_loss=0.03918, over 3337302.31 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:46:43,892 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3833, 2.1690, 1.7883, 1.9779, 2.4714, 2.2598, 2.3650, 2.5837], device='cuda:3'), covar=tensor([0.0283, 0.0452, 0.0611, 0.0532, 0.0295, 0.0397, 0.0230, 0.0348], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0248, 0.0235, 0.0238, 0.0249, 0.0247, 0.0250, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:47:30,016 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.124e+02 2.460e+02 2.748e+02 9.578e+02, threshold=4.920e+02, percent-clipped=1.0 2023-05-01 23:47:45,773 INFO [train.py:904] (3/8) Epoch 25, batch 3550, loss[loss=0.1631, simple_loss=0.2552, pruned_loss=0.03547, over 17119.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2523, pruned_loss=0.03907, over 3332661.41 frames. ], batch size: 47, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:48:07,888 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 23:48:55,167 INFO [train.py:904] (3/8) Epoch 25, batch 3600, loss[loss=0.1613, simple_loss=0.255, pruned_loss=0.0338, over 17132.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2511, pruned_loss=0.03888, over 3337749.88 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:48:58,324 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 23:49:49,108 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 3650, loss[loss=0.1457, simple_loss=0.2385, pruned_loss=0.02644, over 17230.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2499, pruned_loss=0.0391, over 3327966.41 frames. ], batch size: 44, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:50:59,470 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:51:02,304 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 23:51:17,358 INFO [train.py:904] (3/8) Epoch 25, batch 3700, loss[loss=0.1829, simple_loss=0.2565, pruned_loss=0.05469, over 16892.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2489, pruned_loss=0.04066, over 3304173.12 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:52:01,348 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:52:12,945 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.292e+02 2.615e+02 3.085e+02 5.082e+02, threshold=5.229e+02, percent-clipped=0.0 2023-05-01 23:52:28,375 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6794, 4.5510, 4.7414, 4.8876, 5.0038, 4.5456, 4.9661, 5.0056], device='cuda:3'), covar=tensor([0.1959, 0.1424, 0.1578, 0.0971, 0.0821, 0.1166, 0.1657, 0.1338], device='cuda:3'), in_proj_covar=tensor([0.0691, 0.0853, 0.0982, 0.0860, 0.0659, 0.0679, 0.0706, 0.0828], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:52:30,272 INFO [train.py:904] (3/8) Epoch 25, batch 3750, loss[loss=0.1676, simple_loss=0.252, pruned_loss=0.04161, over 16451.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2492, pruned_loss=0.04177, over 3285783.38 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:53:06,785 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:10,243 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:33,352 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-05-01 23:53:42,910 INFO [train.py:904] (3/8) Epoch 25, batch 3800, loss[loss=0.1663, simple_loss=0.2455, pruned_loss=0.04359, over 16873.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2504, pruned_loss=0.04299, over 3281852.50 frames. ], batch size: 90, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:54:06,893 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0199, 2.2005, 2.3177, 3.6902, 2.1151, 2.4322, 2.2656, 2.3693], device='cuda:3'), covar=tensor([0.1625, 0.3844, 0.3129, 0.0700, 0.4232, 0.2676, 0.3911, 0.3175], device='cuda:3'), in_proj_covar=tensor([0.0416, 0.0465, 0.0383, 0.0337, 0.0444, 0.0533, 0.0438, 0.0545], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:54:16,051 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247426.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:35,825 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247440.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:38,773 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.614e+02 2.181e+02 2.707e+02 3.148e+02 5.686e+02, threshold=5.414e+02, percent-clipped=1.0 2023-05-01 23:54:54,709 INFO [train.py:904] (3/8) Epoch 25, batch 3850, loss[loss=0.1659, simple_loss=0.2461, pruned_loss=0.04287, over 16868.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2508, pruned_loss=0.04366, over 3268515.26 frames. ], batch size: 96, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:55:23,205 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 23:55:40,068 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9378, 2.2023, 2.5546, 2.9018, 2.9422, 3.0337, 2.1884, 3.2076], device='cuda:3'), covar=tensor([0.0198, 0.0467, 0.0358, 0.0281, 0.0304, 0.0279, 0.0524, 0.0156], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0196, 0.0185, 0.0189, 0.0205, 0.0163, 0.0201, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-01 23:55:43,403 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247487.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:02,986 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:06,079 INFO [train.py:904] (3/8) Epoch 25, batch 3900, loss[loss=0.1606, simple_loss=0.2399, pruned_loss=0.04069, over 16873.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2499, pruned_loss=0.0437, over 3282131.24 frames. ], batch size: 90, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:56:39,775 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:03,110 INFO [optim.py:368] (3/8) 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 23:57:14,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5888, 4.6466, 4.8090, 4.5650, 4.6960, 5.2460, 4.7725, 4.4186], device='cuda:3'), covar=tensor([0.1581, 0.2377, 0.2318, 0.2293, 0.2509, 0.1044, 0.1679, 0.2704], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0639, 0.0697, 0.0518, 0.0691, 0.0726, 0.0542, 0.0693], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-01 23:57:17,853 INFO [train.py:904] (3/8) Epoch 25, batch 3950, loss[loss=0.167, simple_loss=0.2328, pruned_loss=0.05063, over 16675.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2491, pruned_loss=0.04445, over 3294900.51 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:57:19,963 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:30,948 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:05,353 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:12,120 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247590.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:28,588 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-01 23:58:30,195 INFO [train.py:904] (3/8) Epoch 25, batch 4000, loss[loss=0.1644, simple_loss=0.2474, pruned_loss=0.04069, over 16219.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2495, pruned_loss=0.04482, over 3288055.94 frames. ], batch size: 35, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:58:49,592 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:21,868 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:27,469 INFO [optim.py:368] (3/8) 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 23:59:44,176 INFO [train.py:904] (3/8) Epoch 25, batch 4050, loss[loss=0.17, simple_loss=0.2569, pruned_loss=0.04156, over 16476.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2502, pruned_loss=0.04428, over 3297425.59 frames. ], batch size: 75, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:59:59,690 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 00:00:52,241 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 00:00:59,185 INFO [train.py:904] (3/8) Epoch 25, batch 4100, loss[loss=0.1947, simple_loss=0.29, pruned_loss=0.04968, over 16782.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2518, pruned_loss=0.04393, over 3295183.72 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:01:00,675 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-02 00:01:25,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0365, 5.3963, 5.6131, 5.3109, 5.4378, 5.9722, 5.4776, 5.1107], device='cuda:3'), covar=tensor([0.0894, 0.1573, 0.1788, 0.1867, 0.2025, 0.0765, 0.1219, 0.2181], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0635, 0.0690, 0.0515, 0.0684, 0.0719, 0.0537, 0.0687], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 00:01:43,477 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0779, 4.7849, 4.7940, 5.2541, 5.4281, 4.8891, 5.3811, 5.4100], device='cuda:3'), covar=tensor([0.1701, 0.1225, 0.2292, 0.0914, 0.0679, 0.1023, 0.0881, 0.0886], device='cuda:3'), in_proj_covar=tensor([0.0681, 0.0836, 0.0964, 0.0845, 0.0647, 0.0666, 0.0692, 0.0811], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:01:48,423 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247735.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:01:57,928 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 4150, loss[loss=0.2302, simple_loss=0.3034, pruned_loss=0.07849, over 11462.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2587, pruned_loss=0.04642, over 3247300.87 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:02:42,287 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8024, 2.8523, 2.7368, 4.7950, 3.7622, 4.0588, 1.6307, 3.0280], device='cuda:3'), covar=tensor([0.1308, 0.0794, 0.1194, 0.0129, 0.0256, 0.0418, 0.1699, 0.0864], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0179, 0.0197, 0.0197, 0.0206, 0.0218, 0.0207, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:03:01,851 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:04,915 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247784.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:32,728 INFO [train.py:904] (3/8) Epoch 25, batch 4200, loss[loss=0.201, simple_loss=0.2932, pruned_loss=0.0544, over 16671.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2661, pruned_loss=0.04852, over 3197666.96 frames. ], batch size: 76, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:30,292 INFO [optim.py:368] (3/8) 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 00:04:35,247 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 4250, loss[loss=0.1945, simple_loss=0.2813, pruned_loss=0.05385, over 16439.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2689, pruned_loss=0.0476, over 3208132.32 frames. ], batch size: 35, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:52,105 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247856.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:28,708 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:32,131 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 00:06:01,866 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9913, 1.9065, 2.5823, 2.9096, 2.7890, 3.4766, 2.0694, 3.3338], device='cuda:3'), covar=tensor([0.0244, 0.0637, 0.0387, 0.0346, 0.0365, 0.0184, 0.0642, 0.0187], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0195, 0.0184, 0.0188, 0.0203, 0.0162, 0.0199, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:06:02,544 INFO [train.py:904] (3/8) Epoch 25, batch 4300, loss[loss=0.1977, simple_loss=0.2829, pruned_loss=0.05619, over 16596.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.27, pruned_loss=0.04665, over 3201424.87 frames. ], batch size: 57, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:06:13,457 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:07:05,048 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8388, 3.8880, 2.4515, 4.7720, 3.0290, 4.6464, 2.6538, 3.1994], device='cuda:3'), covar=tensor([0.0313, 0.0390, 0.1685, 0.0127, 0.0836, 0.0447, 0.1432, 0.0757], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0181, 0.0196, 0.0172, 0.0179, 0.0223, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:07:17,410 INFO [train.py:904] (3/8) Epoch 25, batch 4350, loss[loss=0.1825, simple_loss=0.2711, pruned_loss=0.04689, over 17215.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2735, pruned_loss=0.04764, over 3203287.80 frames. ], batch size: 44, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:08:05,111 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247984.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:23,775 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 4400, loss[loss=0.1924, simple_loss=0.2817, pruned_loss=0.05158, over 16956.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2755, pruned_loss=0.04891, over 3191765.45 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:24,767 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:34,455 INFO [optim.py:368] (3/8) 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 00:09:39,297 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 4450, loss[loss=0.2208, simple_loss=0.3092, pruned_loss=0.06619, over 16834.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2798, pruned_loss=0.05054, over 3197448.65 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:56,533 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,136 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,278 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:34,919 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248083.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:04,506 INFO [train.py:904] (3/8) Epoch 25, batch 4500, loss[loss=0.1982, simple_loss=0.2854, pruned_loss=0.05551, over 16216.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2799, pruned_loss=0.051, over 3196987.92 frames. ], batch size: 165, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:11:44,502 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:44,641 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0720, 5.1341, 4.9544, 4.5677, 4.6713, 5.0519, 4.8091, 4.7290], device='cuda:3'), covar=tensor([0.0491, 0.0315, 0.0229, 0.0263, 0.0759, 0.0281, 0.0346, 0.0536], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0456, 0.0356, 0.0361, 0.0360, 0.0414, 0.0244, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:12:00,013 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.873e+02 2.071e+02 2.399e+02 5.095e+02, threshold=4.142e+02, percent-clipped=1.0 2023-05-02 00:12:04,316 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:15,937 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 00:12:17,509 INFO [train.py:904] (3/8) Epoch 25, batch 4550, loss[loss=0.1976, simple_loss=0.2822, pruned_loss=0.0565, over 16902.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2811, pruned_loss=0.05184, over 3224981.25 frames. ], batch size: 109, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:12:22,789 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:00,083 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248181.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:19,548 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 00:13:23,230 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 00:13:32,543 INFO [train.py:904] (3/8) Epoch 25, batch 4600, loss[loss=0.1893, simple_loss=0.2791, pruned_loss=0.04974, over 17118.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2819, pruned_loss=0.05191, over 3231227.93 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:13:34,234 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:11,936 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248229.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:29,187 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.740e+02 1.903e+02 2.276e+02 6.899e+02, threshold=3.806e+02, percent-clipped=0.0 2023-05-02 00:14:36,196 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5997, 4.3544, 4.1857, 2.8862, 3.8052, 4.3250, 3.7673, 2.5964], device='cuda:3'), covar=tensor([0.0544, 0.0033, 0.0047, 0.0402, 0.0097, 0.0075, 0.0101, 0.0405], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 00:14:46,954 INFO [train.py:904] (3/8) Epoch 25, batch 4650, loss[loss=0.199, simple_loss=0.2789, pruned_loss=0.0596, over 16664.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2812, pruned_loss=0.05207, over 3217122.09 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:14:54,066 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 4700, loss[loss=0.1755, simple_loss=0.2703, pruned_loss=0.04038, over 16815.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.278, pruned_loss=0.05069, over 3223331.12 frames. ], batch size: 116, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:16:44,327 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:50,517 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6664, 4.8878, 5.0393, 4.8226, 4.8501, 5.4124, 4.8821, 4.5960], device='cuda:3'), covar=tensor([0.1224, 0.1624, 0.2027, 0.1887, 0.2290, 0.0892, 0.1481, 0.2476], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0622, 0.0678, 0.0506, 0.0672, 0.0709, 0.0527, 0.0677], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 00:16:56,009 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:17:14,747 INFO [train.py:904] (3/8) Epoch 25, batch 4750, loss[loss=0.1839, simple_loss=0.2742, pruned_loss=0.04678, over 16293.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2743, pruned_loss=0.04853, over 3225242.14 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:17:59,884 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:13,293 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:29,681 INFO [train.py:904] (3/8) Epoch 25, batch 4800, loss[loss=0.1851, simple_loss=0.274, pruned_loss=0.04813, over 16890.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2707, pruned_loss=0.04659, over 3230725.90 frames. ], batch size: 109, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:19:20,372 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0444, 4.1083, 4.3374, 4.3194, 4.3111, 4.0947, 4.0953, 4.0834], device='cuda:3'), covar=tensor([0.0318, 0.0478, 0.0386, 0.0359, 0.0381, 0.0365, 0.0730, 0.0451], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0463, 0.0454, 0.0415, 0.0497, 0.0474, 0.0552, 0.0377], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 00:19:22,239 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:24,713 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:19:31,516 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:43,834 INFO [train.py:904] (3/8) Epoch 25, batch 4850, loss[loss=0.2054, simple_loss=0.2944, pruned_loss=0.0582, over 12010.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2718, pruned_loss=0.04604, over 3210662.95 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:19:50,290 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2836, 4.3397, 4.6166, 4.5930, 4.5875, 4.3445, 4.3505, 4.2727], device='cuda:3'), covar=tensor([0.0315, 0.0459, 0.0334, 0.0336, 0.0383, 0.0332, 0.0744, 0.0417], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0461, 0.0452, 0.0413, 0.0495, 0.0473, 0.0551, 0.0376], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 00:20:36,774 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:20:58,431 INFO [train.py:904] (3/8) Epoch 25, batch 4900, loss[loss=0.1788, simple_loss=0.273, pruned_loss=0.04236, over 16864.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2707, pruned_loss=0.04479, over 3196320.89 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:22:03,445 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 4950, loss[loss=0.1834, simple_loss=0.2777, pruned_loss=0.04458, over 15359.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2702, pruned_loss=0.04397, over 3205924.94 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:31,264 INFO [train.py:904] (3/8) Epoch 25, batch 5000, loss[loss=0.1835, simple_loss=0.275, pruned_loss=0.04601, over 16895.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2724, pruned_loss=0.04441, over 3213368.38 frames. ], batch size: 109, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:35,296 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8181, 3.2738, 3.2158, 1.8791, 2.7884, 2.2389, 3.3334, 3.4747], device='cuda:3'), covar=tensor([0.0284, 0.0744, 0.0767, 0.2163, 0.0950, 0.1008, 0.0683, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0168, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:24:25,285 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248640.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:27,922 INFO [optim.py:368] (3/8) 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 00:24:42,933 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:43,832 INFO [train.py:904] (3/8) Epoch 25, batch 5050, loss[loss=0.1617, simple_loss=0.2544, pruned_loss=0.03453, over 16415.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2737, pruned_loss=0.04495, over 3194901.10 frames. ], batch size: 35, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:25:02,182 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5924, 4.6790, 4.4531, 4.1300, 4.1427, 4.5701, 4.3221, 4.2735], device='cuda:3'), covar=tensor([0.0620, 0.0471, 0.0342, 0.0338, 0.1061, 0.0513, 0.0487, 0.0658], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0457, 0.0356, 0.0360, 0.0360, 0.0416, 0.0243, 0.0427], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:25:35,140 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:35,153 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:39,579 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:56,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3692, 3.5229, 3.7336, 2.1021, 3.2071, 2.5055, 3.8141, 3.7306], device='cuda:3'), covar=tensor([0.0232, 0.0818, 0.0578, 0.2121, 0.0818, 0.0926, 0.0609, 0.0940], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0168, 0.0170, 0.0156, 0.0147, 0.0132, 0.0145, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 00:25:57,292 INFO [train.py:904] (3/8) Epoch 25, batch 5100, loss[loss=0.172, simple_loss=0.253, pruned_loss=0.04549, over 17034.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2713, pruned_loss=0.04398, over 3216453.92 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:26:15,814 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:49,979 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:52,089 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:27:12,742 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:13,424 INFO [train.py:904] (3/8) Epoch 25, batch 5150, loss[loss=0.1668, simple_loss=0.2698, pruned_loss=0.03186, over 16882.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2714, pruned_loss=0.04352, over 3192143.43 frames. ], batch size: 96, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:27:47,474 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:03,429 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:13,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7495, 3.1515, 3.3008, 2.0577, 2.8945, 2.2887, 3.3116, 3.3723], device='cuda:3'), covar=tensor([0.0270, 0.0766, 0.0621, 0.1940, 0.0870, 0.0961, 0.0609, 0.0750], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0167, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:28:28,598 INFO [train.py:904] (3/8) Epoch 25, batch 5200, loss[loss=0.1735, simple_loss=0.2529, pruned_loss=0.047, over 16578.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2692, pruned_loss=0.04264, over 3199591.65 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:28:44,935 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6627, 4.6166, 4.5536, 3.1611, 4.5893, 1.6271, 4.2238, 4.2274], device='cuda:3'), covar=tensor([0.0183, 0.0156, 0.0231, 0.0884, 0.0153, 0.3680, 0.0247, 0.0385], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0167, 0.0206, 0.0183, 0.0183, 0.0213, 0.0196, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:29:25,194 INFO [optim.py:368] (3/8) 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 00:29:41,117 INFO [train.py:904] (3/8) Epoch 25, batch 5250, loss[loss=0.1869, simple_loss=0.2764, pruned_loss=0.04873, over 16727.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2667, pruned_loss=0.04207, over 3206981.92 frames. ], batch size: 124, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:30:53,680 INFO [train.py:904] (3/8) Epoch 25, batch 5300, loss[loss=0.1481, simple_loss=0.2372, pruned_loss=0.02955, over 16903.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2631, pruned_loss=0.0411, over 3211945.69 frames. ], batch size: 96, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:31:45,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8237, 4.3722, 3.0943, 2.5158, 2.9164, 2.7340, 4.8048, 3.6859], device='cuda:3'), covar=tensor([0.2888, 0.0531, 0.1831, 0.2607, 0.2472, 0.1848, 0.0368, 0.1199], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0273, 0.0310, 0.0319, 0.0302, 0.0269, 0.0302, 0.0346], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 00:31:50,599 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6617, 2.6380, 1.8853, 2.7587, 2.0950, 2.8396, 2.1500, 2.3995], device='cuda:3'), covar=tensor([0.0318, 0.0367, 0.1410, 0.0229, 0.0691, 0.0448, 0.1234, 0.0662], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0179, 0.0194, 0.0168, 0.0177, 0.0218, 0.0202, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:31:51,342 INFO [optim.py:368] (3/8) 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 00:32:07,351 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2656, 2.4068, 2.4371, 3.9517, 2.2435, 2.7578, 2.4485, 2.5964], device='cuda:3'), covar=tensor([0.1398, 0.3346, 0.2823, 0.0542, 0.3945, 0.2292, 0.3472, 0.2928], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0463, 0.0380, 0.0334, 0.0442, 0.0531, 0.0435, 0.0540], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:32:07,997 INFO [train.py:904] (3/8) Epoch 25, batch 5350, loss[loss=0.1774, simple_loss=0.2648, pruned_loss=0.04503, over 16507.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2615, pruned_loss=0.04059, over 3210026.11 frames. ], batch size: 75, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:33:01,130 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248988.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:33:09,081 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 00:33:22,803 INFO [train.py:904] (3/8) Epoch 25, batch 5400, loss[loss=0.1781, simple_loss=0.28, pruned_loss=0.0381, over 16639.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2637, pruned_loss=0.04112, over 3206576.35 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:33:28,649 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 00:34:11,324 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:17,300 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 5450, loss[loss=0.1754, simple_loss=0.2576, pruned_loss=0.04659, over 17028.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2667, pruned_loss=0.04261, over 3193062.56 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:35:08,681 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:34,927 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249087.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:58,385 INFO [train.py:904] (3/8) Epoch 25, batch 5500, loss[loss=0.2073, simple_loss=0.297, pruned_loss=0.05884, over 16805.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2743, pruned_loss=0.04738, over 3145462.93 frames. ], batch size: 39, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:37:00,872 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 5550, loss[loss=0.225, simple_loss=0.3103, pruned_loss=0.06991, over 16912.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2808, pruned_loss=0.05159, over 3137288.66 frames. ], batch size: 109, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:38:41,394 INFO [train.py:904] (3/8) Epoch 25, batch 5600, loss[loss=0.244, simple_loss=0.3147, pruned_loss=0.08661, over 15292.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2861, pruned_loss=0.05626, over 3092604.21 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:39:02,184 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:39:47,608 INFO [optim.py:368] (3/8) 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 00:39:51,610 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8197, 3.1276, 3.2683, 2.0024, 2.8209, 2.2221, 3.2771, 3.4575], device='cuda:3'), covar=tensor([0.0277, 0.0823, 0.0598, 0.2090, 0.0911, 0.0993, 0.0685, 0.0928], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0167, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:40:04,581 INFO [train.py:904] (3/8) Epoch 25, batch 5650, loss[loss=0.2606, simple_loss=0.3263, pruned_loss=0.09748, over 11298.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.291, pruned_loss=0.06003, over 3074556.72 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:40:18,439 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9286, 4.2203, 3.1311, 2.5562, 2.9627, 2.7828, 4.6640, 3.6592], device='cuda:3'), covar=tensor([0.2823, 0.0635, 0.1810, 0.2653, 0.2530, 0.1877, 0.0400, 0.1315], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0273, 0.0311, 0.0321, 0.0304, 0.0270, 0.0303, 0.0346], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 00:40:41,286 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:40:57,660 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-02 00:41:22,856 INFO [train.py:904] (3/8) Epoch 25, batch 5700, loss[loss=0.2125, simple_loss=0.3049, pruned_loss=0.06001, over 16238.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2928, pruned_loss=0.06152, over 3071823.65 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:41:51,256 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1745, 5.4376, 5.1925, 5.2067, 4.9155, 4.8719, 4.8340, 5.5788], device='cuda:3'), covar=tensor([0.1223, 0.0910, 0.1137, 0.0984, 0.0854, 0.0888, 0.1304, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0694, 0.0836, 0.0692, 0.0646, 0.0532, 0.0533, 0.0703, 0.0654], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:42:25,338 INFO [optim.py:368] (3/8) 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 00:42:34,491 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:42:43,872 INFO [train.py:904] (3/8) Epoch 25, batch 5750, loss[loss=0.2112, simple_loss=0.2835, pruned_loss=0.06942, over 11473.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2955, pruned_loss=0.0629, over 3047114.14 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:43:13,839 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:43:53,992 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 5800, loss[loss=0.1732, simple_loss=0.2666, pruned_loss=0.03987, over 16653.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2946, pruned_loss=0.06139, over 3056456.26 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:44:26,245 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 00:44:32,818 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 00:45:26,289 INFO [train.py:904] (3/8) Epoch 25, batch 5850, loss[loss=0.2291, simple_loss=0.304, pruned_loss=0.07708, over 11569.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2923, pruned_loss=0.0599, over 3065468.89 frames. ], batch size: 250, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:46:22,498 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 00:46:46,208 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2382, 3.9754, 4.5533, 2.3266, 4.7283, 4.7267, 3.4338, 3.6398], device='cuda:3'), covar=tensor([0.0692, 0.0281, 0.0160, 0.1143, 0.0070, 0.0134, 0.0368, 0.0413], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0141, 0.0085, 0.0131, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:46:46,939 INFO [train.py:904] (3/8) Epoch 25, batch 5900, loss[loss=0.1823, simple_loss=0.2735, pruned_loss=0.04556, over 17051.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2917, pruned_loss=0.05966, over 3069120.18 frames. ], batch size: 41, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:47:24,623 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 00:47:47,875 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0752, 3.9801, 4.1588, 4.2597, 4.4017, 3.9915, 4.3511, 4.4246], device='cuda:3'), covar=tensor([0.1896, 0.1218, 0.1372, 0.0737, 0.0680, 0.1502, 0.0948, 0.0738], device='cuda:3'), in_proj_covar=tensor([0.0660, 0.0812, 0.0933, 0.0818, 0.0627, 0.0645, 0.0673, 0.0786], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:47:49,723 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8911, 3.6732, 4.1673, 2.0193, 4.3249, 4.3745, 3.2838, 3.3052], device='cuda:3'), covar=tensor([0.0736, 0.0301, 0.0169, 0.1298, 0.0079, 0.0170, 0.0377, 0.0442], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0101, 0.0141, 0.0085, 0.0131, 0.0130, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:47:52,231 INFO [optim.py:368] (3/8) 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 00:48:08,064 INFO [train.py:904] (3/8) Epoch 25, batch 5950, loss[loss=0.1924, simple_loss=0.2871, pruned_loss=0.04891, over 16776.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2924, pruned_loss=0.05867, over 3064492.02 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:48:36,946 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:48:58,277 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0227, 2.4114, 2.0444, 2.2507, 2.7428, 2.4591, 2.6304, 2.9279], device='cuda:3'), covar=tensor([0.0186, 0.0431, 0.0570, 0.0482, 0.0294, 0.0395, 0.0259, 0.0297], device='cuda:3'), in_proj_covar=tensor([0.0220, 0.0239, 0.0227, 0.0230, 0.0239, 0.0238, 0.0240, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:49:28,143 INFO [train.py:904] (3/8) Epoch 25, batch 6000, loss[loss=0.1912, simple_loss=0.2834, pruned_loss=0.04944, over 16570.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2908, pruned_loss=0.05776, over 3070015.55 frames. ], batch size: 75, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:49:28,144 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 00:49:37,139 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8951, 2.6537, 2.6057, 4.0099, 2.7097, 3.9925, 1.6801, 3.1028], device='cuda:3'), covar=tensor([0.1352, 0.0803, 0.1203, 0.0199, 0.0133, 0.0399, 0.1700, 0.0739], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0206, 0.0216, 0.0206, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:49:38,599 INFO [train.py:938] (3/8) Epoch 25, validation: loss=0.1487, simple_loss=0.2613, pruned_loss=0.01811, over 944034.00 frames. 2023-05-02 00:49:38,599 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 00:49:40,405 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4497, 2.9321, 3.0620, 2.0350, 2.7421, 2.0952, 3.1260, 3.2103], device='cuda:3'), covar=tensor([0.0241, 0.0774, 0.0585, 0.1977, 0.0860, 0.1025, 0.0587, 0.0894], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:50:36,538 INFO [optim.py:368] (3/8) 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 00:50:38,120 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 00:50:54,725 INFO [train.py:904] (3/8) Epoch 25, batch 6050, loss[loss=0.1838, simple_loss=0.2795, pruned_loss=0.04406, over 16846.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2894, pruned_loss=0.05685, over 3098478.10 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:51:35,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0316, 5.0971, 4.9134, 4.4896, 4.5406, 4.9760, 4.8305, 4.6493], device='cuda:3'), covar=tensor([0.0805, 0.0874, 0.0398, 0.0434, 0.1157, 0.0646, 0.0571, 0.0891], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0449, 0.0349, 0.0354, 0.0354, 0.0408, 0.0239, 0.0422], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:51:43,346 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 00:52:12,397 INFO [train.py:904] (3/8) Epoch 25, batch 6100, loss[loss=0.1855, simple_loss=0.2847, pruned_loss=0.0431, over 16901.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2886, pruned_loss=0.05581, over 3099324.69 frames. ], batch size: 96, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:53:15,555 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 6150, loss[loss=0.1693, simple_loss=0.2631, pruned_loss=0.03779, over 16757.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2868, pruned_loss=0.05486, over 3132566.36 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:54:51,553 INFO [train.py:904] (3/8) Epoch 25, batch 6200, loss[loss=0.2007, simple_loss=0.2854, pruned_loss=0.05795, over 16630.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2848, pruned_loss=0.05454, over 3118162.44 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:54:57,713 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.65 vs. limit=5.0 2023-05-02 00:55:03,136 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0220, 5.0144, 4.8398, 4.1089, 4.9281, 1.9284, 4.6961, 4.5768], device='cuda:3'), covar=tensor([0.0110, 0.0098, 0.0198, 0.0447, 0.0109, 0.2794, 0.0148, 0.0259], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0167, 0.0207, 0.0184, 0.0184, 0.0213, 0.0196, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 00:55:10,535 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:55:55,230 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.750e+02 3.284e+02 4.072e+02 6.772e+02, threshold=6.569e+02, percent-clipped=1.0 2023-05-02 00:56:10,145 INFO [train.py:904] (3/8) Epoch 25, batch 6250, loss[loss=0.1974, simple_loss=0.2974, pruned_loss=0.04873, over 16596.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2844, pruned_loss=0.05421, over 3126615.18 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:56:38,966 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249871.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:56:40,165 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3277, 3.1125, 3.4680, 1.8350, 3.5546, 3.6209, 2.8504, 2.6887], device='cuda:3'), covar=tensor([0.0857, 0.0298, 0.0180, 0.1241, 0.0099, 0.0187, 0.0449, 0.0521], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0140, 0.0085, 0.0130, 0.0130, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 00:56:44,485 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:57:26,264 INFO [train.py:904] (3/8) Epoch 25, batch 6300, loss[loss=0.2129, simple_loss=0.284, pruned_loss=0.07093, over 11588.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2845, pruned_loss=0.0541, over 3122510.55 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:57:52,578 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 6350, loss[loss=0.1985, simple_loss=0.2856, pruned_loss=0.05571, over 16965.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2858, pruned_loss=0.05567, over 3092534.96 frames. ], batch size: 109, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:58:48,578 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249955.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 00:58:54,454 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 00:58:59,146 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7976, 4.8488, 5.1947, 5.1764, 5.2153, 4.8994, 4.8463, 4.6837], device='cuda:3'), covar=tensor([0.0330, 0.0587, 0.0379, 0.0423, 0.0468, 0.0404, 0.0949, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0471, 0.0458, 0.0418, 0.0503, 0.0480, 0.0558, 0.0383], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 00:59:01,744 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 00:59:20,608 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249976.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:00:04,120 INFO [train.py:904] (3/8) Epoch 25, batch 6400, loss[loss=0.1715, simple_loss=0.2656, pruned_loss=0.03877, over 16714.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2861, pruned_loss=0.05701, over 3084537.54 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:00:13,464 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-02 01:00:23,802 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250016.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:00:46,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8306, 3.7433, 4.2547, 2.1143, 4.4000, 4.4173, 3.3473, 3.2651], device='cuda:3'), covar=tensor([0.0777, 0.0274, 0.0157, 0.1230, 0.0067, 0.0165, 0.0377, 0.0459], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0141, 0.0085, 0.0131, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:00:56,490 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:01:01,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8659, 5.1337, 4.8894, 4.9407, 4.6576, 4.6346, 4.5827, 5.2179], device='cuda:3'), covar=tensor([0.1210, 0.0841, 0.1088, 0.0983, 0.0850, 0.1097, 0.1243, 0.0921], device='cuda:3'), in_proj_covar=tensor([0.0694, 0.0838, 0.0692, 0.0647, 0.0533, 0.0534, 0.0705, 0.0656], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:01:05,937 INFO [optim.py:368] (3/8) 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 01:01:16,820 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 01:01:20,199 INFO [train.py:904] (3/8) Epoch 25, batch 6450, loss[loss=0.1954, simple_loss=0.2746, pruned_loss=0.05804, over 11622.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2863, pruned_loss=0.05607, over 3089615.03 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:01:33,128 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8728, 1.4446, 1.7780, 1.7140, 1.8027, 1.8904, 1.6871, 1.8007], device='cuda:3'), covar=tensor([0.0251, 0.0409, 0.0245, 0.0300, 0.0345, 0.0224, 0.0471, 0.0179], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0195, 0.0183, 0.0187, 0.0202, 0.0161, 0.0200, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:02:36,758 INFO [train.py:904] (3/8) Epoch 25, batch 6500, loss[loss=0.2199, simple_loss=0.2863, pruned_loss=0.07674, over 11761.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.285, pruned_loss=0.05639, over 3052721.98 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:03:30,065 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 01:03:40,563 INFO [optim.py:368] (3/8) 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 01:03:52,688 INFO [train.py:904] (3/8) Epoch 25, batch 6550, loss[loss=0.1872, simple_loss=0.294, pruned_loss=0.04019, over 16925.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2874, pruned_loss=0.05632, over 3081499.68 frames. ], batch size: 96, lr: 2.69e-03, grad_scale: 4.0 2023-05-02 01:04:18,811 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:04:30,945 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 01:05:05,603 INFO [train.py:904] (3/8) Epoch 25, batch 6600, loss[loss=0.2506, simple_loss=0.3225, pruned_loss=0.08933, over 11612.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2901, pruned_loss=0.05687, over 3098269.99 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:05:13,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6451, 4.6983, 5.0527, 5.0237, 5.0428, 4.7457, 4.6830, 4.5420], device='cuda:3'), covar=tensor([0.0330, 0.0568, 0.0359, 0.0408, 0.0449, 0.0422, 0.0999, 0.0536], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0470, 0.0457, 0.0418, 0.0501, 0.0478, 0.0557, 0.0382], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 01:06:06,344 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7730, 1.7992, 2.4230, 2.7660, 2.6513, 3.1381, 1.9705, 3.1079], device='cuda:3'), covar=tensor([0.0223, 0.0641, 0.0354, 0.0339, 0.0381, 0.0222, 0.0651, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0195, 0.0182, 0.0186, 0.0202, 0.0161, 0.0199, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:06:08,494 INFO [optim.py:368] (3/8) 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 01:06:10,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-05-02 01:06:21,898 INFO [train.py:904] (3/8) Epoch 25, batch 6650, loss[loss=0.1969, simple_loss=0.282, pruned_loss=0.05584, over 16683.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2899, pruned_loss=0.05713, over 3106749.44 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:06:49,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4368, 2.6481, 2.3144, 2.4945, 2.9467, 2.5881, 2.9941, 3.1796], device='cuda:3'), covar=tensor([0.0130, 0.0427, 0.0504, 0.0447, 0.0273, 0.0411, 0.0234, 0.0252], device='cuda:3'), in_proj_covar=tensor([0.0218, 0.0237, 0.0227, 0.0228, 0.0238, 0.0237, 0.0237, 0.0235], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:07:37,544 INFO [train.py:904] (3/8) Epoch 25, batch 6700, loss[loss=0.1879, simple_loss=0.2786, pruned_loss=0.04858, over 17050.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2885, pruned_loss=0.05691, over 3120179.51 frames. ], batch size: 55, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:50,661 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250311.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:08:23,224 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:08:41,139 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 6750, loss[loss=0.2365, simple_loss=0.3065, pruned_loss=0.08324, over 11842.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2874, pruned_loss=0.05726, over 3107448.71 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:10:10,573 INFO [train.py:904] (3/8) Epoch 25, batch 6800, loss[loss=0.2063, simple_loss=0.2973, pruned_loss=0.05768, over 16782.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2879, pruned_loss=0.05747, over 3086894.00 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:10:36,328 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7405, 1.8037, 1.6485, 1.5199, 1.9366, 1.5692, 1.6222, 1.9188], device='cuda:3'), covar=tensor([0.0200, 0.0291, 0.0395, 0.0337, 0.0203, 0.0283, 0.0164, 0.0205], device='cuda:3'), in_proj_covar=tensor([0.0219, 0.0239, 0.0228, 0.0229, 0.0239, 0.0238, 0.0238, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:11:16,509 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.847e+02 3.369e+02 4.013e+02 7.021e+02, threshold=6.738e+02, percent-clipped=2.0 2023-05-02 01:11:27,500 INFO [train.py:904] (3/8) Epoch 25, batch 6850, loss[loss=0.2209, simple_loss=0.3318, pruned_loss=0.05496, over 16613.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2895, pruned_loss=0.05812, over 3084227.83 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:52,829 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:12:15,553 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 01:12:43,350 INFO [train.py:904] (3/8) Epoch 25, batch 6900, loss[loss=0.1815, simple_loss=0.2831, pruned_loss=0.03992, over 16635.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2905, pruned_loss=0.05651, over 3104176.11 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:13:06,142 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:13:47,080 INFO [optim.py:368] (3/8) 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 01:14:00,191 INFO [train.py:904] (3/8) Epoch 25, batch 6950, loss[loss=0.2204, simple_loss=0.3124, pruned_loss=0.06417, over 16405.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2926, pruned_loss=0.0587, over 3090969.82 frames. ], batch size: 68, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:14:43,074 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8285, 3.7204, 3.8976, 3.9958, 4.0923, 3.6939, 4.0323, 4.1123], device='cuda:3'), covar=tensor([0.1686, 0.1267, 0.1402, 0.0785, 0.0660, 0.1868, 0.1000, 0.0801], device='cuda:3'), in_proj_covar=tensor([0.0651, 0.0800, 0.0925, 0.0809, 0.0621, 0.0639, 0.0672, 0.0781], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:15:16,490 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 01:15:18,378 INFO [train.py:904] (3/8) Epoch 25, batch 7000, loss[loss=0.2022, simple_loss=0.2946, pruned_loss=0.05493, over 17057.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2925, pruned_loss=0.05789, over 3097041.46 frames. ], batch size: 55, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:31,368 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250611.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:03,412 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:16:22,012 INFO [optim.py:368] (3/8) 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 01:16:29,075 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4335, 3.2709, 2.5232, 2.0514, 2.2637, 2.1181, 3.4152, 2.9740], device='cuda:3'), covar=tensor([0.3289, 0.0801, 0.2183, 0.2961, 0.3031, 0.2562, 0.0574, 0.1589], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0273, 0.0311, 0.0320, 0.0303, 0.0270, 0.0301, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 01:16:35,090 INFO [train.py:904] (3/8) Epoch 25, batch 7050, loss[loss=0.187, simple_loss=0.2796, pruned_loss=0.04719, over 17183.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2933, pruned_loss=0.05817, over 3087561.93 frames. ], batch size: 46, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:16:44,701 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250659.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:49,612 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:16:53,680 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 01:17:14,214 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0875, 3.2892, 3.4812, 2.1679, 2.9444, 2.2531, 3.4130, 3.5613], device='cuda:3'), covar=tensor([0.0284, 0.0882, 0.0658, 0.2169, 0.0979, 0.1093, 0.0731, 0.1155], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0168, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 01:17:17,436 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250680.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:51,099 INFO [train.py:904] (3/8) Epoch 25, batch 7100, loss[loss=0.1988, simple_loss=0.2882, pruned_loss=0.05468, over 15263.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.292, pruned_loss=0.05791, over 3097198.23 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:17:56,449 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0117, 4.0996, 3.9129, 3.6247, 3.6796, 4.0254, 3.6656, 3.7965], device='cuda:3'), covar=tensor([0.0606, 0.0610, 0.0349, 0.0332, 0.0774, 0.0504, 0.1183, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0297, 0.0447, 0.0347, 0.0352, 0.0353, 0.0405, 0.0239, 0.0421], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:18:00,374 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0475, 2.3330, 2.3299, 2.8230, 2.0336, 3.2368, 1.8637, 2.7481], device='cuda:3'), covar=tensor([0.1089, 0.0597, 0.1015, 0.0180, 0.0108, 0.0330, 0.1331, 0.0661], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0180, 0.0198, 0.0197, 0.0207, 0.0217, 0.0208, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:18:23,411 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250723.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:18:56,841 INFO [optim.py:368] (3/8) 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 01:19:09,287 INFO [train.py:904] (3/8) Epoch 25, batch 7150, loss[loss=0.2324, simple_loss=0.2984, pruned_loss=0.0832, over 11245.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2895, pruned_loss=0.05752, over 3086201.68 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:20:22,681 INFO [train.py:904] (3/8) Epoch 25, batch 7200, loss[loss=0.1599, simple_loss=0.2578, pruned_loss=0.031, over 15425.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.287, pruned_loss=0.05585, over 3074869.32 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:21:08,899 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,112 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 7250, loss[loss=0.1723, simple_loss=0.2587, pruned_loss=0.04298, over 16849.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2842, pruned_loss=0.05422, over 3107313.68 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:22:42,138 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:22:55,036 INFO [train.py:904] (3/8) Epoch 25, batch 7300, loss[loss=0.2418, simple_loss=0.3051, pruned_loss=0.08927, over 11467.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.284, pruned_loss=0.05471, over 3090432.83 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:23:00,827 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250906.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:23:59,649 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 7350, loss[loss=0.2132, simple_loss=0.3018, pruned_loss=0.06225, over 16240.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2858, pruned_loss=0.05621, over 3068975.34 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:27,062 INFO [train.py:904] (3/8) Epoch 25, batch 7400, loss[loss=0.211, simple_loss=0.2955, pruned_loss=0.06331, over 16380.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2867, pruned_loss=0.05716, over 3052818.77 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:51,014 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:26:35,984 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 7450, loss[loss=0.2072, simple_loss=0.3073, pruned_loss=0.05358, over 16250.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2876, pruned_loss=0.05748, over 3078378.11 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:26:47,437 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7722, 1.4507, 1.7361, 1.6833, 1.8445, 1.8889, 1.6258, 1.8503], device='cuda:3'), covar=tensor([0.0245, 0.0349, 0.0212, 0.0268, 0.0248, 0.0205, 0.0408, 0.0142], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0194, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:28:06,268 INFO [train.py:904] (3/8) Epoch 25, batch 7500, loss[loss=0.1987, simple_loss=0.2839, pruned_loss=0.05675, over 16858.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2877, pruned_loss=0.05667, over 3061073.88 frames. ], batch size: 42, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:46,850 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8752, 4.1362, 3.9822, 4.0187, 3.6558, 3.7590, 3.8344, 4.1450], device='cuda:3'), covar=tensor([0.1158, 0.0908, 0.1056, 0.0906, 0.0850, 0.1898, 0.0955, 0.0992], device='cuda:3'), in_proj_covar=tensor([0.0693, 0.0840, 0.0692, 0.0647, 0.0532, 0.0538, 0.0703, 0.0654], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:29:12,776 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.915e+02 2.911e+02 3.474e+02 4.178e+02 7.715e+02, threshold=6.948e+02, percent-clipped=3.0 2023-05-02 01:29:23,951 INFO [train.py:904] (3/8) Epoch 25, batch 7550, loss[loss=0.1856, simple_loss=0.281, pruned_loss=0.0451, over 16672.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2872, pruned_loss=0.05734, over 3034157.15 frames. ], batch size: 89, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:30:00,939 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 01:30:19,160 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:32,691 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:37,923 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251201.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:30:39,949 INFO [train.py:904] (3/8) Epoch 25, batch 7600, loss[loss=0.1839, simple_loss=0.2833, pruned_loss=0.04229, over 16681.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2867, pruned_loss=0.05756, over 3038836.77 frames. ], batch size: 89, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:30:46,581 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0886, 1.9878, 2.5555, 3.0448, 2.8576, 3.5902, 2.1785, 3.4677], device='cuda:3'), covar=tensor([0.0241, 0.0581, 0.0411, 0.0337, 0.0358, 0.0154, 0.0626, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0195, 0.0183, 0.0186, 0.0201, 0.0160, 0.0199, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:30:49,333 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 01:31:43,657 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 7650, loss[loss=0.1847, simple_loss=0.2766, pruned_loss=0.04643, over 16275.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2868, pruned_loss=0.05778, over 3049822.79 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:32:02,572 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251259.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:32:58,493 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0994, 5.1078, 4.9304, 4.5276, 4.6127, 5.0087, 4.9168, 4.6656], device='cuda:3'), covar=tensor([0.0556, 0.0457, 0.0299, 0.0341, 0.0976, 0.0479, 0.0293, 0.0718], device='cuda:3'), in_proj_covar=tensor([0.0296, 0.0445, 0.0344, 0.0350, 0.0350, 0.0402, 0.0238, 0.0418], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:33:07,010 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 7700, loss[loss=0.1794, simple_loss=0.2746, pruned_loss=0.04212, over 16899.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2863, pruned_loss=0.05752, over 3070251.63 frames. ], batch size: 96, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:33:31,789 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 7750, loss[loss=0.2461, simple_loss=0.3173, pruned_loss=0.08746, over 11926.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2866, pruned_loss=0.05729, over 3070201.23 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:34:39,727 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:45,579 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:35:35,163 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5860, 2.6936, 2.2592, 2.5085, 3.0562, 2.6827, 3.1005, 3.2795], device='cuda:3'), covar=tensor([0.0133, 0.0391, 0.0553, 0.0430, 0.0276, 0.0404, 0.0262, 0.0226], device='cuda:3'), in_proj_covar=tensor([0.0216, 0.0235, 0.0225, 0.0227, 0.0236, 0.0234, 0.0234, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:35:40,750 INFO [train.py:904] (3/8) Epoch 25, batch 7800, loss[loss=0.188, simple_loss=0.2787, pruned_loss=0.04864, over 17051.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2869, pruned_loss=0.05749, over 3091453.67 frames. ], batch size: 55, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:35:42,387 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251404.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:36:45,155 INFO [optim.py:368] (3/8) 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 01:36:55,300 INFO [train.py:904] (3/8) Epoch 25, batch 7850, loss[loss=0.1694, simple_loss=0.2667, pruned_loss=0.03601, over 16818.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.288, pruned_loss=0.05649, over 3115936.38 frames. ], batch size: 102, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:36:58,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3401, 2.8271, 3.0865, 1.9432, 2.7522, 2.1037, 2.9932, 3.0657], device='cuda:3'), covar=tensor([0.0281, 0.0852, 0.0598, 0.2149, 0.0884, 0.1076, 0.0687, 0.0919], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:37:14,496 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251465.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:28,295 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6936, 1.7786, 1.6027, 1.4909, 1.8739, 1.5393, 1.5339, 1.8492], device='cuda:3'), covar=tensor([0.0226, 0.0307, 0.0437, 0.0338, 0.0225, 0.0287, 0.0172, 0.0215], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0235, 0.0225, 0.0226, 0.0236, 0.0234, 0.0234, 0.0233], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:37:50,393 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:38:07,398 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251501.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:38:09,239 INFO [train.py:904] (3/8) Epoch 25, batch 7900, loss[loss=0.2151, simple_loss=0.3042, pruned_loss=0.06302, over 16703.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2881, pruned_loss=0.05682, over 3105864.88 frames. ], batch size: 124, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:38:40,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7422, 3.7600, 3.9190, 3.6972, 3.8738, 4.2304, 3.8716, 3.6341], device='cuda:3'), covar=tensor([0.2454, 0.2308, 0.2450, 0.2479, 0.2702, 0.1749, 0.1715, 0.2686], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0615, 0.0679, 0.0504, 0.0670, 0.0703, 0.0528, 0.0677], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 01:39:04,226 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:17,277 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 2.635e+02 3.162e+02 3.936e+02 7.872e+02, threshold=6.324e+02, percent-clipped=1.0 2023-05-02 01:39:22,129 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:28,549 INFO [train.py:904] (3/8) Epoch 25, batch 7950, loss[loss=0.2091, simple_loss=0.2853, pruned_loss=0.0665, over 11507.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2886, pruned_loss=0.05763, over 3087597.68 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:39:30,109 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:40:46,595 INFO [train.py:904] (3/8) Epoch 25, batch 8000, loss[loss=0.1946, simple_loss=0.2926, pruned_loss=0.04826, over 15353.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2894, pruned_loss=0.05904, over 3058765.12 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:41:44,798 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 8050, loss[loss=0.2091, simple_loss=0.2919, pruned_loss=0.0632, over 15454.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2895, pruned_loss=0.05885, over 3061438.18 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:42:09,017 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:42:54,700 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7504, 4.7831, 5.1189, 5.0847, 5.1249, 4.7879, 4.7883, 4.6005], device='cuda:3'), covar=tensor([0.0326, 0.0534, 0.0369, 0.0412, 0.0524, 0.0380, 0.1039, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0475, 0.0460, 0.0423, 0.0507, 0.0486, 0.0561, 0.0387], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 01:43:17,030 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 8100, loss[loss=0.1811, simple_loss=0.2686, pruned_loss=0.04679, over 16645.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2888, pruned_loss=0.05779, over 3079772.66 frames. ], batch size: 62, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:43:27,261 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251709.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:44:04,228 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2620, 3.3269, 2.0435, 3.5689, 2.4639, 3.5982, 2.1748, 2.7254], device='cuda:3'), covar=tensor([0.0326, 0.0430, 0.1693, 0.0285, 0.0880, 0.0724, 0.1569, 0.0855], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0168, 0.0178, 0.0219, 0.0204, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:44:22,959 INFO [optim.py:368] (3/8) 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 01:44:29,015 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 01:44:33,793 INFO [train.py:904] (3/8) Epoch 25, batch 8150, loss[loss=0.1875, simple_loss=0.2777, pruned_loss=0.0487, over 16693.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2862, pruned_loss=0.0569, over 3081291.17 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:44:45,086 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251760.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:00,599 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:00,977 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 01:45:50,905 INFO [train.py:904] (3/8) Epoch 25, batch 8200, loss[loss=0.206, simple_loss=0.2933, pruned_loss=0.05936, over 15227.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2841, pruned_loss=0.05614, over 3088126.47 frames. ], batch size: 190, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:46:59,192 INFO [optim.py:368] (3/8) 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 01:47:01,866 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3962, 3.5171, 2.1949, 3.8424, 2.6124, 3.8152, 2.2482, 2.8637], device='cuda:3'), covar=tensor([0.0322, 0.0383, 0.1591, 0.0275, 0.0876, 0.0591, 0.1592, 0.0745], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0167, 0.0177, 0.0218, 0.0204, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:47:10,655 INFO [train.py:904] (3/8) Epoch 25, batch 8250, loss[loss=0.1811, simple_loss=0.2788, pruned_loss=0.0417, over 16281.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2832, pruned_loss=0.05382, over 3078016.88 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:47:12,519 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:47:15,443 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7044, 4.7308, 4.5584, 4.1047, 4.1979, 4.6241, 4.4697, 4.3100], device='cuda:3'), covar=tensor([0.0605, 0.0633, 0.0364, 0.0433, 0.1034, 0.0607, 0.0460, 0.0796], device='cuda:3'), in_proj_covar=tensor([0.0295, 0.0445, 0.0344, 0.0349, 0.0349, 0.0400, 0.0238, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:48:29,881 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:30,710 INFO [train.py:904] (3/8) Epoch 25, batch 8300, loss[loss=0.1572, simple_loss=0.2618, pruned_loss=0.02633, over 16707.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2807, pruned_loss=0.05097, over 3071365.47 frames. ], batch size: 76, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:41,163 INFO [optim.py:368] (3/8) 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 01:49:41,693 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6655, 4.0045, 3.9555, 2.7940, 3.5161, 3.9930, 3.6082, 2.1907], device='cuda:3'), covar=tensor([0.0508, 0.0064, 0.0061, 0.0413, 0.0121, 0.0117, 0.0098, 0.0600], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0086, 0.0087, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 01:49:52,624 INFO [train.py:904] (3/8) Epoch 25, batch 8350, loss[loss=0.1596, simple_loss=0.2628, pruned_loss=0.02823, over 16852.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2793, pruned_loss=0.04857, over 3071090.17 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:59,393 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:50:09,652 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 01:50:41,575 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0085, 4.4540, 3.2318, 2.4271, 2.7770, 2.7646, 4.7463, 3.7366], device='cuda:3'), covar=tensor([0.2726, 0.0472, 0.1799, 0.3188, 0.3020, 0.2031, 0.0348, 0.1289], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0269, 0.0307, 0.0316, 0.0300, 0.0267, 0.0297, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 01:51:05,350 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 8400, loss[loss=0.176, simple_loss=0.2737, pruned_loss=0.03912, over 16671.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2764, pruned_loss=0.04674, over 3052624.97 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:51:22,059 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 01:52:40,175 INFO [train.py:904] (3/8) Epoch 25, batch 8450, loss[loss=0.1696, simple_loss=0.2655, pruned_loss=0.03686, over 15363.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2745, pruned_loss=0.04488, over 3045054.89 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:52:51,809 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:53:14,041 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0623, 1.6057, 1.9609, 2.1183, 2.2564, 2.3105, 1.8785, 2.2780], device='cuda:3'), covar=tensor([0.0248, 0.0502, 0.0300, 0.0303, 0.0320, 0.0219, 0.0454, 0.0160], device='cuda:3'), in_proj_covar=tensor([0.0190, 0.0193, 0.0181, 0.0183, 0.0199, 0.0159, 0.0196, 0.0158], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:54:01,927 INFO [train.py:904] (3/8) Epoch 25, batch 8500, loss[loss=0.1544, simple_loss=0.2412, pruned_loss=0.03382, over 12261.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2717, pruned_loss=0.04335, over 3060693.92 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:54:11,540 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252108.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:55:15,325 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 8550, loss[loss=0.1755, simple_loss=0.2601, pruned_loss=0.04548, over 12108.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2689, pruned_loss=0.04224, over 3039303.37 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:56:00,482 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8356, 3.1507, 2.6181, 4.8009, 3.5116, 4.4032, 1.6185, 3.2063], device='cuda:3'), covar=tensor([0.1595, 0.0742, 0.1295, 0.0180, 0.0196, 0.0387, 0.1880, 0.0704], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0177, 0.0195, 0.0193, 0.0204, 0.0214, 0.0205, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:57:09,035 INFO [train.py:904] (3/8) Epoch 25, batch 8600, loss[loss=0.1632, simple_loss=0.2601, pruned_loss=0.03315, over 16859.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.269, pruned_loss=0.0412, over 3053422.21 frames. ], batch size: 90, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:57:16,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2465, 4.2314, 4.1248, 3.3455, 4.1855, 1.6884, 3.9549, 3.7757], device='cuda:3'), covar=tensor([0.0115, 0.0113, 0.0188, 0.0302, 0.0102, 0.2889, 0.0152, 0.0310], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0163, 0.0202, 0.0179, 0.0179, 0.0210, 0.0191, 0.0172], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:58:04,986 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8819, 2.1642, 2.4186, 3.1287, 2.1720, 2.3076, 2.3347, 2.2567], device='cuda:3'), covar=tensor([0.1439, 0.3900, 0.2888, 0.0794, 0.4655, 0.2875, 0.3778, 0.3813], device='cuda:3'), in_proj_covar=tensor([0.0404, 0.0454, 0.0372, 0.0325, 0.0433, 0.0517, 0.0425, 0.0529], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 01:58:34,356 INFO [optim.py:368] (3/8) 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 01:58:39,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0567, 3.2419, 3.7948, 1.9415, 3.0836, 2.3227, 3.5330, 3.4806], device='cuda:3'), covar=tensor([0.0263, 0.0969, 0.0470, 0.2396, 0.0824, 0.1045, 0.0693, 0.1062], device='cuda:3'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0152, 0.0144, 0.0128, 0.0142, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 01:58:48,811 INFO [train.py:904] (3/8) Epoch 25, batch 8650, loss[loss=0.1655, simple_loss=0.267, pruned_loss=0.03194, over 16659.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2669, pruned_loss=0.03987, over 3035507.36 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:58:52,538 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 02:00:23,910 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252297.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:00:33,689 INFO [train.py:904] (3/8) Epoch 25, batch 8700, loss[loss=0.1683, simple_loss=0.2547, pruned_loss=0.04098, over 12634.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2646, pruned_loss=0.03864, over 3034432.55 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:01:15,844 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 02:01:53,641 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 8750, loss[loss=0.2026, simple_loss=0.2937, pruned_loss=0.05579, over 15402.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2639, pruned_loss=0.03844, over 3011505.48 frames. ], batch size: 192, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:02:41,197 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252365.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:03:47,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6941, 2.6890, 1.9190, 2.8602, 2.1240, 2.8660, 2.1401, 2.4234], device='cuda:3'), covar=tensor([0.0311, 0.0369, 0.1306, 0.0282, 0.0712, 0.0482, 0.1297, 0.0619], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0174, 0.0191, 0.0162, 0.0173, 0.0211, 0.0200, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 02:04:01,511 INFO [train.py:904] (3/8) Epoch 25, batch 8800, loss[loss=0.1787, simple_loss=0.2737, pruned_loss=0.04188, over 16412.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2632, pruned_loss=0.03755, over 3041187.64 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:04:21,824 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:30,827 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252417.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:56,273 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8927, 3.9184, 4.0404, 3.8347, 3.9392, 4.3653, 3.9889, 3.7413], device='cuda:3'), covar=tensor([0.2045, 0.1885, 0.1770, 0.2136, 0.2625, 0.1316, 0.1569, 0.2397], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0595, 0.0661, 0.0491, 0.0650, 0.0686, 0.0513, 0.0656], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 02:05:07,182 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:31,498 INFO [optim.py:368] (3/8) 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 02:05:39,617 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4311, 3.4895, 3.6566, 3.6602, 3.6857, 3.5096, 3.5526, 3.5418], device='cuda:3'), covar=tensor([0.0337, 0.0624, 0.0546, 0.0447, 0.0408, 0.0530, 0.0653, 0.0485], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0462, 0.0449, 0.0411, 0.0494, 0.0472, 0.0545, 0.0377], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 02:05:46,035 INFO [train.py:904] (3/8) Epoch 25, batch 8850, loss[loss=0.1658, simple_loss=0.2709, pruned_loss=0.03036, over 16182.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2657, pruned_loss=0.0369, over 3034286.38 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:06:39,677 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252478.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:06:48,945 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 02:07:12,115 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4492, 3.7457, 3.7522, 2.6804, 3.3598, 3.7927, 3.5365, 2.2507], device='cuda:3'), covar=tensor([0.0488, 0.0051, 0.0046, 0.0345, 0.0103, 0.0077, 0.0079, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0084, 0.0086, 0.0132, 0.0098, 0.0109, 0.0094, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 02:07:12,299 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 02:07:16,679 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:31,608 INFO [train.py:904] (3/8) Epoch 25, batch 8900, loss[loss=0.1782, simple_loss=0.2723, pruned_loss=0.04209, over 15268.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2656, pruned_loss=0.03617, over 3030740.93 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:09:18,926 INFO [optim.py:368] (3/8) 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 02:09:34,563 INFO [train.py:904] (3/8) Epoch 25, batch 8950, loss[loss=0.1561, simple_loss=0.2542, pruned_loss=0.02896, over 16303.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2653, pruned_loss=0.03654, over 3037514.57 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:10:11,678 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 02:10:18,846 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4586, 3.0215, 2.7196, 2.2287, 2.2247, 2.3188, 3.0683, 2.8137], device='cuda:3'), covar=tensor([0.2667, 0.0806, 0.1751, 0.2920, 0.2853, 0.2254, 0.0486, 0.1742], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0266, 0.0305, 0.0314, 0.0295, 0.0265, 0.0293, 0.0337], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 02:10:59,641 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252592.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:11:21,899 INFO [train.py:904] (3/8) Epoch 25, batch 9000, loss[loss=0.163, simple_loss=0.2523, pruned_loss=0.0368, over 11959.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2626, pruned_loss=0.03565, over 3036680.43 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:11:21,899 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 02:11:31,594 INFO [train.py:938] (3/8) Epoch 25, validation: loss=0.1442, simple_loss=0.248, pruned_loss=0.02014, over 944034.00 frames. 2023-05-02 02:11:31,595 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 02:11:53,595 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:11:57,926 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8639, 3.8070, 3.9528, 3.7760, 3.9424, 4.3089, 3.9611, 3.6936], device='cuda:3'), covar=tensor([0.1921, 0.2270, 0.2197, 0.2261, 0.2553, 0.1632, 0.1514, 0.2488], device='cuda:3'), in_proj_covar=tensor([0.0405, 0.0593, 0.0657, 0.0489, 0.0646, 0.0684, 0.0510, 0.0653], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 02:13:01,732 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 9050, loss[loss=0.1847, simple_loss=0.271, pruned_loss=0.04919, over 12766.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2634, pruned_loss=0.03605, over 3054568.91 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:13:15,970 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252653.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:47,225 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:56,461 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:14:13,487 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0023, 3.8517, 3.9536, 4.1875, 4.2654, 3.9656, 4.3194, 4.3433], device='cuda:3'), covar=tensor([0.1812, 0.1364, 0.1971, 0.0920, 0.0795, 0.1492, 0.0817, 0.0960], device='cuda:3'), in_proj_covar=tensor([0.0621, 0.0764, 0.0879, 0.0778, 0.0593, 0.0615, 0.0644, 0.0746], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:14:58,382 INFO [train.py:904] (3/8) Epoch 25, batch 9100, loss[loss=0.1676, simple_loss=0.2519, pruned_loss=0.04168, over 12548.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2631, pruned_loss=0.03656, over 3049256.68 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:16:01,335 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:16:41,921 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 25, batch 9150, loss[loss=0.1558, simple_loss=0.253, pruned_loss=0.02927, over 16680.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2625, pruned_loss=0.03606, over 3031058.83 frames. ], batch size: 76, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:17:29,236 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9520, 4.7905, 4.9714, 5.1649, 5.3730, 4.7358, 5.3928, 5.4102], device='cuda:3'), covar=tensor([0.2209, 0.1272, 0.1957, 0.0845, 0.0617, 0.0967, 0.0508, 0.0729], device='cuda:3'), in_proj_covar=tensor([0.0622, 0.0763, 0.0879, 0.0777, 0.0592, 0.0614, 0.0642, 0.0746], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:17:31,590 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-02 02:17:42,773 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,320 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 9200, loss[loss=0.1813, simple_loss=0.2675, pruned_loss=0.04757, over 16664.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2585, pruned_loss=0.03512, over 3069601.12 frames. ], batch size: 62, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:20:05,204 INFO [optim.py:368] (3/8) 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 02:20:19,112 INFO [train.py:904] (3/8) Epoch 25, batch 9250, loss[loss=0.1469, simple_loss=0.2331, pruned_loss=0.03032, over 12644.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2587, pruned_loss=0.03537, over 3067274.54 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 16.0 2023-05-02 02:20:28,204 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 9300, loss[loss=0.1578, simple_loss=0.2426, pruned_loss=0.03652, over 12083.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2568, pruned_loss=0.0346, over 3047718.88 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:22:47,945 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:22:53,204 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2116, 4.3217, 4.1446, 3.8185, 3.8572, 4.2339, 3.8970, 4.0108], device='cuda:3'), covar=tensor([0.0594, 0.0679, 0.0320, 0.0334, 0.0769, 0.0629, 0.0824, 0.0608], device='cuda:3'), in_proj_covar=tensor([0.0291, 0.0435, 0.0339, 0.0342, 0.0340, 0.0393, 0.0234, 0.0406], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:23:14,953 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4302, 5.3736, 5.1048, 4.6418, 5.2571, 1.9782, 4.9513, 4.9355], device='cuda:3'), covar=tensor([0.0086, 0.0087, 0.0232, 0.0355, 0.0102, 0.2738, 0.0143, 0.0235], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0162, 0.0200, 0.0175, 0.0177, 0.0208, 0.0188, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:23:45,074 INFO [optim.py:368] (3/8) 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 02:23:45,938 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4179, 4.6984, 4.5281, 4.5169, 4.2272, 4.2274, 4.1955, 4.7324], device='cuda:3'), covar=tensor([0.1146, 0.0866, 0.0892, 0.0790, 0.0733, 0.1295, 0.1059, 0.0823], device='cuda:3'), in_proj_covar=tensor([0.0675, 0.0815, 0.0672, 0.0631, 0.0518, 0.0525, 0.0682, 0.0637], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:23:47,879 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252948.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:23:55,323 INFO [train.py:904] (3/8) Epoch 25, batch 9350, loss[loss=0.1758, simple_loss=0.2673, pruned_loss=0.04219, over 16163.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2565, pruned_loss=0.03451, over 3039584.38 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:24:29,917 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:25:16,704 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 9400, loss[loss=0.1812, simple_loss=0.2848, pruned_loss=0.03876, over 16391.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2568, pruned_loss=0.03445, over 3043835.02 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:26:19,759 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:05,785 INFO [optim.py:368] (3/8) 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 02:27:17,026 INFO [train.py:904] (3/8) Epoch 25, batch 9450, loss[loss=0.171, simple_loss=0.2626, pruned_loss=0.03967, over 16906.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2587, pruned_loss=0.03509, over 3024894.19 frames. ], batch size: 109, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:27:19,812 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:57,727 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:31,189 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:54,674 INFO [train.py:904] (3/8) Epoch 25, batch 9500, loss[loss=0.1403, simple_loss=0.239, pruned_loss=0.02083, over 16886.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2578, pruned_loss=0.03439, over 3059402.79 frames. ], batch size: 96, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:29:33,158 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:26,751 INFO [optim.py:368] (3/8) 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 02:30:40,904 INFO [train.py:904] (3/8) Epoch 25, batch 9550, loss[loss=0.1467, simple_loss=0.2453, pruned_loss=0.02401, over 16739.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2575, pruned_loss=0.03447, over 3050522.50 frames. ], batch size: 76, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:22,177 INFO [train.py:904] (3/8) Epoch 25, batch 9600, loss[loss=0.1825, simple_loss=0.2793, pruned_loss=0.04288, over 16867.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2587, pruned_loss=0.03524, over 3031359.83 frames. ], batch size: 116, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:42,619 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253213.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:33:21,138 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9104, 4.1923, 4.0377, 4.0522, 3.6849, 3.8241, 3.8253, 4.1957], device='cuda:3'), covar=tensor([0.1240, 0.0939, 0.1124, 0.0924, 0.0844, 0.1635, 0.1071, 0.1057], device='cuda:3'), in_proj_covar=tensor([0.0675, 0.0814, 0.0671, 0.0630, 0.0517, 0.0524, 0.0681, 0.0638], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:33:55,685 INFO [optim.py:368] (3/8) 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 02:33:59,551 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 9650, loss[loss=0.1461, simple_loss=0.2447, pruned_loss=0.02377, over 16496.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2609, pruned_loss=0.03562, over 3042685.05 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:34:52,568 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:35:46,089 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 9700, loss[loss=0.1633, simple_loss=0.2614, pruned_loss=0.03253, over 16354.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2601, pruned_loss=0.03521, over 3062120.31 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:36:27,591 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:42,577 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:56,111 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 02:37:37,140 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253349.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:44,444 INFO [train.py:904] (3/8) Epoch 25, batch 9750, loss[loss=0.1483, simple_loss=0.2439, pruned_loss=0.02629, over 16454.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2589, pruned_loss=0.03537, over 3064698.43 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:37:51,444 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6060, 3.6678, 3.4742, 3.1322, 3.2997, 3.5556, 3.3730, 3.4069], device='cuda:3'), covar=tensor([0.0581, 0.0688, 0.0295, 0.0247, 0.0502, 0.0427, 0.1292, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0288, 0.0431, 0.0338, 0.0339, 0.0339, 0.0389, 0.0233, 0.0402], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:38:16,704 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 02:38:21,278 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:38:38,926 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253380.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:00,933 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253390.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:23,675 INFO [train.py:904] (3/8) Epoch 25, batch 9800, loss[loss=0.1515, simple_loss=0.2593, pruned_loss=0.02179, over 16539.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2594, pruned_loss=0.03451, over 3078685.47 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:39:53,067 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:38,521 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 1.999e+02 2.298e+02 2.745e+02 1.319e+03, threshold=4.597e+02, percent-clipped=1.0 2023-05-02 02:41:06,296 INFO [train.py:904] (3/8) Epoch 25, batch 9850, loss[loss=0.1622, simple_loss=0.2605, pruned_loss=0.03189, over 15303.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2597, pruned_loss=0.03392, over 3079986.92 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:42:03,967 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 25, batch 9900, loss[loss=0.1702, simple_loss=0.2725, pruned_loss=0.03398, over 16704.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2608, pruned_loss=0.03407, over 3068465.20 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:43:25,506 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253513.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:31,405 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 02:43:41,889 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:44:46,254 INFO [optim.py:368] (3/8) 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 02:44:47,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4841, 4.3023, 4.5290, 4.6828, 4.8453, 4.3491, 4.8495, 4.8665], device='cuda:3'), covar=tensor([0.1923, 0.1197, 0.1673, 0.0786, 0.0609, 0.1069, 0.0567, 0.0635], device='cuda:3'), in_proj_covar=tensor([0.0623, 0.0762, 0.0876, 0.0776, 0.0596, 0.0615, 0.0644, 0.0747], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:44:47,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7775, 3.5363, 4.0230, 1.9572, 4.0899, 4.2283, 3.1564, 3.1335], device='cuda:3'), covar=tensor([0.0702, 0.0253, 0.0133, 0.1255, 0.0068, 0.0118, 0.0391, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0106, 0.0093, 0.0134, 0.0080, 0.0122, 0.0124, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 02:44:58,925 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6742, 2.7201, 2.4459, 4.3616, 2.7364, 4.0587, 1.5794, 2.9488], device='cuda:3'), covar=tensor([0.1367, 0.0766, 0.1205, 0.0145, 0.0128, 0.0338, 0.1585, 0.0720], device='cuda:3'), in_proj_covar=tensor([0.0169, 0.0175, 0.0194, 0.0191, 0.0197, 0.0212, 0.0204, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 02:44:59,712 INFO [train.py:904] (3/8) Epoch 25, batch 9950, loss[loss=0.1789, simple_loss=0.2743, pruned_loss=0.04178, over 16986.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2629, pruned_loss=0.03442, over 3074220.12 frames. ], batch size: 109, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:45:18,703 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:14,342 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:30,367 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1206, 2.0622, 2.5917, 3.2445, 2.8351, 3.6182, 2.3240, 3.5919], device='cuda:3'), covar=tensor([0.0248, 0.0596, 0.0422, 0.0258, 0.0382, 0.0150, 0.0531, 0.0157], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0192, 0.0180, 0.0181, 0.0198, 0.0156, 0.0195, 0.0155], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:47:02,502 INFO [train.py:904] (3/8) Epoch 25, batch 10000, loss[loss=0.1415, simple_loss=0.2412, pruned_loss=0.02093, over 17134.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2613, pruned_loss=0.03374, over 3098854.20 frames. ], batch size: 49, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:47:52,996 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253628.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:03,393 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 02:48:36,792 INFO [optim.py:368] (3/8) 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 02:48:39,828 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:41,200 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253650.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:46,077 INFO [train.py:904] (3/8) Epoch 25, batch 10050, loss[loss=0.1681, simple_loss=0.2682, pruned_loss=0.03402, over 16362.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2617, pruned_loss=0.0337, over 3111503.38 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:49:48,434 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 25, batch 10100, loss[loss=0.1568, simple_loss=0.2518, pruned_loss=0.03085, over 16276.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2614, pruned_loss=0.03379, over 3097995.38 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:50:35,648 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 02:51:36,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5822, 3.5703, 3.5189, 2.8951, 3.4353, 1.9927, 3.2535, 2.9038], device='cuda:3'), covar=tensor([0.0140, 0.0146, 0.0194, 0.0215, 0.0113, 0.2522, 0.0129, 0.0268], device='cuda:3'), in_proj_covar=tensor([0.0167, 0.0160, 0.0196, 0.0172, 0.0174, 0.0205, 0.0186, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 02:52:07,292 INFO [train.py:904] (3/8) Epoch 26, batch 0, loss[loss=0.161, simple_loss=0.2487, pruned_loss=0.03662, over 16842.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2487, pruned_loss=0.03662, over 16842.00 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 8.0 2023-05-02 02:52:07,292 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 02:52:14,674 INFO [train.py:938] (3/8) Epoch 26, validation: loss=0.1437, simple_loss=0.2472, pruned_loss=0.02009, over 944034.00 frames. 2023-05-02 02:52:14,675 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 02:52:44,911 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253775.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:53:23,654 INFO [train.py:904] (3/8) Epoch 26, batch 50, loss[loss=0.1944, simple_loss=0.275, pruned_loss=0.05685, over 16447.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.266, pruned_loss=0.04609, over 750557.05 frames. ], batch size: 146, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:54:28,067 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 100, loss[loss=0.1888, simple_loss=0.2719, pruned_loss=0.05286, over 16272.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2631, pruned_loss=0.04437, over 1312722.74 frames. ], batch size: 165, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:55:03,716 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:55:39,325 INFO [train.py:904] (3/8) Epoch 26, batch 150, loss[loss=0.1679, simple_loss=0.2702, pruned_loss=0.03286, over 17115.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2609, pruned_loss=0.04361, over 1761272.04 frames. ], batch size: 49, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:08,419 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:56:19,640 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 02:56:46,628 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.138e+02 2.556e+02 3.055e+02 5.694e+02, threshold=5.112e+02, percent-clipped=1.0 2023-05-02 02:56:49,068 INFO [train.py:904] (3/8) Epoch 26, batch 200, loss[loss=0.1593, simple_loss=0.2555, pruned_loss=0.03157, over 16731.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2601, pruned_loss=0.04346, over 2109887.56 frames. ], batch size: 57, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:57:33,640 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253985.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:03,918 INFO [train.py:904] (3/8) Epoch 26, batch 250, loss[loss=0.1676, simple_loss=0.2524, pruned_loss=0.04139, over 16699.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2595, pruned_loss=0.04299, over 2387292.55 frames. ], batch size: 83, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:58:07,776 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:47,173 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:50,822 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:10,750 INFO [optim.py:368] (3/8) 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 02:59:14,068 INFO [train.py:904] (3/8) Epoch 26, batch 300, loss[loss=0.2001, simple_loss=0.2657, pruned_loss=0.06725, over 15896.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2573, pruned_loss=0.04208, over 2601315.70 frames. ], batch size: 35, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:59:45,412 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254075.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:57,061 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:00:23,588 INFO [train.py:904] (3/8) Epoch 26, batch 350, loss[loss=0.1369, simple_loss=0.2217, pruned_loss=0.02605, over 16776.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.255, pruned_loss=0.04071, over 2762648.15 frames. ], batch size: 39, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 03:00:51,987 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:00:52,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7163, 3.7611, 3.9126, 2.8643, 3.6519, 4.0724, 3.7169, 2.4875], device='cuda:3'), covar=tensor([0.0536, 0.0340, 0.0066, 0.0395, 0.0123, 0.0115, 0.0118, 0.0471], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0101, 0.0111, 0.0096, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 03:01:30,800 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 400, loss[loss=0.2025, simple_loss=0.2728, pruned_loss=0.06608, over 16793.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2534, pruned_loss=0.04115, over 2882260.89 frames. ], batch size: 124, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:02:07,819 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:02:14,803 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1927, 5.0836, 5.0360, 4.5258, 4.6901, 5.1017, 4.9993, 4.7287], device='cuda:3'), covar=tensor([0.0568, 0.0483, 0.0345, 0.0393, 0.1147, 0.0379, 0.0373, 0.0759], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0446, 0.0349, 0.0352, 0.0350, 0.0403, 0.0239, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:02:44,706 INFO [train.py:904] (3/8) Epoch 26, batch 450, loss[loss=0.1829, simple_loss=0.2676, pruned_loss=0.04913, over 16657.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2526, pruned_loss=0.04052, over 2981274.03 frames. ], batch size: 57, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:03:12,894 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:15,095 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 26, batch 500, loss[loss=0.1654, simple_loss=0.25, pruned_loss=0.04036, over 16720.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2512, pruned_loss=0.03974, over 3057679.38 frames. ], batch size: 89, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:04:18,987 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254271.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:04:57,431 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6949, 3.8219, 2.6039, 4.3889, 3.0711, 4.3499, 2.8126, 3.3046], device='cuda:3'), covar=tensor([0.0349, 0.0462, 0.1499, 0.0356, 0.0810, 0.0577, 0.1333, 0.0712], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0179, 0.0197, 0.0169, 0.0178, 0.0217, 0.0204, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:05:01,741 INFO [train.py:904] (3/8) Epoch 26, batch 550, loss[loss=0.1715, simple_loss=0.2467, pruned_loss=0.04817, over 16838.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.249, pruned_loss=0.03856, over 3116738.34 frames. ], batch size: 96, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:05:05,514 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:47,903 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 03:06:08,396 INFO [optim.py:368] (3/8) 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 03:06:11,648 INFO [train.py:904] (3/8) Epoch 26, batch 600, loss[loss=0.1454, simple_loss=0.2367, pruned_loss=0.02707, over 17198.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2486, pruned_loss=0.03899, over 3150803.31 frames. ], batch size: 46, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:06:13,034 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 650, loss[loss=0.1446, simple_loss=0.2361, pruned_loss=0.02659, over 16873.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2472, pruned_loss=0.03824, over 3195028.81 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:07:22,521 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 03:07:34,447 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-05-02 03:08:28,766 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 700, loss[loss=0.172, simple_loss=0.246, pruned_loss=0.04896, over 16853.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2484, pruned_loss=0.03871, over 3222299.80 frames. ], batch size: 116, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:08:34,520 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6720, 2.6091, 2.3054, 2.5066, 2.9749, 2.6204, 3.2235, 3.1023], device='cuda:3'), covar=tensor([0.0219, 0.0552, 0.0638, 0.0572, 0.0381, 0.0531, 0.0346, 0.0380], device='cuda:3'), in_proj_covar=tensor([0.0226, 0.0246, 0.0234, 0.0234, 0.0245, 0.0244, 0.0241, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:09:41,479 INFO [train.py:904] (3/8) Epoch 26, batch 750, loss[loss=0.1521, simple_loss=0.2386, pruned_loss=0.03285, over 16857.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2484, pruned_loss=0.03888, over 3235926.37 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:10:27,468 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 26, batch 800, loss[loss=0.1656, simple_loss=0.2434, pruned_loss=0.04386, over 16923.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2485, pruned_loss=0.03852, over 3265144.09 frames. ], batch size: 109, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:10:50,934 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4141, 4.4243, 4.5650, 4.3676, 4.5017, 5.0285, 4.5529, 4.2169], device='cuda:3'), covar=tensor([0.1789, 0.2432, 0.2940, 0.2567, 0.2838, 0.1276, 0.1823, 0.2791], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0622, 0.0691, 0.0512, 0.0677, 0.0713, 0.0535, 0.0681], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 03:11:20,489 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254574.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:51,482 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 850, loss[loss=0.1575, simple_loss=0.2368, pruned_loss=0.03909, over 16834.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2482, pruned_loss=0.03807, over 3275741.33 frames. ], batch size: 116, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:12:10,645 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2028, 3.2668, 3.3733, 2.3206, 3.1339, 3.4572, 3.1908, 2.0492], device='cuda:3'), covar=tensor([0.0567, 0.0121, 0.0073, 0.0445, 0.0135, 0.0117, 0.0122, 0.0529], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0087, 0.0088, 0.0135, 0.0101, 0.0112, 0.0097, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 03:12:45,175 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 2.023e+02 2.316e+02 2.771e+02 4.972e+02, threshold=4.632e+02, percent-clipped=1.0 2023-05-02 03:13:10,208 INFO [train.py:904] (3/8) Epoch 26, batch 900, loss[loss=0.1709, simple_loss=0.259, pruned_loss=0.04134, over 16443.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2475, pruned_loss=0.03744, over 3284157.47 frames. ], batch size: 68, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:13:41,828 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0456, 5.1136, 5.5108, 5.4921, 5.4992, 5.1468, 5.0938, 4.9097], device='cuda:3'), covar=tensor([0.0354, 0.0486, 0.0405, 0.0404, 0.0437, 0.0413, 0.0903, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0481, 0.0469, 0.0429, 0.0515, 0.0493, 0.0568, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 03:14:19,298 INFO [train.py:904] (3/8) Epoch 26, batch 950, loss[loss=0.1552, simple_loss=0.2472, pruned_loss=0.03162, over 17021.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.248, pruned_loss=0.03781, over 3291933.53 frames. ], batch size: 55, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:14:23,648 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5529, 3.5192, 2.7607, 2.1486, 2.2662, 2.2750, 3.6234, 3.0231], device='cuda:3'), covar=tensor([0.2903, 0.0665, 0.1840, 0.3155, 0.2797, 0.2298, 0.0568, 0.1711], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0272, 0.0311, 0.0320, 0.0301, 0.0271, 0.0301, 0.0346], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 03:15:07,727 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8632, 2.5736, 2.4420, 4.0082, 3.1050, 3.9343, 1.8100, 2.8095], device='cuda:3'), covar=tensor([0.1441, 0.0841, 0.1385, 0.0249, 0.0184, 0.0521, 0.1595, 0.0980], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0198, 0.0203, 0.0218, 0.0208, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:15:14,708 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8794, 3.4924, 3.9219, 2.2208, 3.9756, 4.0225, 3.2326, 3.0968], device='cuda:3'), covar=tensor([0.0724, 0.0285, 0.0200, 0.1113, 0.0123, 0.0231, 0.0395, 0.0473], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0140, 0.0085, 0.0130, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:15:24,090 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 1000, loss[loss=0.1452, simple_loss=0.2399, pruned_loss=0.02521, over 17114.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2464, pruned_loss=0.03829, over 3302594.79 frames. ], batch size: 47, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:15:31,179 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:28,367 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:35,899 INFO [train.py:904] (3/8) Epoch 26, batch 1050, loss[loss=0.1624, simple_loss=0.2386, pruned_loss=0.04313, over 16274.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2458, pruned_loss=0.03775, over 3310849.38 frames. ], batch size: 165, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:16:55,110 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:42,920 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:43,594 INFO [optim.py:368] (3/8) 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 03:17:44,786 INFO [train.py:904] (3/8) Epoch 26, batch 1100, loss[loss=0.154, simple_loss=0.2517, pruned_loss=0.02816, over 17146.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2453, pruned_loss=0.03756, over 3315754.84 frames. ], batch size: 48, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:17:53,544 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:22,598 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 03:18:36,860 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 1150, loss[loss=0.1645, simple_loss=0.2433, pruned_loss=0.04281, over 16805.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2452, pruned_loss=0.03709, over 3320838.96 frames. ], batch size: 83, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:18:53,534 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0353, 3.1381, 2.9378, 5.1895, 4.2887, 4.4333, 1.8866, 3.4357], device='cuda:3'), covar=tensor([0.1301, 0.0741, 0.1164, 0.0226, 0.0219, 0.0413, 0.1603, 0.0706], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0180, 0.0198, 0.0198, 0.0203, 0.0219, 0.0209, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:19:04,319 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254912.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:29,276 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:59,610 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 1200, loss[loss=0.1681, simple_loss=0.2608, pruned_loss=0.03766, over 16735.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2457, pruned_loss=0.03702, over 3322258.07 frames. ], batch size: 57, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:10,674 INFO [train.py:904] (3/8) Epoch 26, batch 1250, loss[loss=0.1593, simple_loss=0.2358, pruned_loss=0.04138, over 16340.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2459, pruned_loss=0.03726, over 3309917.64 frames. ], batch size: 165, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:11,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1595, 2.6869, 2.1613, 2.4291, 3.0428, 2.7715, 3.0881, 3.0998], device='cuda:3'), covar=tensor([0.0236, 0.0421, 0.0545, 0.0467, 0.0260, 0.0346, 0.0262, 0.0314], device='cuda:3'), in_proj_covar=tensor([0.0228, 0.0247, 0.0234, 0.0235, 0.0246, 0.0245, 0.0244, 0.0243], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:21:45,796 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4803, 3.5919, 3.7107, 1.8486, 2.9836, 2.0705, 3.8263, 3.8128], device='cuda:3'), covar=tensor([0.0244, 0.0928, 0.0614, 0.2494, 0.1022, 0.1255, 0.0618, 0.0942], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0168, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 03:21:50,242 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 03:22:19,973 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.021e+02 2.372e+02 2.873e+02 4.328e+02, threshold=4.744e+02, percent-clipped=0.0 2023-05-02 03:22:21,123 INFO [train.py:904] (3/8) Epoch 26, batch 1300, loss[loss=0.1738, simple_loss=0.2526, pruned_loss=0.04754, over 16462.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2452, pruned_loss=0.03726, over 3312832.50 frames. ], batch size: 146, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:22:41,379 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 03:23:04,620 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0290, 4.7737, 5.0639, 5.2601, 5.4672, 4.7722, 5.4221, 5.4521], device='cuda:3'), covar=tensor([0.1970, 0.1342, 0.1749, 0.0723, 0.0542, 0.0905, 0.0530, 0.0621], device='cuda:3'), in_proj_covar=tensor([0.0674, 0.0824, 0.0954, 0.0842, 0.0638, 0.0662, 0.0695, 0.0807], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:23:30,835 INFO [train.py:904] (3/8) Epoch 26, batch 1350, loss[loss=0.1572, simple_loss=0.2392, pruned_loss=0.0376, over 16814.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2449, pruned_loss=0.03705, over 3307926.26 frames. ], batch size: 102, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:42,882 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:24:38,899 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 1400, loss[loss=0.1612, simple_loss=0.2386, pruned_loss=0.04189, over 16705.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2454, pruned_loss=0.03732, over 3300928.56 frames. ], batch size: 134, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:24:41,766 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:23,411 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9070, 1.9541, 2.5942, 2.9230, 2.7376, 3.4685, 2.4236, 3.4254], device='cuda:3'), covar=tensor([0.0310, 0.0653, 0.0405, 0.0384, 0.0430, 0.0226, 0.0516, 0.0203], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0190, 0.0206, 0.0164, 0.0202, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:25:34,092 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 1450, loss[loss=0.149, simple_loss=0.2442, pruned_loss=0.02689, over 17040.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2447, pruned_loss=0.03726, over 3303369.94 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:25:54,147 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:25,735 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0529, 2.1244, 2.7430, 3.0298, 2.9010, 3.5186, 2.4769, 3.5392], device='cuda:3'), covar=tensor([0.0301, 0.0601, 0.0394, 0.0375, 0.0413, 0.0242, 0.0558, 0.0201], device='cuda:3'), in_proj_covar=tensor([0.0197, 0.0200, 0.0188, 0.0190, 0.0206, 0.0164, 0.0202, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:26:26,883 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255230.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:40,748 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:56,773 INFO [optim.py:368] (3/8) 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 03:26:57,916 INFO [train.py:904] (3/8) Epoch 26, batch 1500, loss[loss=0.1517, simple_loss=0.245, pruned_loss=0.02924, over 16864.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2451, pruned_loss=0.03684, over 3306548.04 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:27:19,655 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6584, 3.9042, 2.4818, 4.4787, 2.9833, 4.4091, 2.3973, 3.1416], device='cuda:3'), covar=tensor([0.0358, 0.0438, 0.1701, 0.0327, 0.0917, 0.0575, 0.1648, 0.0852], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0183, 0.0201, 0.0175, 0.0182, 0.0223, 0.0208, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:27:31,537 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:28:04,972 INFO [train.py:904] (3/8) Epoch 26, batch 1550, loss[loss=0.1674, simple_loss=0.2557, pruned_loss=0.03953, over 17068.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2458, pruned_loss=0.03778, over 3309172.31 frames. ], batch size: 53, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:29:12,930 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.260e+02 2.709e+02 3.380e+02 1.198e+03, threshold=5.419e+02, percent-clipped=4.0 2023-05-02 03:29:14,102 INFO [train.py:904] (3/8) Epoch 26, batch 1600, loss[loss=0.1825, simple_loss=0.2548, pruned_loss=0.05515, over 16754.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2468, pruned_loss=0.03817, over 3312515.59 frames. ], batch size: 134, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:29:51,058 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0796, 4.2349, 2.9833, 4.7793, 3.4483, 4.7856, 3.0718, 3.5935], device='cuda:3'), covar=tensor([0.0298, 0.0374, 0.1405, 0.0312, 0.0730, 0.0438, 0.1322, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0183, 0.0200, 0.0174, 0.0181, 0.0222, 0.0207, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:30:03,619 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:30:20,469 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3578, 3.4942, 3.9936, 2.2155, 3.2443, 2.3074, 3.9124, 3.7702], device='cuda:3'), covar=tensor([0.0262, 0.0938, 0.0455, 0.2072, 0.0804, 0.1060, 0.0525, 0.1009], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0169, 0.0171, 0.0157, 0.0148, 0.0132, 0.0146, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 03:30:23,420 INFO [train.py:904] (3/8) Epoch 26, batch 1650, loss[loss=0.1785, simple_loss=0.2645, pruned_loss=0.04623, over 16635.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2479, pruned_loss=0.03916, over 3312923.29 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:30:35,814 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:30:36,325 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 03:31:16,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9189, 3.5230, 3.9944, 2.2406, 4.1354, 4.1889, 3.2434, 3.2170], device='cuda:3'), covar=tensor([0.0714, 0.0300, 0.0222, 0.1110, 0.0100, 0.0221, 0.0443, 0.0452], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0113, 0.0100, 0.0141, 0.0086, 0.0131, 0.0131, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 03:31:28,976 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=255449.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:32,745 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 1700, loss[loss=0.1377, simple_loss=0.2258, pruned_loss=0.02473, over 16762.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2508, pruned_loss=0.04001, over 3294468.28 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:31:34,384 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255459.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:50,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-05-02 03:32:40,142 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 1750, loss[loss=0.1533, simple_loss=0.2386, pruned_loss=0.03396, over 16861.00 frames. ], tot_loss[loss=0.166, simple_loss=0.252, pruned_loss=0.04002, over 3299716.26 frames. ], batch size: 96, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:32:47,996 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255507.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:33:49,124 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.153e+02 2.652e+02 3.222e+02 7.615e+02, threshold=5.303e+02, percent-clipped=5.0 2023-05-02 03:33:51,389 INFO [train.py:904] (3/8) Epoch 26, batch 1800, loss[loss=0.1567, simple_loss=0.2498, pruned_loss=0.03181, over 17130.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.253, pruned_loss=0.03967, over 3304203.54 frames. ], batch size: 47, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:33:54,879 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:34:49,176 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 03:34:59,034 INFO [train.py:904] (3/8) Epoch 26, batch 1850, loss[loss=0.1805, simple_loss=0.2558, pruned_loss=0.05264, over 16690.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2542, pruned_loss=0.04003, over 3302072.95 frames. ], batch size: 124, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:05,021 INFO [optim.py:368] (3/8) 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 03:36:06,160 INFO [train.py:904] (3/8) Epoch 26, batch 1900, loss[loss=0.1522, simple_loss=0.2378, pruned_loss=0.03327, over 16795.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2533, pruned_loss=0.03941, over 3298144.63 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:32,135 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6402, 2.7494, 2.2832, 2.5382, 3.0036, 2.7325, 3.1555, 3.1627], device='cuda:3'), covar=tensor([0.0189, 0.0445, 0.0581, 0.0482, 0.0335, 0.0423, 0.0335, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0249, 0.0237, 0.0237, 0.0248, 0.0248, 0.0246, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:36:32,599 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 03:36:44,049 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5279, 3.5433, 3.7695, 2.7993, 3.4820, 3.8908, 3.5775, 2.3321], device='cuda:3'), covar=tensor([0.0522, 0.0232, 0.0068, 0.0391, 0.0139, 0.0104, 0.0107, 0.0497], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0136, 0.0102, 0.0112, 0.0097, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 03:37:16,451 INFO [train.py:904] (3/8) Epoch 26, batch 1950, loss[loss=0.1639, simple_loss=0.2623, pruned_loss=0.0327, over 16541.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2536, pruned_loss=0.03876, over 3303070.57 frames. ], batch size: 68, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:55,253 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8087, 1.9903, 2.6513, 2.8514, 2.7188, 3.3166, 2.3876, 3.4201], device='cuda:3'), covar=tensor([0.0340, 0.0621, 0.0380, 0.0392, 0.0423, 0.0247, 0.0545, 0.0206], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0199, 0.0187, 0.0190, 0.0206, 0.0164, 0.0201, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:38:08,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8287, 4.2889, 4.2680, 3.0965, 3.5934, 4.2507, 3.8452, 2.4596], device='cuda:3'), covar=tensor([0.0508, 0.0076, 0.0059, 0.0399, 0.0173, 0.0103, 0.0102, 0.0513], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0136, 0.0102, 0.0112, 0.0097, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 03:38:12,955 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:38:24,265 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.052e+02 2.408e+02 2.894e+02 5.578e+02, threshold=4.816e+02, percent-clipped=3.0 2023-05-02 03:38:25,464 INFO [train.py:904] (3/8) Epoch 26, batch 2000, loss[loss=0.1863, simple_loss=0.2755, pruned_loss=0.04857, over 15379.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2533, pruned_loss=0.03851, over 3307290.06 frames. ], batch size: 190, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:39:35,298 INFO [train.py:904] (3/8) Epoch 26, batch 2050, loss[loss=0.1764, simple_loss=0.2659, pruned_loss=0.04349, over 15552.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2524, pruned_loss=0.03839, over 3311360.66 frames. ], batch size: 191, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:40:10,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2512, 2.4154, 2.4176, 4.0573, 2.3233, 2.7601, 2.4642, 2.5285], device='cuda:3'), covar=tensor([0.1542, 0.3652, 0.3171, 0.0662, 0.4040, 0.2532, 0.3622, 0.3513], device='cuda:3'), in_proj_covar=tensor([0.0419, 0.0469, 0.0385, 0.0338, 0.0445, 0.0536, 0.0440, 0.0548], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:40:44,482 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.146e+02 2.478e+02 2.962e+02 5.196e+02, threshold=4.956e+02, percent-clipped=1.0 2023-05-02 03:40:45,698 INFO [train.py:904] (3/8) Epoch 26, batch 2100, loss[loss=0.1427, simple_loss=0.2343, pruned_loss=0.02556, over 17199.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2536, pruned_loss=0.03901, over 3308509.36 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:40:51,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6499, 2.6606, 2.3060, 2.5632, 2.9670, 2.6586, 3.1727, 3.1632], device='cuda:3'), covar=tensor([0.0182, 0.0493, 0.0601, 0.0523, 0.0339, 0.0494, 0.0312, 0.0322], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0249, 0.0237, 0.0237, 0.0249, 0.0249, 0.0247, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:41:54,461 INFO [train.py:904] (3/8) Epoch 26, batch 2150, loss[loss=0.1566, simple_loss=0.2465, pruned_loss=0.03338, over 17174.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2544, pruned_loss=0.03979, over 3312858.02 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:43:04,641 INFO [optim.py:368] (3/8) 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 03:43:05,582 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-02 03:43:05,921 INFO [train.py:904] (3/8) Epoch 26, batch 2200, loss[loss=0.16, simple_loss=0.2576, pruned_loss=0.03121, over 17127.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2545, pruned_loss=0.03959, over 3315444.08 frames. ], batch size: 47, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:43:07,521 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4738, 5.4269, 5.3077, 4.8156, 4.9660, 5.3687, 5.2620, 4.9240], device='cuda:3'), covar=tensor([0.0584, 0.0545, 0.0297, 0.0358, 0.1067, 0.0446, 0.0333, 0.0818], device='cuda:3'), in_proj_covar=tensor([0.0319, 0.0474, 0.0369, 0.0374, 0.0372, 0.0427, 0.0254, 0.0446], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 03:43:10,453 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255956.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:44:19,346 INFO [train.py:904] (3/8) Epoch 26, batch 2250, loss[loss=0.1592, simple_loss=0.2463, pruned_loss=0.03602, over 15861.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2545, pruned_loss=0.03926, over 3325695.58 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:44:40,071 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:09,552 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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 03:45:29,018 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-02 03:45:29,154 INFO [train.py:904] (3/8) Epoch 26, batch 2300, loss[loss=0.1569, simple_loss=0.2485, pruned_loss=0.03261, over 16840.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2538, pruned_loss=0.03905, over 3326738.25 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:46:09,843 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:14,177 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:24,421 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256092.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:34,501 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 2350, loss[loss=0.1797, simple_loss=0.2832, pruned_loss=0.03806, over 17023.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2552, pruned_loss=0.04008, over 3329053.49 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:19,389 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:24,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4947, 4.8050, 5.1420, 5.0980, 5.1119, 4.7808, 4.4581, 4.5548], device='cuda:3'), covar=tensor([0.0645, 0.0753, 0.0610, 0.0653, 0.0916, 0.0668, 0.1890, 0.0693], device='cuda:3'), in_proj_covar=tensor([0.0433, 0.0489, 0.0472, 0.0432, 0.0520, 0.0498, 0.0574, 0.0398], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 03:47:25,938 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 03:47:32,069 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:36,950 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:44,531 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8549, 4.5435, 4.5121, 5.0204, 5.2496, 4.6686, 5.1835, 5.2325], device='cuda:3'), covar=tensor([0.1899, 0.1561, 0.2748, 0.1110, 0.0869, 0.1307, 0.1096, 0.0955], device='cuda:3'), in_proj_covar=tensor([0.0690, 0.0847, 0.0978, 0.0859, 0.0651, 0.0678, 0.0711, 0.0825], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:47:45,350 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.073e+02 2.391e+02 2.953e+02 7.650e+02, threshold=4.783e+02, percent-clipped=2.0 2023-05-02 03:47:46,518 INFO [train.py:904] (3/8) Epoch 26, batch 2400, loss[loss=0.1883, simple_loss=0.2783, pruned_loss=0.04913, over 17049.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.256, pruned_loss=0.04074, over 3331106.52 frames. ], batch size: 55, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:48:26,039 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8612, 3.8049, 3.9459, 4.0244, 4.0880, 3.7023, 3.9676, 4.1185], device='cuda:3'), covar=tensor([0.1572, 0.1052, 0.1128, 0.0646, 0.0597, 0.1843, 0.1829, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0688, 0.0845, 0.0976, 0.0858, 0.0650, 0.0677, 0.0710, 0.0823], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:48:42,724 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256193.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 03:48:55,676 INFO [train.py:904] (3/8) Epoch 26, batch 2450, loss[loss=0.2016, simple_loss=0.2728, pruned_loss=0.06523, over 16360.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2576, pruned_loss=0.04107, over 3329939.34 frames. ], batch size: 145, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:01,725 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 2500, loss[loss=0.1765, simple_loss=0.2757, pruned_loss=0.03869, over 16686.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2579, pruned_loss=0.04091, over 3328630.98 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:05,749 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 03:50:48,276 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 2550, loss[loss=0.1666, simple_loss=0.2545, pruned_loss=0.03931, over 16753.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2579, pruned_loss=0.04062, over 3318707.26 frames. ], batch size: 83, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:51:26,210 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256312.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:14,139 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.118e+02 2.541e+02 2.890e+02 5.508e+02, threshold=5.081e+02, percent-clipped=1.0 2023-05-02 03:52:23,667 INFO [train.py:904] (3/8) Epoch 26, batch 2600, loss[loss=0.1558, simple_loss=0.2416, pruned_loss=0.03499, over 16838.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2578, pruned_loss=0.04066, over 3324001.88 frames. ], batch size: 96, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:52:51,299 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:10,147 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 03:53:21,459 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 2650, loss[loss=0.159, simple_loss=0.2533, pruned_loss=0.03235, over 17127.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2588, pruned_loss=0.0402, over 3326348.45 frames. ], batch size: 48, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:54:14,471 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256433.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:18,544 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7584, 6.0975, 5.8497, 5.9303, 5.4733, 5.5096, 5.4644, 6.2593], device='cuda:3'), covar=tensor([0.1341, 0.0891, 0.1053, 0.0872, 0.0928, 0.0655, 0.1366, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0713, 0.0867, 0.0707, 0.0668, 0.0550, 0.0549, 0.0727, 0.0675], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 03:54:20,783 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 26, batch 2700, loss[loss=0.1937, simple_loss=0.2829, pruned_loss=0.05227, over 15585.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2586, pruned_loss=0.03948, over 3329307.95 frames. ], batch size: 191, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:55:06,449 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:55:28,867 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256488.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 03:55:48,614 INFO [train.py:904] (3/8) Epoch 26, batch 2750, loss[loss=0.1956, simple_loss=0.2673, pruned_loss=0.06188, over 16855.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2587, pruned_loss=0.03917, over 3329189.84 frames. ], batch size: 109, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:55:55,149 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 03:56:25,294 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-05-02 03:56:29,971 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256532.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:56:36,988 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 03:56:56,716 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 2.155e+02 2.522e+02 3.116e+02 8.730e+02, threshold=5.045e+02, percent-clipped=2.0 2023-05-02 03:56:58,713 INFO [train.py:904] (3/8) Epoch 26, batch 2800, loss[loss=0.1538, simple_loss=0.2522, pruned_loss=0.02771, over 17110.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2589, pruned_loss=0.03873, over 3330449.34 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:58:07,641 INFO [train.py:904] (3/8) Epoch 26, batch 2850, loss[loss=0.1606, simple_loss=0.263, pruned_loss=0.02906, over 17271.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2579, pruned_loss=0.03848, over 3335567.02 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:58:21,216 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256641.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:15,039 INFO [optim.py:368] (3/8) 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 03:59:16,864 INFO [train.py:904] (3/8) Epoch 26, batch 2900, loss[loss=0.1389, simple_loss=0.2291, pruned_loss=0.02438, over 16836.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2574, pruned_loss=0.03913, over 3325721.12 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:59:27,182 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256660.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:45,850 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6099, 4.7629, 4.8941, 4.7217, 4.7088, 5.3205, 4.8364, 4.5361], device='cuda:3'), covar=tensor([0.1468, 0.2122, 0.2663, 0.2148, 0.2839, 0.1113, 0.1591, 0.2645], device='cuda:3'), in_proj_covar=tensor([0.0436, 0.0643, 0.0711, 0.0526, 0.0696, 0.0733, 0.0550, 0.0703], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 03:59:45,898 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7950, 5.1269, 4.8807, 4.8922, 4.6231, 4.5742, 4.6130, 5.1901], device='cuda:3'), covar=tensor([0.1280, 0.0872, 0.1135, 0.0865, 0.0844, 0.1258, 0.1295, 0.0968], device='cuda:3'), in_proj_covar=tensor([0.0715, 0.0868, 0.0711, 0.0669, 0.0552, 0.0550, 0.0729, 0.0676], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:00:13,322 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256694.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:00:25,265 INFO [train.py:904] (3/8) Epoch 26, batch 2950, loss[loss=0.1716, simple_loss=0.2506, pruned_loss=0.04625, over 16681.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.257, pruned_loss=0.03979, over 3323565.01 frames. ], batch size: 89, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:01:01,149 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256728.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:14,580 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:18,028 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256742.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:35,075 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 3000, loss[loss=0.1433, simple_loss=0.2343, pruned_loss=0.02615, over 17221.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2578, pruned_loss=0.04039, over 3319643.89 frames. ], batch size: 43, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:01:35,091 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 04:01:44,001 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 04:02:27,414 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:29,705 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-02 04:02:31,512 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,567 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:53,042 INFO [train.py:904] (3/8) Epoch 26, batch 3050, loss[loss=0.1366, simple_loss=0.2253, pruned_loss=0.02395, over 16832.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2568, pruned_loss=0.04025, over 3317946.79 frames. ], batch size: 39, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:03:03,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4284, 5.3910, 5.2756, 4.7891, 4.9290, 5.3640, 5.2663, 4.9885], device='cuda:3'), covar=tensor([0.0566, 0.0509, 0.0303, 0.0376, 0.1065, 0.0449, 0.0276, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0472, 0.0370, 0.0375, 0.0373, 0.0428, 0.0253, 0.0445], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 04:03:27,147 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256827.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:03:27,322 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256827.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:03:38,906 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:04:02,821 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 3100, loss[loss=0.1472, simple_loss=0.2271, pruned_loss=0.03372, over 16545.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2553, pruned_loss=0.03966, over 3323119.55 frames. ], batch size: 75, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:04:47,239 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256883.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:04:53,619 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256888.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:04:56,753 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2239, 5.1893, 4.9447, 4.3819, 5.0387, 1.8741, 4.7582, 4.7728], device='cuda:3'), covar=tensor([0.0091, 0.0087, 0.0241, 0.0436, 0.0105, 0.2923, 0.0149, 0.0272], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0174, 0.0212, 0.0187, 0.0188, 0.0219, 0.0201, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:05:03,341 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-05-02 04:05:06,583 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1801, 5.1592, 4.9070, 4.3742, 4.9978, 1.7516, 4.6893, 4.7481], device='cuda:3'), covar=tensor([0.0138, 0.0108, 0.0269, 0.0469, 0.0124, 0.2978, 0.0165, 0.0258], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0174, 0.0212, 0.0187, 0.0189, 0.0219, 0.0201, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:05:13,355 INFO [train.py:904] (3/8) Epoch 26, batch 3150, loss[loss=0.1591, simple_loss=0.2621, pruned_loss=0.02804, over 17279.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2535, pruned_loss=0.03871, over 3333566.06 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:06:05,224 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256941.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:06:09,344 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256944.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:06:21,402 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 3200, loss[loss=0.1427, simple_loss=0.237, pruned_loss=0.02421, over 16990.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.252, pruned_loss=0.03841, over 3333871.15 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:06:44,264 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 04:07:11,955 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:29,905 INFO [train.py:904] (3/8) Epoch 26, batch 3250, loss[loss=0.1786, simple_loss=0.2694, pruned_loss=0.04384, over 16746.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2522, pruned_loss=0.03899, over 3335291.72 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:07:43,651 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:45,544 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 04:07:56,070 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-02 04:07:58,642 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-02 04:08:03,342 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:38,566 INFO [optim.py:368] (3/8) 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 04:08:38,581 INFO [train.py:904] (3/8) Epoch 26, batch 3300, loss[loss=0.1706, simple_loss=0.2654, pruned_loss=0.03791, over 16746.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2544, pruned_loss=0.04032, over 3329814.30 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:08:41,735 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 04:09:01,500 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 04:09:07,484 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257074.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:09,485 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257076.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:48,082 INFO [train.py:904] (3/8) Epoch 26, batch 3350, loss[loss=0.1636, simple_loss=0.2637, pruned_loss=0.03168, over 17065.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2546, pruned_loss=0.03956, over 3326464.45 frames. ], batch size: 53, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:10:20,884 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:10:56,576 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.179e+02 2.494e+02 3.258e+02 7.595e+02, threshold=4.988e+02, percent-clipped=3.0 2023-05-02 04:10:56,591 INFO [train.py:904] (3/8) Epoch 26, batch 3400, loss[loss=0.1467, simple_loss=0.2416, pruned_loss=0.02592, over 17039.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2549, pruned_loss=0.03979, over 3326068.07 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:11:02,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3356, 3.5729, 3.9910, 2.1645, 3.1123, 2.4399, 3.8541, 3.7640], device='cuda:3'), covar=tensor([0.0307, 0.0974, 0.0482, 0.2118, 0.0856, 0.1038, 0.0581, 0.1015], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 04:11:13,672 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257165.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:27,518 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257175.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:39,259 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257183.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:12:06,425 INFO [train.py:904] (3/8) Epoch 26, batch 3450, loss[loss=0.1855, simple_loss=0.2731, pruned_loss=0.04894, over 16233.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2534, pruned_loss=0.03913, over 3319554.70 frames. ], batch size: 165, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:12:07,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7093, 2.6285, 2.1304, 2.2511, 2.8860, 2.6559, 3.3639, 3.2423], device='cuda:3'), covar=tensor([0.0202, 0.0593, 0.0789, 0.0672, 0.0452, 0.0581, 0.0328, 0.0377], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0249, 0.0237, 0.0239, 0.0250, 0.0249, 0.0249, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:12:23,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1531, 5.1893, 5.5684, 5.5575, 5.6067, 5.2781, 5.1807, 5.0047], device='cuda:3'), covar=tensor([0.0327, 0.0632, 0.0426, 0.0423, 0.0448, 0.0387, 0.0943, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0445, 0.0500, 0.0485, 0.0444, 0.0532, 0.0510, 0.0591, 0.0410], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 04:12:33,996 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:39,906 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:57,439 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257239.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:12:59,822 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:13:16,458 INFO [optim.py:368] (3/8) 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 04:13:16,481 INFO [train.py:904] (3/8) Epoch 26, batch 3500, loss[loss=0.152, simple_loss=0.2463, pruned_loss=0.02884, over 17133.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2526, pruned_loss=0.03854, over 3320807.18 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:13:29,553 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 04:13:58,965 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:25,656 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 3550, loss[loss=0.1401, simple_loss=0.2264, pruned_loss=0.02692, over 15769.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2508, pruned_loss=0.03767, over 3317608.25 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:14:55,618 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3697, 5.3088, 5.0426, 4.5063, 5.1840, 1.9914, 4.8920, 4.8806], device='cuda:3'), covar=tensor([0.0093, 0.0087, 0.0247, 0.0420, 0.0098, 0.2785, 0.0159, 0.0246], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0174, 0.0211, 0.0187, 0.0188, 0.0218, 0.0201, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:15:34,858 INFO [train.py:904] (3/8) Epoch 26, batch 3600, loss[loss=0.1597, simple_loss=0.2504, pruned_loss=0.03446, over 17231.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2496, pruned_loss=0.0373, over 3316346.40 frames. ], batch size: 44, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:15:35,973 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257369.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:16:47,814 INFO [train.py:904] (3/8) Epoch 26, batch 3650, loss[loss=0.1402, simple_loss=0.2294, pruned_loss=0.02552, over 16764.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2487, pruned_loss=0.03765, over 3315825.92 frames. ], batch size: 39, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:31,487 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0489, 5.5622, 5.6946, 5.3661, 5.4140, 6.0334, 5.4843, 5.1122], device='cuda:3'), covar=tensor([0.0913, 0.1669, 0.1696, 0.1908, 0.2376, 0.0828, 0.1395, 0.2281], device='cuda:3'), in_proj_covar=tensor([0.0434, 0.0642, 0.0707, 0.0524, 0.0697, 0.0732, 0.0549, 0.0701], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 04:17:58,062 INFO [train.py:904] (3/8) Epoch 26, batch 3700, loss[loss=0.1636, simple_loss=0.2439, pruned_loss=0.04162, over 16524.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2478, pruned_loss=0.03926, over 3306627.22 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:59,893 INFO [optim.py:368] (3/8) 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 04:18:18,542 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:18:41,395 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257483.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:18:59,896 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 04:19:09,128 INFO [train.py:904] (3/8) Epoch 26, batch 3750, loss[loss=0.1634, simple_loss=0.2416, pruned_loss=0.04263, over 16893.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2483, pruned_loss=0.04064, over 3295618.36 frames. ], batch size: 90, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:19:33,996 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:19:44,829 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257528.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:19:47,967 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1107, 3.3407, 3.4356, 2.1689, 3.0037, 2.3858, 3.6238, 3.7026], device='cuda:3'), covar=tensor([0.0256, 0.0865, 0.0653, 0.2118, 0.0869, 0.1076, 0.0539, 0.0805], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 04:19:48,975 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257531.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:01,915 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257539.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:19,792 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 04:20:20,045 INFO [train.py:904] (3/8) Epoch 26, batch 3800, loss[loss=0.1481, simple_loss=0.2276, pruned_loss=0.03431, over 16831.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2492, pruned_loss=0.04191, over 3287713.16 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:20:22,165 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.842e+02 2.272e+02 2.465e+02 2.848e+02 4.858e+02, threshold=4.931e+02, percent-clipped=0.0 2023-05-02 04:20:55,722 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:20:55,797 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257587.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:09,810 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6316, 3.6265, 2.2721, 3.8985, 2.9019, 3.8346, 2.3731, 3.0047], device='cuda:3'), covar=tensor([0.0277, 0.0439, 0.1536, 0.0307, 0.0816, 0.0768, 0.1413, 0.0702], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0183, 0.0199, 0.0177, 0.0182, 0.0224, 0.0207, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 04:21:22,616 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 3850, loss[loss=0.1579, simple_loss=0.2464, pruned_loss=0.03465, over 16805.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2506, pruned_loss=0.04333, over 3275116.45 frames. ], batch size: 39, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:20,399 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257639.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:22:41,231 INFO [train.py:904] (3/8) Epoch 26, batch 3900, loss[loss=0.147, simple_loss=0.2233, pruned_loss=0.03535, over 16725.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2503, pruned_loss=0.04375, over 3284526.99 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:42,472 INFO [optim.py:368] (3/8) 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 04:23:04,477 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:23:14,821 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 04:23:51,421 INFO [train.py:904] (3/8) Epoch 26, batch 3950, loss[loss=0.1665, simple_loss=0.2516, pruned_loss=0.04064, over 16656.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2498, pruned_loss=0.04427, over 3278040.20 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:24:12,552 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:25:02,900 INFO [train.py:904] (3/8) Epoch 26, batch 4000, loss[loss=0.1843, simple_loss=0.2721, pruned_loss=0.0483, over 16743.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2495, pruned_loss=0.04466, over 3275654.97 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:25:03,989 INFO [optim.py:368] (3/8) 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 04:25:42,659 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-02 04:26:13,230 INFO [train.py:904] (3/8) Epoch 26, batch 4050, loss[loss=0.1667, simple_loss=0.2493, pruned_loss=0.04211, over 16640.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2504, pruned_loss=0.04434, over 3285137.21 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:26:40,068 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:26:42,420 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257823.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:27:01,657 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-05-02 04:27:25,375 INFO [train.py:904] (3/8) Epoch 26, batch 4100, loss[loss=0.1709, simple_loss=0.2587, pruned_loss=0.0416, over 16411.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2525, pruned_loss=0.04418, over 3259378.48 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:27:26,540 INFO [optim.py:368] (3/8) 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 04:27:31,754 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 04:27:35,106 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:27:40,483 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 04:27:49,110 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257869.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:02,831 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:10,500 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257883.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:30,688 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:39,710 INFO [train.py:904] (3/8) Epoch 26, batch 4150, loss[loss=0.1895, simple_loss=0.2835, pruned_loss=0.04777, over 16642.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2599, pruned_loss=0.04661, over 3226423.45 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:07,290 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:15,911 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:28,837 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257934.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:29:44,246 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257944.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:29:45,298 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:52,857 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:56,805 INFO [train.py:904] (3/8) Epoch 26, batch 4200, loss[loss=0.1814, simple_loss=0.2808, pruned_loss=0.04098, over 17029.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2666, pruned_loss=0.04786, over 3209490.11 frames. ], batch size: 50, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:58,476 INFO [optim.py:368] (3/8) 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 04:30:14,975 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 04:30:47,829 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 04:31:15,317 INFO [train.py:904] (3/8) Epoch 26, batch 4250, loss[loss=0.1795, simple_loss=0.2622, pruned_loss=0.04839, over 12012.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2695, pruned_loss=0.04761, over 3184266.70 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:31:29,162 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:32:29,606 INFO [train.py:904] (3/8) Epoch 26, batch 4300, loss[loss=0.1833, simple_loss=0.2741, pruned_loss=0.04626, over 11819.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2708, pruned_loss=0.0468, over 3188295.50 frames. ], batch size: 248, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:32:31,420 INFO [optim.py:368] (3/8) 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 04:33:24,270 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3179, 5.2907, 5.0396, 4.4267, 5.2495, 1.8273, 4.9223, 4.7300], device='cuda:3'), covar=tensor([0.0052, 0.0049, 0.0162, 0.0335, 0.0056, 0.3045, 0.0093, 0.0225], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0173, 0.0210, 0.0186, 0.0187, 0.0217, 0.0200, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:33:45,781 INFO [train.py:904] (3/8) Epoch 26, batch 4350, loss[loss=0.1893, simple_loss=0.2794, pruned_loss=0.04967, over 16739.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2736, pruned_loss=0.0474, over 3195913.07 frames. ], batch size: 124, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:34:15,451 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:34:51,688 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8771, 2.7737, 2.6478, 1.9222, 2.6033, 2.7663, 2.6006, 1.9797], device='cuda:3'), covar=tensor([0.0475, 0.0085, 0.0079, 0.0397, 0.0137, 0.0132, 0.0147, 0.0389], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0088, 0.0090, 0.0136, 0.0102, 0.0114, 0.0098, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 04:34:58,991 INFO [train.py:904] (3/8) Epoch 26, batch 4400, loss[loss=0.1785, simple_loss=0.267, pruned_loss=0.04496, over 17250.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.276, pruned_loss=0.04882, over 3191042.26 frames. ], batch size: 52, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:35:00,098 INFO [optim.py:368] (3/8) 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 04:35:14,241 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1300, 3.0963, 1.9465, 3.4618, 2.3284, 3.5071, 2.0680, 2.5652], device='cuda:3'), covar=tensor([0.0314, 0.0400, 0.1599, 0.0174, 0.0897, 0.0438, 0.1495, 0.0760], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0181, 0.0197, 0.0172, 0.0180, 0.0220, 0.0204, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 04:35:25,197 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 4450, loss[loss=0.1834, simple_loss=0.2865, pruned_loss=0.04016, over 16917.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2794, pruned_loss=0.05016, over 3200325.94 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:36:28,127 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:36:34,057 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7975, 4.6839, 4.5164, 3.1884, 3.9058, 4.4874, 3.8583, 2.8317], device='cuda:3'), covar=tensor([0.0529, 0.0030, 0.0049, 0.0383, 0.0109, 0.0101, 0.0113, 0.0408], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0088, 0.0090, 0.0137, 0.0102, 0.0114, 0.0099, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 04:36:36,163 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 04:36:55,327 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258234.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:37:03,330 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258239.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:37:22,665 INFO [train.py:904] (3/8) Epoch 26, batch 4500, loss[loss=0.1874, simple_loss=0.2774, pruned_loss=0.04864, over 16484.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2801, pruned_loss=0.05121, over 3201322.48 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:37:23,847 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 1.900e+02 2.188e+02 2.602e+02 4.588e+02, threshold=4.376e+02, percent-clipped=0.0 2023-05-02 04:37:48,157 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-02 04:37:57,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3327, 2.4183, 2.8551, 3.2233, 3.2080, 3.8133, 2.5063, 3.7887], device='cuda:3'), covar=tensor([0.0208, 0.0498, 0.0317, 0.0306, 0.0260, 0.0129, 0.0526, 0.0104], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0187, 0.0191, 0.0206, 0.0164, 0.0203, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:38:05,642 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258282.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:38:08,654 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 04:38:35,272 INFO [train.py:904] (3/8) Epoch 26, batch 4550, loss[loss=0.1794, simple_loss=0.2767, pruned_loss=0.04109, over 16736.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.281, pruned_loss=0.05203, over 3215326.04 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:38:39,768 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:38:44,884 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7157, 4.6061, 4.7825, 4.9254, 5.0400, 4.5900, 5.0801, 5.0853], device='cuda:3'), covar=tensor([0.1623, 0.1128, 0.1367, 0.0611, 0.0456, 0.0825, 0.0526, 0.0514], device='cuda:3'), in_proj_covar=tensor([0.0674, 0.0826, 0.0952, 0.0836, 0.0634, 0.0663, 0.0689, 0.0800], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:38:45,031 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1943, 3.5204, 3.6294, 2.2077, 3.0762, 2.1325, 3.4823, 3.7199], device='cuda:3'), covar=tensor([0.0232, 0.0738, 0.0545, 0.2050, 0.0821, 0.1109, 0.0599, 0.0852], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 04:39:48,615 INFO [train.py:904] (3/8) Epoch 26, batch 4600, loss[loss=0.1966, simple_loss=0.2753, pruned_loss=0.05895, over 11067.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.282, pruned_loss=0.05214, over 3223836.04 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:39:50,256 INFO [optim.py:368] (3/8) 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 04:40:00,027 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 04:41:03,103 INFO [train.py:904] (3/8) Epoch 26, batch 4650, loss[loss=0.1639, simple_loss=0.252, pruned_loss=0.03786, over 16537.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2804, pruned_loss=0.05139, over 3236223.46 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:42:11,296 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2894, 5.9602, 6.1408, 5.7222, 5.8409, 6.3939, 5.9270, 5.6823], device='cuda:3'), covar=tensor([0.0814, 0.1543, 0.1913, 0.1863, 0.2582, 0.0872, 0.1309, 0.2046], device='cuda:3'), in_proj_covar=tensor([0.0424, 0.0624, 0.0684, 0.0509, 0.0673, 0.0713, 0.0531, 0.0679], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 04:42:14,211 INFO [train.py:904] (3/8) Epoch 26, batch 4700, loss[loss=0.1763, simple_loss=0.261, pruned_loss=0.04583, over 16885.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2777, pruned_loss=0.05013, over 3242599.70 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:42:16,007 INFO [optim.py:368] (3/8) 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 04:42:16,465 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:26,791 INFO [train.py:904] (3/8) Epoch 26, batch 4750, loss[loss=0.1903, simple_loss=0.2743, pruned_loss=0.05317, over 11952.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2742, pruned_loss=0.04844, over 3230850.86 frames. ], batch size: 248, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:43:38,861 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,039 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,081 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:21,184 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258539.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:41,653 INFO [train.py:904] (3/8) Epoch 26, batch 4800, loss[loss=0.1628, simple_loss=0.2542, pruned_loss=0.03566, over 16509.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2706, pruned_loss=0.04647, over 3230040.18 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:44:43,328 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258563.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:11,418 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258587.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:57,755 INFO [train.py:904] (3/8) Epoch 26, batch 4850, loss[loss=0.1856, simple_loss=0.2811, pruned_loss=0.04501, over 16460.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2713, pruned_loss=0.04561, over 3218480.14 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:46:01,950 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4118, 4.4403, 4.7225, 4.6939, 4.7096, 4.4440, 4.4307, 4.3132], device='cuda:3'), covar=tensor([0.0292, 0.0468, 0.0360, 0.0402, 0.0453, 0.0340, 0.0876, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0473, 0.0460, 0.0423, 0.0508, 0.0485, 0.0563, 0.0391], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 04:46:03,360 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 4900, loss[loss=0.1686, simple_loss=0.2658, pruned_loss=0.0357, over 15461.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2704, pruned_loss=0.04419, over 3197225.56 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:47:16,710 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:47:59,872 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8870, 4.9452, 5.3105, 5.2397, 5.3100, 4.9831, 4.9093, 4.7450], device='cuda:3'), covar=tensor([0.0299, 0.0519, 0.0291, 0.0412, 0.0442, 0.0392, 0.0922, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0475, 0.0462, 0.0424, 0.0511, 0.0487, 0.0565, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 04:48:21,175 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8710, 3.9335, 4.1398, 4.0988, 4.1207, 3.9237, 3.9138, 3.9314], device='cuda:3'), covar=tensor([0.0324, 0.0555, 0.0413, 0.0451, 0.0457, 0.0413, 0.0776, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0475, 0.0462, 0.0424, 0.0510, 0.0486, 0.0564, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 04:48:29,883 INFO [train.py:904] (3/8) Epoch 26, batch 4950, loss[loss=0.1798, simple_loss=0.2772, pruned_loss=0.04123, over 16573.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2702, pruned_loss=0.04364, over 3180874.61 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:48:49,820 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2803, 4.1238, 4.0243, 2.5177, 3.5921, 4.0782, 3.5300, 2.3384], device='cuda:3'), covar=tensor([0.0619, 0.0044, 0.0048, 0.0463, 0.0104, 0.0081, 0.0122, 0.0468], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0135, 0.0102, 0.0113, 0.0098, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 04:49:17,274 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5177, 4.5316, 4.8224, 4.7700, 4.8226, 4.5492, 4.4725, 4.4233], device='cuda:3'), covar=tensor([0.0315, 0.0546, 0.0382, 0.0475, 0.0465, 0.0387, 0.0955, 0.0486], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0477, 0.0464, 0.0426, 0.0512, 0.0488, 0.0567, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 04:49:41,062 INFO [train.py:904] (3/8) Epoch 26, batch 5000, loss[loss=0.1596, simple_loss=0.2662, pruned_loss=0.02653, over 16737.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2715, pruned_loss=0.0436, over 3186448.21 frames. ], batch size: 89, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:42,186 INFO [optim.py:368] (3/8) 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 04:50:30,559 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-02 04:50:54,489 INFO [train.py:904] (3/8) Epoch 26, batch 5050, loss[loss=0.1829, simple_loss=0.2775, pruned_loss=0.04413, over 16421.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2727, pruned_loss=0.04413, over 3184414.57 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:51:05,003 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:51:35,438 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5791, 3.6472, 3.4087, 3.0953, 3.2692, 3.5027, 3.3447, 3.3111], device='cuda:3'), covar=tensor([0.0546, 0.0548, 0.0306, 0.0285, 0.0590, 0.0466, 0.1374, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0305, 0.0456, 0.0356, 0.0360, 0.0359, 0.0414, 0.0243, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:52:02,407 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6607, 3.7942, 2.2695, 4.4417, 2.8799, 4.3665, 2.4665, 3.0534], device='cuda:3'), covar=tensor([0.0338, 0.0363, 0.1785, 0.0155, 0.0852, 0.0373, 0.1576, 0.0824], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0180, 0.0196, 0.0170, 0.0179, 0.0218, 0.0204, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 04:52:07,189 INFO [train.py:904] (3/8) Epoch 26, batch 5100, loss[loss=0.1574, simple_loss=0.2472, pruned_loss=0.0338, over 17040.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2705, pruned_loss=0.0433, over 3198790.47 frames. ], batch size: 50, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:52:08,933 INFO [optim.py:368] (3/8) 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 04:52:20,371 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5770, 3.6504, 3.4118, 3.1145, 3.2513, 3.5100, 3.3667, 3.3253], device='cuda:3'), covar=tensor([0.0555, 0.0538, 0.0321, 0.0303, 0.0605, 0.0485, 0.1157, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0455, 0.0356, 0.0359, 0.0359, 0.0413, 0.0242, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 04:52:26,609 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8321, 3.8535, 3.9892, 3.6957, 3.8894, 4.3104, 3.9880, 3.6390], device='cuda:3'), covar=tensor([0.2156, 0.2088, 0.2373, 0.2364, 0.2508, 0.1631, 0.1484, 0.2437], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0617, 0.0678, 0.0505, 0.0667, 0.0705, 0.0529, 0.0674], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 04:52:27,817 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:53:21,028 INFO [train.py:904] (3/8) Epoch 26, batch 5150, loss[loss=0.1776, simple_loss=0.2684, pruned_loss=0.04343, over 16550.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2705, pruned_loss=0.04281, over 3178855.23 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:53:45,427 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 04:54:35,099 INFO [train.py:904] (3/8) Epoch 26, batch 5200, loss[loss=0.1583, simple_loss=0.2461, pruned_loss=0.03528, over 16895.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2684, pruned_loss=0.04189, over 3186519.05 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:36,752 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 1.846e+02 2.126e+02 2.536e+02 3.966e+02, threshold=4.251e+02, percent-clipped=0.0 2023-05-02 04:54:53,526 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:55:48,686 INFO [train.py:904] (3/8) Epoch 26, batch 5250, loss[loss=0.216, simple_loss=0.2949, pruned_loss=0.06859, over 12410.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.267, pruned_loss=0.04203, over 3179859.51 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:56:23,078 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259026.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:57:03,179 INFO [train.py:904] (3/8) Epoch 26, batch 5300, loss[loss=0.1529, simple_loss=0.2441, pruned_loss=0.03089, over 16912.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2637, pruned_loss=0.04112, over 3199147.92 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:57:04,412 INFO [optim.py:368] (3/8) 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 04:58:06,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1741, 3.2482, 1.7090, 3.5058, 2.3277, 3.5409, 1.8667, 2.5862], device='cuda:3'), covar=tensor([0.0359, 0.0418, 0.2073, 0.0209, 0.0967, 0.0534, 0.1886, 0.0888], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0180, 0.0195, 0.0170, 0.0178, 0.0218, 0.0204, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 04:58:14,375 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 04:58:18,025 INFO [train.py:904] (3/8) Epoch 26, batch 5350, loss[loss=0.1832, simple_loss=0.2777, pruned_loss=0.04432, over 16166.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2624, pruned_loss=0.04043, over 3211147.45 frames. ], batch size: 165, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:58:27,843 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:31,423 INFO [train.py:904] (3/8) Epoch 26, batch 5400, loss[loss=0.1734, simple_loss=0.2726, pruned_loss=0.03711, over 15426.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2641, pruned_loss=0.04079, over 3227652.40 frames. ], batch size: 191, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:59:32,572 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 1.944e+02 2.298e+02 2.661e+02 4.475e+02, threshold=4.595e+02, percent-clipped=1.0 2023-05-02 04:59:38,341 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:52,963 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:58,612 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2705, 5.2429, 5.1078, 4.3799, 5.1563, 1.7120, 4.8965, 4.8439], device='cuda:3'), covar=tensor([0.0092, 0.0080, 0.0179, 0.0481, 0.0098, 0.3023, 0.0127, 0.0240], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0171, 0.0209, 0.0184, 0.0185, 0.0216, 0.0199, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:00:48,641 INFO [train.py:904] (3/8) Epoch 26, batch 5450, loss[loss=0.2038, simple_loss=0.2815, pruned_loss=0.06308, over 16890.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2663, pruned_loss=0.04207, over 3200629.98 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:01:08,446 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:01:17,617 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1519, 5.1414, 5.0092, 4.6288, 4.6233, 5.0488, 4.9666, 4.7095], device='cuda:3'), covar=tensor([0.0643, 0.0768, 0.0293, 0.0324, 0.1147, 0.0589, 0.0310, 0.0704], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0459, 0.0359, 0.0362, 0.0361, 0.0418, 0.0244, 0.0433], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:01:24,993 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:05,214 INFO [train.py:904] (3/8) Epoch 26, batch 5500, loss[loss=0.278, simple_loss=0.3412, pruned_loss=0.1074, over 11717.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2739, pruned_loss=0.04732, over 3140038.48 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:02:07,103 INFO [optim.py:368] (3/8) 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 05:02:23,166 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259264.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:58,171 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:22,300 INFO [train.py:904] (3/8) Epoch 26, batch 5550, loss[loss=0.2097, simple_loss=0.2975, pruned_loss=0.06094, over 16563.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2811, pruned_loss=0.05207, over 3123825.17 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:03:40,919 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 05:03:50,533 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:57,671 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259325.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:03:58,825 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2306, 4.3540, 4.5292, 4.3208, 4.4120, 4.8527, 4.4113, 4.1609], device='cuda:3'), covar=tensor([0.1709, 0.1922, 0.2117, 0.1952, 0.2300, 0.1041, 0.1673, 0.2395], device='cuda:3'), in_proj_covar=tensor([0.0419, 0.0615, 0.0676, 0.0505, 0.0666, 0.0704, 0.0527, 0.0672], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 05:04:40,412 INFO [train.py:904] (3/8) Epoch 26, batch 5600, loss[loss=0.1947, simple_loss=0.283, pruned_loss=0.05324, over 16641.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2856, pruned_loss=0.05546, over 3104836.07 frames. ], batch size: 134, lr: 2.58e-03, grad_scale: 16.0 2023-05-02 05:04:41,807 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 3.012e+02 3.705e+02 4.647e+02 1.087e+03, threshold=7.410e+02, percent-clipped=8.0 2023-05-02 05:06:04,635 INFO [train.py:904] (3/8) Epoch 26, batch 5650, loss[loss=0.1952, simple_loss=0.2799, pruned_loss=0.05522, over 16671.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2901, pruned_loss=0.05937, over 3085201.86 frames. ], batch size: 57, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:06:22,265 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259414.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:07:18,695 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-02 05:07:22,187 INFO [train.py:904] (3/8) Epoch 26, batch 5700, loss[loss=0.2251, simple_loss=0.3132, pruned_loss=0.06853, over 15248.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2915, pruned_loss=0.06031, over 3085992.28 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:07:25,089 INFO [optim.py:368] (3/8) 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 05:07:29,829 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1435, 3.2094, 1.8449, 3.4465, 2.4047, 3.4643, 2.1488, 2.6415], device='cuda:3'), covar=tensor([0.0350, 0.0442, 0.1868, 0.0253, 0.0909, 0.0649, 0.1616, 0.0892], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0181, 0.0198, 0.0171, 0.0179, 0.0220, 0.0205, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 05:07:56,917 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259475.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:08:39,968 INFO [train.py:904] (3/8) Epoch 26, batch 5750, loss[loss=0.2088, simple_loss=0.3003, pruned_loss=0.05864, over 16441.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2942, pruned_loss=0.062, over 3056230.87 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:02,533 INFO [train.py:904] (3/8) Epoch 26, batch 5800, loss[loss=0.2027, simple_loss=0.2944, pruned_loss=0.0555, over 17143.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2941, pruned_loss=0.06148, over 3038345.75 frames. ], batch size: 48, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:05,675 INFO [optim.py:368] (3/8) 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 05:10:06,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9074, 2.1707, 2.4150, 3.1456, 2.1696, 2.3550, 2.3268, 2.2512], device='cuda:3'), covar=tensor([0.1478, 0.3374, 0.2718, 0.0918, 0.4417, 0.2567, 0.3267, 0.3652], device='cuda:3'), in_proj_covar=tensor([0.0412, 0.0463, 0.0377, 0.0331, 0.0440, 0.0529, 0.0434, 0.0541], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:10:47,016 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:10:51,182 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2184, 4.1246, 4.2826, 4.4397, 4.5548, 4.1287, 4.4892, 4.5661], device='cuda:3'), covar=tensor([0.2033, 0.1336, 0.1624, 0.0750, 0.0623, 0.1303, 0.0870, 0.0747], device='cuda:3'), in_proj_covar=tensor([0.0666, 0.0816, 0.0942, 0.0825, 0.0630, 0.0655, 0.0681, 0.0794], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:11:19,028 INFO [train.py:904] (3/8) Epoch 26, batch 5850, loss[loss=0.1883, simple_loss=0.2827, pruned_loss=0.04699, over 16786.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2919, pruned_loss=0.05985, over 3032879.44 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:11:44,959 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259620.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:11:46,971 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:12:39,963 INFO [train.py:904] (3/8) Epoch 26, batch 5900, loss[loss=0.2024, simple_loss=0.2801, pruned_loss=0.06232, over 15332.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2912, pruned_loss=0.06, over 3038316.21 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:12:43,692 INFO [optim.py:368] (3/8) 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 05:13:01,480 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 05:13:08,192 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 5950, loss[loss=0.2172, simple_loss=0.3047, pruned_loss=0.06485, over 16574.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2919, pruned_loss=0.0591, over 3040859.47 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:14:25,471 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5834, 2.5239, 2.2590, 3.7070, 2.5492, 3.7514, 1.3465, 2.6722], device='cuda:3'), covar=tensor([0.1495, 0.0834, 0.1398, 0.0196, 0.0213, 0.0423, 0.1907, 0.0916], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0179, 0.0199, 0.0198, 0.0207, 0.0217, 0.0208, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 05:15:12,591 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 6000, loss[loss=0.194, simple_loss=0.2767, pruned_loss=0.05565, over 17013.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2905, pruned_loss=0.0581, over 3067961.00 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:15:18,036 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 05:15:28,183 INFO [train.py:938] (3/8) Epoch 26, validation: loss=0.1485, simple_loss=0.2607, pruned_loss=0.01818, over 944034.00 frames. 2023-05-02 05:15:28,184 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 05:15:30,551 INFO [optim.py:368] (3/8) 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 05:15:55,082 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:13,696 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:46,427 INFO [train.py:904] (3/8) Epoch 26, batch 6050, loss[loss=0.2033, simple_loss=0.2927, pruned_loss=0.05695, over 16903.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2888, pruned_loss=0.05687, over 3097986.20 frames. ], batch size: 109, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:16:59,619 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259810.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:17:51,027 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259843.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:18:02,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0518, 3.0908, 2.5314, 3.0111, 3.4314, 3.1147, 3.5741, 3.6605], device='cuda:3'), covar=tensor([0.0096, 0.0388, 0.0532, 0.0368, 0.0261, 0.0344, 0.0252, 0.0250], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0239, 0.0229, 0.0230, 0.0240, 0.0238, 0.0238, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:18:05,774 INFO [train.py:904] (3/8) Epoch 26, batch 6100, loss[loss=0.1869, simple_loss=0.2801, pruned_loss=0.04682, over 16437.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2887, pruned_loss=0.0562, over 3109824.71 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:18:09,301 INFO [optim.py:368] (3/8) 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 05:18:27,816 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:18:53,028 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 6150, loss[loss=0.2211, simple_loss=0.2947, pruned_loss=0.07381, over 11455.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2868, pruned_loss=0.05555, over 3118529.06 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:19:48,995 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:00,767 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:05,847 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:42,381 INFO [train.py:904] (3/8) Epoch 26, batch 6200, loss[loss=0.1663, simple_loss=0.255, pruned_loss=0.03878, over 17235.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2851, pruned_loss=0.05522, over 3113402.86 frames. ], batch size: 52, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:20:44,631 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259968.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:21:57,629 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 05:22:00,546 INFO [train.py:904] (3/8) Epoch 26, batch 6250, loss[loss=0.1879, simple_loss=0.2817, pruned_loss=0.0471, over 16970.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2852, pruned_loss=0.05547, over 3109596.56 frames. ], batch size: 55, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:22:41,204 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:23:15,627 INFO [train.py:904] (3/8) Epoch 26, batch 6300, loss[loss=0.1883, simple_loss=0.2752, pruned_loss=0.05069, over 16687.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2848, pruned_loss=0.05476, over 3121916.82 frames. ], batch size: 134, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:23:19,599 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.817e+02 3.510e+02 4.108e+02 7.742e+02, threshold=7.020e+02, percent-clipped=4.0 2023-05-02 05:23:44,831 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260070.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:07,186 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2572, 5.8814, 6.1236, 5.6768, 5.8804, 6.3534, 5.9235, 5.6446], device='cuda:3'), covar=tensor([0.0858, 0.1713, 0.2247, 0.2082, 0.2249, 0.1088, 0.1488, 0.2426], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0627, 0.0689, 0.0513, 0.0677, 0.0718, 0.0538, 0.0683], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 05:24:11,659 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8269, 1.9443, 2.3749, 2.7097, 2.6568, 3.0811, 1.9269, 3.0172], device='cuda:3'), covar=tensor([0.0222, 0.0592, 0.0379, 0.0318, 0.0374, 0.0208, 0.0643, 0.0162], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0196, 0.0185, 0.0189, 0.0204, 0.0162, 0.0202, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:24:17,926 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:31,743 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0254, 4.9864, 4.8477, 4.1172, 4.9162, 2.0371, 4.6475, 4.5483], device='cuda:3'), covar=tensor([0.0089, 0.0085, 0.0183, 0.0406, 0.0084, 0.2663, 0.0119, 0.0243], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0172, 0.0210, 0.0185, 0.0186, 0.0216, 0.0198, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:24:35,598 INFO [train.py:904] (3/8) Epoch 26, batch 6350, loss[loss=0.1975, simple_loss=0.2922, pruned_loss=0.05137, over 16723.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2852, pruned_loss=0.05548, over 3117006.13 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:24:39,473 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260105.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:24:45,626 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8058, 4.8764, 5.2377, 5.2007, 5.2336, 4.8955, 4.8434, 4.7141], device='cuda:3'), covar=tensor([0.0361, 0.0630, 0.0339, 0.0380, 0.0427, 0.0399, 0.0978, 0.0526], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0480, 0.0466, 0.0427, 0.0513, 0.0490, 0.0568, 0.0394], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 05:24:59,054 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260118.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:25:29,719 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260138.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:25:52,108 INFO [train.py:904] (3/8) Epoch 26, batch 6400, loss[loss=0.1878, simple_loss=0.2715, pruned_loss=0.05206, over 17228.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2852, pruned_loss=0.05647, over 3107528.98 frames. ], batch size: 44, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:25:54,618 INFO [optim.py:368] (3/8) 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 05:25:55,826 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2146, 2.9737, 3.2989, 1.7957, 3.4138, 3.4967, 2.7412, 2.6697], device='cuda:3'), covar=tensor([0.0873, 0.0328, 0.0227, 0.1235, 0.0104, 0.0213, 0.0482, 0.0508], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0100, 0.0140, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 05:26:53,689 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5044, 2.1511, 1.6381, 1.8521, 2.4960, 2.0851, 2.1980, 2.6442], device='cuda:3'), covar=tensor([0.0402, 0.0580, 0.0847, 0.0682, 0.0367, 0.0520, 0.0408, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0240, 0.0230, 0.0231, 0.0240, 0.0239, 0.0239, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:27:08,957 INFO [train.py:904] (3/8) Epoch 26, batch 6450, loss[loss=0.1875, simple_loss=0.2763, pruned_loss=0.04937, over 15255.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2861, pruned_loss=0.05646, over 3087419.60 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:27:39,882 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:27:43,919 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260224.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:28:28,517 INFO [train.py:904] (3/8) Epoch 26, batch 6500, loss[loss=0.1954, simple_loss=0.2695, pruned_loss=0.06061, over 11742.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2841, pruned_loss=0.05599, over 3090067.18 frames. ], batch size: 249, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:28:31,534 INFO [optim.py:368] (3/8) 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 05:28:40,998 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0018, 5.4804, 5.7294, 5.3832, 5.4670, 6.0486, 5.5365, 5.2136], device='cuda:3'), covar=tensor([0.1075, 0.1795, 0.2286, 0.1957, 0.2222, 0.0893, 0.1547, 0.2292], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0628, 0.0690, 0.0513, 0.0679, 0.0718, 0.0538, 0.0684], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 05:29:20,121 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260285.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:29:49,649 INFO [train.py:904] (3/8) Epoch 26, batch 6550, loss[loss=0.2503, simple_loss=0.3176, pruned_loss=0.09154, over 11327.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2866, pruned_loss=0.05678, over 3087590.74 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:30:10,787 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1295, 5.1398, 5.5473, 5.5178, 5.5478, 5.2464, 5.1974, 5.0419], device='cuda:3'), covar=tensor([0.0419, 0.0837, 0.0666, 0.0614, 0.0558, 0.0704, 0.0971, 0.0529], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0477, 0.0463, 0.0424, 0.0510, 0.0486, 0.0563, 0.0390], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 05:30:12,163 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4985, 4.4875, 4.3750, 3.5861, 4.4046, 1.7024, 4.1485, 3.9280], device='cuda:3'), covar=tensor([0.0157, 0.0143, 0.0243, 0.0346, 0.0118, 0.3068, 0.0167, 0.0303], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0171, 0.0209, 0.0184, 0.0185, 0.0216, 0.0198, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:30:40,273 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 05:31:07,550 INFO [train.py:904] (3/8) Epoch 26, batch 6600, loss[loss=0.2204, simple_loss=0.2997, pruned_loss=0.07051, over 15342.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2885, pruned_loss=0.05668, over 3097521.72 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:08,237 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 05:31:09,946 INFO [optim.py:368] (3/8) 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 05:31:27,078 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8185, 1.9109, 2.4607, 2.6944, 2.6744, 3.1162, 2.0203, 3.1094], device='cuda:3'), covar=tensor([0.0231, 0.0550, 0.0319, 0.0345, 0.0337, 0.0203, 0.0586, 0.0150], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0188, 0.0203, 0.0161, 0.0200, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:31:59,445 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:32:26,376 INFO [train.py:904] (3/8) Epoch 26, batch 6650, loss[loss=0.1916, simple_loss=0.2798, pruned_loss=0.05173, over 16399.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2886, pruned_loss=0.05722, over 3097897.50 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:32:30,294 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260405.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:33:21,186 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:33:43,208 INFO [train.py:904] (3/8) Epoch 26, batch 6700, loss[loss=0.23, simple_loss=0.3009, pruned_loss=0.07959, over 11965.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2877, pruned_loss=0.05757, over 3083068.05 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:33:43,550 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260453.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:33:43,703 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5436, 1.7347, 2.2128, 2.4272, 2.4210, 2.7668, 1.8861, 2.7116], device='cuda:3'), covar=tensor([0.0234, 0.0545, 0.0321, 0.0357, 0.0360, 0.0225, 0.0569, 0.0152], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0194, 0.0182, 0.0188, 0.0203, 0.0160, 0.0199, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:33:45,910 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.097e+02 2.665e+02 3.203e+02 3.728e+02 7.997e+02, threshold=6.406e+02, percent-clipped=3.0 2023-05-02 05:34:05,648 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1456, 2.2831, 2.2956, 3.7952, 2.1967, 2.6227, 2.3413, 2.4132], device='cuda:3'), covar=tensor([0.1383, 0.3495, 0.3017, 0.0587, 0.3989, 0.2456, 0.3523, 0.3368], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0464, 0.0378, 0.0332, 0.0442, 0.0531, 0.0436, 0.0543], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:34:35,489 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260486.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:35:01,130 INFO [train.py:904] (3/8) Epoch 26, batch 6750, loss[loss=0.2131, simple_loss=0.2937, pruned_loss=0.06627, over 16193.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2867, pruned_loss=0.05769, over 3070329.31 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:35:31,717 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260522.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:35:55,719 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 05:36:16,941 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8823, 2.1800, 2.4514, 3.1210, 2.1927, 2.3897, 2.3498, 2.3029], device='cuda:3'), covar=tensor([0.1398, 0.3198, 0.2500, 0.0760, 0.3983, 0.2342, 0.3118, 0.3148], device='cuda:3'), in_proj_covar=tensor([0.0415, 0.0465, 0.0379, 0.0332, 0.0443, 0.0533, 0.0437, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:36:19,269 INFO [train.py:904] (3/8) Epoch 26, batch 6800, loss[loss=0.2067, simple_loss=0.2973, pruned_loss=0.05808, over 16439.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2876, pruned_loss=0.05792, over 3055290.58 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:36:21,474 INFO [optim.py:368] (3/8) 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 05:36:45,310 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:37:01,662 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260580.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:37:32,173 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8300, 3.8790, 4.1571, 4.1182, 4.1280, 3.9003, 3.9185, 3.9404], device='cuda:3'), covar=tensor([0.0407, 0.0734, 0.0445, 0.0436, 0.0496, 0.0509, 0.0940, 0.0584], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0478, 0.0464, 0.0425, 0.0512, 0.0486, 0.0564, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 05:37:35,980 INFO [train.py:904] (3/8) Epoch 26, batch 6850, loss[loss=0.1805, simple_loss=0.2912, pruned_loss=0.03486, over 16808.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2885, pruned_loss=0.05829, over 3056415.73 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:07,425 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9882, 2.2096, 2.1650, 3.7105, 2.0353, 2.5057, 2.2324, 2.3200], device='cuda:3'), covar=tensor([0.1613, 0.3871, 0.3279, 0.0611, 0.4282, 0.2612, 0.4154, 0.3239], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0463, 0.0377, 0.0331, 0.0441, 0.0530, 0.0435, 0.0541], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:38:50,099 INFO [train.py:904] (3/8) Epoch 26, batch 6900, loss[loss=0.1878, simple_loss=0.2837, pruned_loss=0.04599, over 16758.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2904, pruned_loss=0.05713, over 3081201.00 frames. ], batch size: 83, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:53,861 INFO [optim.py:368] (3/8) 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 05:39:40,138 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260686.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:39:52,538 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-05-02 05:40:05,852 INFO [train.py:904] (3/8) Epoch 26, batch 6950, loss[loss=0.1765, simple_loss=0.2699, pruned_loss=0.0415, over 16868.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.292, pruned_loss=0.05847, over 3074672.87 frames. ], batch size: 96, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:40:48,886 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6145, 5.6356, 5.4747, 5.0469, 5.0554, 5.5206, 5.5080, 5.2217], device='cuda:3'), covar=tensor([0.0656, 0.0514, 0.0336, 0.0374, 0.1222, 0.0490, 0.0307, 0.0773], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0452, 0.0352, 0.0354, 0.0352, 0.0408, 0.0242, 0.0426], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:40:48,942 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260734.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:21,377 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:23,879 INFO [train.py:904] (3/8) Epoch 26, batch 7000, loss[loss=0.1933, simple_loss=0.2926, pruned_loss=0.04699, over 16467.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2918, pruned_loss=0.05794, over 3063916.64 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:41:29,415 INFO [optim.py:368] (3/8) 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 05:42:24,292 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:42:42,008 INFO [train.py:904] (3/8) Epoch 26, batch 7050, loss[loss=0.2108, simple_loss=0.297, pruned_loss=0.06226, over 16742.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2921, pruned_loss=0.05716, over 3084923.13 frames. ], batch size: 124, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:42:42,560 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7178, 1.8213, 1.6568, 1.5082, 1.9407, 1.6031, 1.6302, 1.9135], device='cuda:3'), covar=tensor([0.0227, 0.0315, 0.0435, 0.0353, 0.0237, 0.0282, 0.0175, 0.0215], device='cuda:3'), in_proj_covar=tensor([0.0223, 0.0239, 0.0231, 0.0231, 0.0241, 0.0239, 0.0238, 0.0238], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:42:55,688 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:42:57,566 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.58 vs. limit=5.0 2023-05-02 05:43:59,434 INFO [train.py:904] (3/8) Epoch 26, batch 7100, loss[loss=0.1931, simple_loss=0.2847, pruned_loss=0.05077, over 17041.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2907, pruned_loss=0.05741, over 3080239.22 frames. ], batch size: 50, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:44:05,368 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.003e+02 2.616e+02 2.972e+02 3.634e+02 7.546e+02, threshold=5.943e+02, percent-clipped=1.0 2023-05-02 05:44:42,694 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260880.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:44:44,308 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 7150, loss[loss=0.2032, simple_loss=0.2819, pruned_loss=0.06225, over 16886.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2881, pruned_loss=0.05654, over 3092209.39 frames. ], batch size: 109, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:45:53,426 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260928.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:46:13,541 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260942.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:46:29,364 INFO [train.py:904] (3/8) Epoch 26, batch 7200, loss[loss=0.1783, simple_loss=0.2716, pruned_loss=0.0425, over 15268.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2857, pruned_loss=0.05441, over 3126152.75 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:46:35,554 INFO [optim.py:368] (3/8) 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 05:46:46,366 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4399, 3.3199, 2.6104, 2.1645, 2.1984, 2.2991, 3.4938, 3.0359], device='cuda:3'), covar=tensor([0.3220, 0.0796, 0.2043, 0.2819, 0.2772, 0.2356, 0.0559, 0.1471], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0275, 0.0311, 0.0324, 0.0305, 0.0273, 0.0302, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 05:47:50,258 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5135, 3.5874, 3.3249, 2.9524, 3.2138, 3.4789, 3.3273, 3.2830], device='cuda:3'), covar=tensor([0.0579, 0.0671, 0.0278, 0.0273, 0.0487, 0.0497, 0.1257, 0.0479], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0451, 0.0351, 0.0354, 0.0351, 0.0407, 0.0241, 0.0425], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:47:53,031 INFO [train.py:904] (3/8) Epoch 26, batch 7250, loss[loss=0.1694, simple_loss=0.2545, pruned_loss=0.04218, over 16302.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2836, pruned_loss=0.05348, over 3115944.28 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:48:35,665 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8010, 2.8670, 2.4314, 2.7229, 3.1477, 2.8121, 3.3388, 3.3623], device='cuda:3'), covar=tensor([0.0115, 0.0386, 0.0505, 0.0417, 0.0264, 0.0390, 0.0217, 0.0276], device='cuda:3'), in_proj_covar=tensor([0.0221, 0.0237, 0.0229, 0.0230, 0.0240, 0.0237, 0.0237, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:48:38,970 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.84 vs. limit=5.0 2023-05-02 05:48:58,407 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.48 vs. limit=5.0 2023-05-02 05:49:09,253 INFO [train.py:904] (3/8) Epoch 26, batch 7300, loss[loss=0.2004, simple_loss=0.2997, pruned_loss=0.05062, over 16695.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2834, pruned_loss=0.05378, over 3120896.22 frames. ], batch size: 134, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:49:15,972 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.610e+02 3.123e+02 3.831e+02 8.496e+02, threshold=6.245e+02, percent-clipped=1.0 2023-05-02 05:49:31,457 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 05:50:00,843 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:50:26,325 INFO [train.py:904] (3/8) Epoch 26, batch 7350, loss[loss=0.1908, simple_loss=0.2744, pruned_loss=0.05364, over 16294.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.285, pruned_loss=0.05499, over 3107522.80 frames. ], batch size: 35, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:50:33,366 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261107.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:51:18,064 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261136.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:51:40,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9293, 4.1781, 4.0134, 4.0559, 3.7556, 3.7955, 3.8352, 4.1762], device='cuda:3'), covar=tensor([0.1096, 0.0829, 0.0974, 0.0833, 0.0728, 0.1804, 0.0946, 0.0948], device='cuda:3'), in_proj_covar=tensor([0.0695, 0.0833, 0.0689, 0.0644, 0.0532, 0.0533, 0.0706, 0.0653], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:51:44,987 INFO [train.py:904] (3/8) Epoch 26, batch 7400, loss[loss=0.2004, simple_loss=0.2993, pruned_loss=0.0508, over 16852.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.286, pruned_loss=0.05539, over 3125588.91 frames. ], batch size: 42, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:51:50,753 INFO [optim.py:368] (3/8) 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 05:52:43,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9300, 2.7720, 2.8498, 2.1543, 2.7055, 2.1415, 2.7489, 2.9658], device='cuda:3'), covar=tensor([0.0292, 0.0769, 0.0588, 0.1756, 0.0798, 0.0949, 0.0601, 0.0690], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 05:52:55,110 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261197.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:53:04,380 INFO [train.py:904] (3/8) Epoch 26, batch 7450, loss[loss=0.2095, simple_loss=0.3066, pruned_loss=0.05621, over 16300.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2867, pruned_loss=0.05573, over 3130948.01 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:53:47,872 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6051, 4.4332, 4.6360, 4.7925, 4.9639, 4.5033, 4.9422, 4.9918], device='cuda:3'), covar=tensor([0.2009, 0.1482, 0.1722, 0.0810, 0.0617, 0.0990, 0.0670, 0.0691], device='cuda:3'), in_proj_covar=tensor([0.0652, 0.0805, 0.0929, 0.0816, 0.0624, 0.0647, 0.0677, 0.0790], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:54:01,620 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:54:27,108 INFO [train.py:904] (3/8) Epoch 26, batch 7500, loss[loss=0.1964, simple_loss=0.2762, pruned_loss=0.05829, over 16531.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2864, pruned_loss=0.05484, over 3115207.03 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:32,597 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2308, 4.3552, 4.5116, 4.2561, 4.3724, 4.8547, 4.4242, 4.1560], device='cuda:3'), covar=tensor([0.1716, 0.2040, 0.2480, 0.2320, 0.2588, 0.1227, 0.1720, 0.2463], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0626, 0.0689, 0.0510, 0.0676, 0.0717, 0.0538, 0.0684], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 05:54:33,503 INFO [optim.py:368] (3/8) 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 05:55:45,646 INFO [train.py:904] (3/8) Epoch 26, batch 7550, loss[loss=0.2088, simple_loss=0.2965, pruned_loss=0.06058, over 16351.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2856, pruned_loss=0.05551, over 3111651.94 frames. ], batch size: 146, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:56:42,568 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 05:57:01,728 INFO [train.py:904] (3/8) Epoch 26, batch 7600, loss[loss=0.2377, simple_loss=0.3076, pruned_loss=0.08387, over 11371.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2856, pruned_loss=0.05673, over 3078726.10 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 05:57:07,520 INFO [optim.py:368] (3/8) 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 05:57:35,105 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9617, 4.9619, 4.7513, 4.0290, 4.8833, 1.7883, 4.6476, 4.3811], device='cuda:3'), covar=tensor([0.0097, 0.0083, 0.0191, 0.0382, 0.0082, 0.2940, 0.0117, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0168, 0.0206, 0.0181, 0.0182, 0.0212, 0.0194, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 05:57:54,670 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:58:20,640 INFO [train.py:904] (3/8) Epoch 26, batch 7650, loss[loss=0.2246, simple_loss=0.3083, pruned_loss=0.07041, over 16445.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2861, pruned_loss=0.05701, over 3085669.37 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 05:58:26,651 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:08,082 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 7700, loss[loss=0.1924, simple_loss=0.2796, pruned_loss=0.05259, over 16676.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2865, pruned_loss=0.05767, over 3067827.42 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 05:59:40,109 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261455.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:42,112 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 06:00:13,685 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1030, 3.4423, 3.4440, 2.0825, 3.0704, 2.2786, 3.5673, 3.7185], device='cuda:3'), covar=tensor([0.0263, 0.0795, 0.0639, 0.2174, 0.0845, 0.1045, 0.0599, 0.0888], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 06:00:36,558 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261492.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 06:00:37,091 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 06:00:52,204 INFO [train.py:904] (3/8) Epoch 26, batch 7750, loss[loss=0.1675, simple_loss=0.268, pruned_loss=0.03345, over 16893.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2873, pruned_loss=0.05745, over 3080760.02 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:01:13,782 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:01:34,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7326, 4.5857, 4.7626, 4.9096, 5.0704, 4.6010, 5.0771, 5.0936], device='cuda:3'), covar=tensor([0.2161, 0.1415, 0.1773, 0.0881, 0.0751, 0.1050, 0.0833, 0.0752], device='cuda:3'), in_proj_covar=tensor([0.0652, 0.0805, 0.0928, 0.0815, 0.0623, 0.0648, 0.0676, 0.0789], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:01:44,979 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:02:09,780 INFO [train.py:904] (3/8) Epoch 26, batch 7800, loss[loss=0.2235, simple_loss=0.3088, pruned_loss=0.06905, over 15295.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2883, pruned_loss=0.05819, over 3066506.88 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:02:19,354 INFO [optim.py:368] (3/8) 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 06:03:00,430 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261585.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:03:16,433 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 06:03:28,317 INFO [train.py:904] (3/8) Epoch 26, batch 7850, loss[loss=0.1805, simple_loss=0.2693, pruned_loss=0.04589, over 16576.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2887, pruned_loss=0.05752, over 3083329.88 frames. ], batch size: 62, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:46,515 INFO [train.py:904] (3/8) Epoch 26, batch 7900, loss[loss=0.2129, simple_loss=0.2875, pruned_loss=0.06911, over 11823.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.287, pruned_loss=0.05664, over 3078716.53 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:55,564 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.825e+02 2.587e+02 3.177e+02 3.708e+02 5.885e+02, threshold=6.353e+02, percent-clipped=0.0 2023-05-02 06:06:05,686 INFO [train.py:904] (3/8) Epoch 26, batch 7950, loss[loss=0.2724, simple_loss=0.3229, pruned_loss=0.111, over 11723.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2867, pruned_loss=0.05696, over 3069886.65 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:06:22,708 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-02 06:07:23,908 INFO [train.py:904] (3/8) Epoch 26, batch 8000, loss[loss=0.204, simple_loss=0.2987, pruned_loss=0.05466, over 16335.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.287, pruned_loss=0.05725, over 3083219.58 frames. ], batch size: 75, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:07:24,740 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 06:07:32,099 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.5656, 2.7202, 2.5595, 4.0279, 2.8622, 3.9667, 1.5947, 2.7478], device='cuda:3'), covar=tensor([0.1504, 0.0792, 0.1219, 0.0195, 0.0274, 0.0423, 0.1743, 0.0915], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0179, 0.0199, 0.0198, 0.0207, 0.0217, 0.0208, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 06:07:32,670 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.545e+02 3.130e+02 3.720e+02 6.505e+02, threshold=6.260e+02, percent-clipped=1.0 2023-05-02 06:07:38,176 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 06:08:18,257 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 06:08:23,667 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261792.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 06:08:40,303 INFO [train.py:904] (3/8) Epoch 26, batch 8050, loss[loss=0.2042, simple_loss=0.288, pruned_loss=0.06022, over 16693.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2869, pruned_loss=0.0574, over 3067975.08 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:08:54,192 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:13,029 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261824.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:27,086 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:37,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261840.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 06:09:57,223 INFO [train.py:904] (3/8) Epoch 26, batch 8100, loss[loss=0.2051, simple_loss=0.2925, pruned_loss=0.05887, over 15450.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2866, pruned_loss=0.05684, over 3069162.94 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:10:06,903 INFO [optim.py:368] (3/8) 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 06:10:14,209 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 06:10:31,907 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8181, 3.9702, 2.4615, 4.7277, 3.2716, 4.6108, 2.5329, 3.1582], device='cuda:3'), covar=tensor([0.0299, 0.0419, 0.1753, 0.0280, 0.0740, 0.0544, 0.1628, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0180, 0.0198, 0.0171, 0.0180, 0.0221, 0.0206, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 06:10:46,828 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:00,373 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:13,578 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3791, 2.9590, 2.7002, 2.2858, 2.3145, 2.3591, 2.9580, 2.8772], device='cuda:3'), covar=tensor([0.2580, 0.0768, 0.1663, 0.2682, 0.2359, 0.2226, 0.0565, 0.1462], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0274, 0.0310, 0.0323, 0.0305, 0.0273, 0.0302, 0.0348], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:11:14,754 INFO [train.py:904] (3/8) Epoch 26, batch 8150, loss[loss=0.2163, simple_loss=0.2871, pruned_loss=0.07276, over 11699.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2837, pruned_loss=0.05552, over 3085620.56 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:20,197 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4798, 2.7381, 2.7366, 4.5598, 2.4954, 2.9716, 2.6466, 2.7984], device='cuda:3'), covar=tensor([0.1435, 0.3162, 0.2723, 0.0472, 0.3927, 0.2316, 0.3410, 0.3119], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0466, 0.0379, 0.0332, 0.0443, 0.0533, 0.0437, 0.0545], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:12:33,067 INFO [train.py:904] (3/8) Epoch 26, batch 8200, loss[loss=0.1699, simple_loss=0.2687, pruned_loss=0.03557, over 16906.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2817, pruned_loss=0.05498, over 3087279.77 frames. ], batch size: 90, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:43,207 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.684e+02 3.208e+02 4.137e+02 6.714e+02, threshold=6.416e+02, percent-clipped=2.0 2023-05-02 06:12:45,181 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5510, 3.8276, 3.9538, 2.6969, 3.5106, 3.9659, 3.5839, 2.1788], device='cuda:3'), covar=tensor([0.0538, 0.0093, 0.0067, 0.0410, 0.0121, 0.0120, 0.0108, 0.0512], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0134, 0.0100, 0.0113, 0.0097, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 06:13:12,977 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261977.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:13:58,686 INFO [train.py:904] (3/8) Epoch 26, batch 8250, loss[loss=0.1957, simple_loss=0.2941, pruned_loss=0.04863, over 16241.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2809, pruned_loss=0.05231, over 3077387.78 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:14:25,638 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 06:14:57,033 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:15:03,468 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-02 06:15:22,683 INFO [train.py:904] (3/8) Epoch 26, batch 8300, loss[loss=0.1707, simple_loss=0.2548, pruned_loss=0.04329, over 12048.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2785, pruned_loss=0.0495, over 3081517.17 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:15:32,832 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.264e+02 2.806e+02 3.280e+02 5.175e+02, threshold=5.611e+02, percent-clipped=0.0 2023-05-02 06:16:44,224 INFO [train.py:904] (3/8) Epoch 26, batch 8350, loss[loss=0.1989, simple_loss=0.2792, pruned_loss=0.0593, over 12073.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2786, pruned_loss=0.04821, over 3071878.88 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:16:55,844 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5495, 1.8349, 2.1417, 2.5782, 2.4509, 2.9345, 1.9225, 2.8643], device='cuda:3'), covar=tensor([0.0268, 0.0588, 0.0440, 0.0350, 0.0437, 0.0217, 0.0653, 0.0167], device='cuda:3'), in_proj_covar=tensor([0.0192, 0.0195, 0.0182, 0.0187, 0.0203, 0.0161, 0.0200, 0.0160], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:16:58,993 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:17:08,696 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0347, 1.8468, 1.6392, 1.4456, 1.9563, 1.6366, 1.5282, 1.9531], device='cuda:3'), covar=tensor([0.0263, 0.0352, 0.0491, 0.0459, 0.0283, 0.0349, 0.0201, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0220, 0.0236, 0.0228, 0.0228, 0.0238, 0.0236, 0.0236, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:18:04,036 INFO [train.py:904] (3/8) Epoch 26, batch 8400, loss[loss=0.1838, simple_loss=0.2806, pruned_loss=0.04351, over 16656.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.276, pruned_loss=0.04618, over 3056766.59 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:18:13,164 INFO [optim.py:368] (3/8) 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 06:18:14,933 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:18:47,907 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262180.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:02,233 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:16,721 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-05-02 06:19:24,256 INFO [train.py:904] (3/8) Epoch 26, batch 8450, loss[loss=0.1661, simple_loss=0.2664, pruned_loss=0.03286, over 16662.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2743, pruned_loss=0.0445, over 3065228.68 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:14,962 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2229, 1.6816, 2.0302, 2.1916, 2.3772, 2.5429, 1.9868, 2.4209], device='cuda:3'), covar=tensor([0.0272, 0.0548, 0.0343, 0.0390, 0.0326, 0.0223, 0.0502, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0186, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:20:48,803 INFO [train.py:904] (3/8) Epoch 26, batch 8500, loss[loss=0.153, simple_loss=0.2557, pruned_loss=0.02512, over 16828.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2706, pruned_loss=0.04247, over 3050404.50 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:57,922 INFO [optim.py:368] (3/8) 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 06:22:02,414 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9636, 2.7263, 2.6038, 1.9746, 2.5809, 2.7459, 2.5953, 1.9525], device='cuda:3'), covar=tensor([0.0430, 0.0097, 0.0081, 0.0374, 0.0150, 0.0122, 0.0114, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0086, 0.0087, 0.0132, 0.0099, 0.0111, 0.0096, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 06:22:09,819 INFO [train.py:904] (3/8) Epoch 26, batch 8550, loss[loss=0.1616, simple_loss=0.2629, pruned_loss=0.03017, over 16915.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2686, pruned_loss=0.04158, over 3042253.74 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:07,970 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 8600, loss[loss=0.1692, simple_loss=0.2696, pruned_loss=0.03439, over 16694.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2686, pruned_loss=0.04048, over 3052796.43 frames. ], batch size: 89, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:59,713 INFO [optim.py:368] (3/8) 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 06:24:13,967 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6943, 2.6525, 1.8810, 2.8268, 2.1537, 2.8407, 2.0859, 2.4514], device='cuda:3'), covar=tensor([0.0328, 0.0367, 0.1233, 0.0300, 0.0671, 0.0546, 0.1350, 0.0578], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0176, 0.0193, 0.0166, 0.0176, 0.0214, 0.0202, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 06:24:22,581 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-05-02 06:25:16,225 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 06:25:27,334 INFO [train.py:904] (3/8) Epoch 26, batch 8650, loss[loss=0.1635, simple_loss=0.2544, pruned_loss=0.03624, over 11998.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2668, pruned_loss=0.03901, over 3046547.39 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:25:34,888 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7668, 3.7476, 3.8946, 3.5895, 3.8767, 4.2651, 3.9196, 3.5951], device='cuda:3'), covar=tensor([0.2374, 0.2653, 0.2476, 0.2668, 0.2709, 0.1797, 0.1541, 0.2537], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0608, 0.0671, 0.0497, 0.0659, 0.0698, 0.0526, 0.0667], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:27:09,693 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 06:27:13,544 INFO [train.py:904] (3/8) Epoch 26, batch 8700, loss[loss=0.1711, simple_loss=0.2622, pruned_loss=0.03996, over 16814.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2645, pruned_loss=0.038, over 3074520.91 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:27:25,316 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 2.080e+02 2.583e+02 3.221e+02 6.009e+02, threshold=5.165e+02, percent-clipped=1.0 2023-05-02 06:28:02,764 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:20,330 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 8750, loss[loss=0.1927, simple_loss=0.2951, pruned_loss=0.0452, over 15335.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2646, pruned_loss=0.0376, over 3069572.64 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:29:48,636 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:09,753 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:09,901 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5597, 3.5460, 3.5221, 2.7453, 3.3614, 2.0347, 3.0866, 2.8351], device='cuda:3'), covar=tensor([0.0147, 0.0126, 0.0179, 0.0197, 0.0104, 0.2581, 0.0150, 0.0238], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0165, 0.0202, 0.0177, 0.0179, 0.0209, 0.0190, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:30:41,540 INFO [train.py:904] (3/8) Epoch 26, batch 8800, loss[loss=0.1902, simple_loss=0.285, pruned_loss=0.04775, over 16751.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2622, pruned_loss=0.03633, over 3067777.97 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:30:52,411 INFO [optim.py:368] (3/8) 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 06:32:25,981 INFO [train.py:904] (3/8) Epoch 26, batch 8850, loss[loss=0.1542, simple_loss=0.2618, pruned_loss=0.02328, over 15447.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2655, pruned_loss=0.03567, over 3083805.17 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:33:05,908 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9346, 2.1674, 2.3158, 2.9559, 1.7932, 3.2513, 1.7511, 2.7293], device='cuda:3'), covar=tensor([0.1291, 0.0795, 0.1093, 0.0160, 0.0090, 0.0334, 0.1532, 0.0747], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0197, 0.0194, 0.0203, 0.0215, 0.0206, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 06:33:32,325 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262633.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:33:46,305 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-05-02 06:34:13,275 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6988, 2.6302, 1.8872, 2.7917, 2.1142, 2.8415, 2.1378, 2.4543], device='cuda:3'), covar=tensor([0.0304, 0.0362, 0.1280, 0.0339, 0.0698, 0.0505, 0.1311, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0165, 0.0175, 0.0212, 0.0201, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 06:34:13,868 INFO [train.py:904] (3/8) Epoch 26, batch 8900, loss[loss=0.1491, simple_loss=0.2441, pruned_loss=0.02707, over 12487.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2658, pruned_loss=0.0353, over 3087606.51 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:34:26,825 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 2.190e+02 2.649e+02 3.221e+02 8.286e+02, threshold=5.298e+02, percent-clipped=4.0 2023-05-02 06:35:20,125 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262681.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:36:18,596 INFO [train.py:904] (3/8) Epoch 26, batch 8950, loss[loss=0.1411, simple_loss=0.2438, pruned_loss=0.01919, over 16725.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2651, pruned_loss=0.03567, over 3081222.19 frames. ], batch size: 76, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:37:49,441 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6499, 2.6197, 1.9288, 2.7875, 2.1374, 2.8180, 2.1040, 2.4300], device='cuda:3'), covar=tensor([0.0322, 0.0391, 0.1295, 0.0237, 0.0722, 0.0531, 0.1374, 0.0613], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0176, 0.0193, 0.0165, 0.0176, 0.0213, 0.0202, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 06:38:00,845 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=262749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:38:04,710 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 06:38:08,102 INFO [train.py:904] (3/8) Epoch 26, batch 9000, loss[loss=0.1397, simple_loss=0.2381, pruned_loss=0.02058, over 16259.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2622, pruned_loss=0.03473, over 3089325.36 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:08,102 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 06:38:17,574 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4542, 3.2827, 2.8231, 2.3471, 2.2359, 2.3935, 3.2351, 2.8941], device='cuda:3'), covar=tensor([0.2898, 0.0683, 0.1822, 0.2988, 0.3060, 0.2331, 0.0393, 0.1696], device='cuda:3'), in_proj_covar=tensor([0.0327, 0.0269, 0.0304, 0.0316, 0.0295, 0.0267, 0.0296, 0.0340], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:38:18,518 INFO [train.py:938] (3/8) Epoch 26, validation: loss=0.1435, simple_loss=0.2475, pruned_loss=0.01976, over 944034.00 frames. 2023-05-02 06:38:18,518 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 06:38:30,746 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.045e+02 2.519e+02 3.077e+02 5.140e+02, threshold=5.037e+02, percent-clipped=0.0 2023-05-02 06:40:03,318 INFO [train.py:904] (3/8) Epoch 26, batch 9050, loss[loss=0.1642, simple_loss=0.2614, pruned_loss=0.03349, over 12242.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2627, pruned_loss=0.03522, over 3076641.98 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:40:18,576 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 06:40:20,432 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:41:48,706 INFO [train.py:904] (3/8) Epoch 26, batch 9100, loss[loss=0.1709, simple_loss=0.258, pruned_loss=0.04185, over 12343.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2619, pruned_loss=0.03547, over 3075734.08 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:42:01,223 INFO [optim.py:368] (3/8) 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 06:43:32,736 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8554, 5.1313, 5.2155, 4.9788, 5.1347, 5.5862, 5.0461, 4.7290], device='cuda:3'), covar=tensor([0.0921, 0.1598, 0.1866, 0.1869, 0.1969, 0.0743, 0.1380, 0.2306], device='cuda:3'), in_proj_covar=tensor([0.0406, 0.0601, 0.0663, 0.0491, 0.0650, 0.0687, 0.0519, 0.0658], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:43:45,862 INFO [train.py:904] (3/8) Epoch 26, batch 9150, loss[loss=0.154, simple_loss=0.2511, pruned_loss=0.0285, over 16963.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2622, pruned_loss=0.03518, over 3053978.87 frames. ], batch size: 109, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:44:00,834 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5322, 3.9115, 3.9773, 2.5837, 3.6047, 3.9690, 3.7253, 2.3011], device='cuda:3'), covar=tensor([0.0530, 0.0062, 0.0046, 0.0441, 0.0105, 0.0087, 0.0078, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0085, 0.0086, 0.0131, 0.0098, 0.0109, 0.0094, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:45:08,675 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4706, 3.4207, 3.5324, 3.5806, 3.6394, 3.3641, 3.6169, 3.6812], device='cuda:3'), covar=tensor([0.1193, 0.0830, 0.0928, 0.0604, 0.0581, 0.2156, 0.0869, 0.0796], device='cuda:3'), in_proj_covar=tensor([0.0628, 0.0778, 0.0893, 0.0788, 0.0600, 0.0624, 0.0652, 0.0764], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:45:27,072 INFO [train.py:904] (3/8) Epoch 26, batch 9200, loss[loss=0.1716, simple_loss=0.2669, pruned_loss=0.0382, over 16234.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2577, pruned_loss=0.03418, over 3051553.54 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:45:35,746 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3909, 4.6897, 4.4774, 4.4875, 4.1944, 4.2208, 4.1148, 4.7322], device='cuda:3'), covar=tensor([0.1123, 0.0926, 0.1067, 0.0848, 0.0863, 0.1556, 0.1328, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0685, 0.0824, 0.0677, 0.0633, 0.0526, 0.0527, 0.0695, 0.0647], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:45:36,599 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 26, batch 9250, loss[loss=0.1565, simple_loss=0.2497, pruned_loss=0.03168, over 16325.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2572, pruned_loss=0.03415, over 3041485.30 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:47:31,504 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8493, 3.8478, 4.1446, 4.1336, 4.1284, 3.9086, 3.9138, 3.9500], device='cuda:3'), covar=tensor([0.0423, 0.0883, 0.0648, 0.0770, 0.0856, 0.0811, 0.0987, 0.0546], device='cuda:3'), in_proj_covar=tensor([0.0411, 0.0466, 0.0453, 0.0415, 0.0501, 0.0477, 0.0548, 0.0381], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 06:48:49,716 INFO [train.py:904] (3/8) Epoch 26, batch 9300, loss[loss=0.1454, simple_loss=0.2452, pruned_loss=0.02283, over 16437.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2559, pruned_loss=0.03371, over 3033483.06 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:49:02,021 INFO [optim.py:368] (3/8) 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 06:49:35,687 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7668, 4.0713, 2.9682, 2.3399, 2.4709, 2.5528, 4.3422, 3.3490], device='cuda:3'), covar=tensor([0.3049, 0.0523, 0.1922, 0.3134, 0.3151, 0.2237, 0.0366, 0.1469], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0267, 0.0304, 0.0316, 0.0294, 0.0267, 0.0296, 0.0339], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:50:20,271 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 9350, loss[loss=0.1673, simple_loss=0.2604, pruned_loss=0.03705, over 16746.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2561, pruned_loss=0.03375, over 3035542.54 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:50:38,185 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263105.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:51,050 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:52,970 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9678, 4.2671, 4.1154, 4.1313, 3.7542, 3.8559, 3.8792, 4.2743], device='cuda:3'), covar=tensor([0.1126, 0.0957, 0.0904, 0.0745, 0.0829, 0.1763, 0.0965, 0.0974], device='cuda:3'), in_proj_covar=tensor([0.0684, 0.0822, 0.0673, 0.0631, 0.0524, 0.0526, 0.0691, 0.0645], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:51:00,416 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9548, 2.2167, 2.4311, 3.2305, 2.1974, 2.4112, 2.3564, 2.2835], device='cuda:3'), covar=tensor([0.1364, 0.3601, 0.2756, 0.0756, 0.4622, 0.2654, 0.3619, 0.3727], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0458, 0.0376, 0.0324, 0.0437, 0.0522, 0.0430, 0.0535], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 06:51:03,892 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-05-02 06:51:48,031 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:14,166 INFO [train.py:904] (3/8) Epoch 26, batch 9400, loss[loss=0.183, simple_loss=0.2866, pruned_loss=0.03968, over 16831.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2563, pruned_loss=0.03363, over 3030506.22 frames. ], batch size: 124, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:52:19,855 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:25,027 INFO [optim.py:368] (3/8) 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 06:52:52,028 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:50,923 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:53,493 INFO [train.py:904] (3/8) Epoch 26, batch 9450, loss[loss=0.1647, simple_loss=0.2578, pruned_loss=0.0358, over 12726.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2579, pruned_loss=0.03383, over 3018399.84 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:54:04,925 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3894, 3.1243, 3.2864, 1.8349, 3.4662, 3.5639, 2.9013, 2.8095], device='cuda:3'), covar=tensor([0.0758, 0.0288, 0.0218, 0.1248, 0.0101, 0.0184, 0.0438, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0107, 0.0095, 0.0135, 0.0082, 0.0123, 0.0126, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 06:54:14,256 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:55:19,948 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 9500, loss[loss=0.1325, simple_loss=0.2214, pruned_loss=0.02184, over 12909.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2571, pruned_loss=0.03328, over 3047714.08 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:55:47,501 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 2.048e+02 2.318e+02 3.052e+02 5.561e+02, threshold=4.636e+02, percent-clipped=2.0 2023-05-02 06:56:19,475 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:56:25,133 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2388, 4.3241, 4.4603, 4.2299, 4.4021, 4.8516, 4.3950, 4.0275], device='cuda:3'), covar=tensor([0.1681, 0.2116, 0.2419, 0.2116, 0.2293, 0.0876, 0.1603, 0.2682], device='cuda:3'), in_proj_covar=tensor([0.0401, 0.0594, 0.0658, 0.0484, 0.0644, 0.0680, 0.0513, 0.0650], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 06:57:17,617 INFO [train.py:904] (3/8) Epoch 26, batch 9550, loss[loss=0.1851, simple_loss=0.2857, pruned_loss=0.04224, over 16670.00 frames. ], tot_loss[loss=0.162, simple_loss=0.257, pruned_loss=0.03347, over 3052058.07 frames. ], batch size: 76, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:57:27,280 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:57:50,689 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0705, 2.8980, 2.9598, 1.7345, 3.1220, 3.3614, 2.8227, 2.5342], device='cuda:3'), covar=tensor([0.1028, 0.0230, 0.0223, 0.1360, 0.0127, 0.0194, 0.0455, 0.0595], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0107, 0.0095, 0.0135, 0.0082, 0.0123, 0.0125, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 06:58:42,335 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 06:58:43,692 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4383, 3.7617, 3.8151, 2.5026, 3.3730, 3.7941, 3.6068, 2.0411], device='cuda:3'), covar=tensor([0.0526, 0.0057, 0.0053, 0.0435, 0.0128, 0.0098, 0.0081, 0.0537], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0086, 0.0087, 0.0133, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 06:58:58,571 INFO [train.py:904] (3/8) Epoch 26, batch 9600, loss[loss=0.1659, simple_loss=0.255, pruned_loss=0.03844, over 12363.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2582, pruned_loss=0.03402, over 3058495.38 frames. ], batch size: 246, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:59:09,908 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.249e+02 2.699e+02 3.019e+02 5.664e+02, threshold=5.399e+02, percent-clipped=4.0 2023-05-02 07:00:45,647 INFO [train.py:904] (3/8) Epoch 26, batch 9650, loss[loss=0.1758, simple_loss=0.2656, pruned_loss=0.04301, over 16911.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2602, pruned_loss=0.03465, over 3042965.94 frames. ], batch size: 116, lr: 2.56e-03, grad_scale: 8.0 2023-05-02 07:00:52,284 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3976, 3.4721, 3.6684, 3.6618, 3.6627, 3.4964, 3.5189, 3.5350], device='cuda:3'), covar=tensor([0.0401, 0.0820, 0.0500, 0.0476, 0.0562, 0.0568, 0.0788, 0.0523], device='cuda:3'), in_proj_covar=tensor([0.0407, 0.0461, 0.0447, 0.0412, 0.0497, 0.0472, 0.0542, 0.0376], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 07:00:52,306 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:01:39,884 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0703, 3.0706, 1.8408, 3.2952, 2.3097, 3.2900, 2.1179, 2.6235], device='cuda:3'), covar=tensor([0.0334, 0.0408, 0.1788, 0.0288, 0.0891, 0.0605, 0.1658, 0.0762], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0173, 0.0190, 0.0163, 0.0174, 0.0210, 0.0199, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:02:02,088 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8341, 1.4109, 1.7899, 1.7193, 1.8937, 1.9608, 1.6886, 1.8977], device='cuda:3'), covar=tensor([0.0285, 0.0484, 0.0265, 0.0350, 0.0326, 0.0249, 0.0499, 0.0172], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0192, 0.0179, 0.0183, 0.0199, 0.0157, 0.0197, 0.0157], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:02:28,499 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263451.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:31,833 INFO [train.py:904] (3/8) Epoch 26, batch 9700, loss[loss=0.1633, simple_loss=0.2597, pruned_loss=0.03342, over 16138.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.259, pruned_loss=0.03473, over 3031685.44 frames. ], batch size: 165, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:02:33,069 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:42,008 INFO [optim.py:368] (3/8) 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 07:02:58,830 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 26, batch 9750, loss[loss=0.166, simple_loss=0.2629, pruned_loss=0.03458, over 17030.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2586, pruned_loss=0.03517, over 3020340.19 frames. ], batch size: 109, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:05:49,270 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5694, 4.8559, 4.6732, 4.6824, 4.3667, 4.3577, 4.2961, 4.9377], device='cuda:3'), covar=tensor([0.1181, 0.0833, 0.0875, 0.0802, 0.0849, 0.1387, 0.1143, 0.0855], device='cuda:3'), in_proj_covar=tensor([0.0683, 0.0824, 0.0673, 0.0632, 0.0526, 0.0526, 0.0693, 0.0647], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:05:51,431 INFO [train.py:904] (3/8) Epoch 26, batch 9800, loss[loss=0.1741, simple_loss=0.2832, pruned_loss=0.0325, over 15410.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2587, pruned_loss=0.03431, over 3037518.02 frames. ], batch size: 190, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:06:03,263 INFO [optim.py:368] (3/8) 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 07:06:23,916 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:07:33,299 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 26, batch 9850, loss[loss=0.1725, simple_loss=0.268, pruned_loss=0.03848, over 16672.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2602, pruned_loss=0.03425, over 3046523.73 frames. ], batch size: 134, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:24,096 INFO [train.py:904] (3/8) Epoch 26, batch 9900, loss[loss=0.1612, simple_loss=0.2662, pruned_loss=0.02811, over 16283.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2604, pruned_loss=0.03402, over 3043441.55 frames. ], batch size: 165, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:36,802 INFO [optim.py:368] (3/8) 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 07:11:20,739 INFO [train.py:904] (3/8) Epoch 26, batch 9950, loss[loss=0.1691, simple_loss=0.2711, pruned_loss=0.03356, over 15356.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2628, pruned_loss=0.03423, over 3055367.12 frames. ], batch size: 191, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:14,892 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6080, 3.6782, 3.4866, 3.1666, 3.3134, 3.5703, 3.3589, 3.4146], device='cuda:3'), covar=tensor([0.0647, 0.1046, 0.0393, 0.0354, 0.0516, 0.0570, 0.1532, 0.0503], device='cuda:3'), in_proj_covar=tensor([0.0294, 0.0433, 0.0339, 0.0340, 0.0337, 0.0392, 0.0234, 0.0406], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:13:18,987 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:13:21,140 INFO [train.py:904] (3/8) Epoch 26, batch 10000, loss[loss=0.1576, simple_loss=0.2579, pruned_loss=0.02861, over 16294.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2614, pruned_loss=0.03402, over 3074674.18 frames. ], batch size: 146, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:34,912 INFO [optim.py:368] (3/8) 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 07:13:40,909 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-05-02 07:13:51,384 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:20,057 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0467, 2.3513, 2.1104, 2.1421, 2.6452, 2.3586, 2.5231, 2.8102], device='cuda:3'), covar=tensor([0.0178, 0.0495, 0.0580, 0.0537, 0.0353, 0.0474, 0.0257, 0.0323], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0235, 0.0226, 0.0227, 0.0236, 0.0235, 0.0231, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:14:49,003 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:53,764 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:00,032 INFO [train.py:904] (3/8) Epoch 26, batch 10050, loss[loss=0.1708, simple_loss=0.2695, pruned_loss=0.03607, over 16679.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2617, pruned_loss=0.03417, over 3071731.83 frames. ], batch size: 134, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:15:01,224 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 07:15:23,649 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263815.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:57,722 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263833.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:16:05,853 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-02 07:16:17,280 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:16:17,453 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3413, 2.9591, 3.0730, 1.9074, 3.2590, 3.3671, 2.7549, 2.6582], device='cuda:3'), covar=tensor([0.0717, 0.0285, 0.0223, 0.1154, 0.0112, 0.0182, 0.0466, 0.0469], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0106, 0.0094, 0.0134, 0.0081, 0.0122, 0.0125, 0.0124], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 07:16:20,153 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1862, 3.2180, 2.0750, 3.5265, 2.4512, 3.4759, 2.1584, 2.6971], device='cuda:3'), covar=tensor([0.0357, 0.0434, 0.1620, 0.0216, 0.0900, 0.0598, 0.1693, 0.0835], device='cuda:3'), in_proj_covar=tensor([0.0168, 0.0172, 0.0188, 0.0161, 0.0172, 0.0208, 0.0198, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:16:30,282 INFO [train.py:904] (3/8) Epoch 26, batch 10100, loss[loss=0.1594, simple_loss=0.2465, pruned_loss=0.0362, over 12985.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2612, pruned_loss=0.03392, over 3060831.75 frames. ], batch size: 250, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:16:39,584 INFO [optim.py:368] (3/8) 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 07:16:41,142 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0723, 2.2501, 1.9712, 2.0548, 2.6038, 2.2748, 2.4160, 2.7759], device='cuda:3'), covar=tensor([0.0170, 0.0571, 0.0661, 0.0631, 0.0381, 0.0527, 0.0232, 0.0343], device='cuda:3'), in_proj_covar=tensor([0.0215, 0.0235, 0.0226, 0.0227, 0.0236, 0.0234, 0.0231, 0.0231], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:17:00,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5010, 4.5683, 4.3926, 4.0507, 4.0986, 4.4805, 4.2760, 4.1908], device='cuda:3'), covar=tensor([0.0584, 0.0575, 0.0350, 0.0345, 0.0863, 0.0571, 0.0533, 0.0666], device='cuda:3'), in_proj_covar=tensor([0.0292, 0.0430, 0.0337, 0.0338, 0.0335, 0.0390, 0.0232, 0.0404], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:3') 2023-05-02 07:17:02,791 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:17:36,277 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263894.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:18:08,997 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:09,862 INFO [train.py:904] (3/8) Epoch 27, batch 0, loss[loss=0.1886, simple_loss=0.2866, pruned_loss=0.04536, over 17128.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2866, pruned_loss=0.04536, over 17128.00 frames. ], batch size: 48, lr: 2.51e-03, grad_scale: 16.0 2023-05-02 07:18:09,862 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 07:18:17,114 INFO [train.py:938] (3/8) Epoch 27, validation: loss=0.1434, simple_loss=0.2467, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 07:18:17,115 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 07:18:38,688 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:22,226 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:26,171 INFO [train.py:904] (3/8) Epoch 27, batch 50, loss[loss=0.1597, simple_loss=0.2475, pruned_loss=0.036, over 16741.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2676, pruned_loss=0.04536, over 752186.01 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:19:36,245 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:20:36,409 INFO [train.py:904] (3/8) Epoch 27, batch 100, loss[loss=0.1477, simple_loss=0.2371, pruned_loss=0.02915, over 17120.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2641, pruned_loss=0.04531, over 1317335.30 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:20:52,775 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7814, 4.3051, 2.9410, 2.3066, 2.6020, 2.5494, 4.6617, 3.4612], device='cuda:3'), covar=tensor([0.3108, 0.0605, 0.1991, 0.3202, 0.3260, 0.2198, 0.0374, 0.1722], device='cuda:3'), in_proj_covar=tensor([0.0326, 0.0268, 0.0306, 0.0318, 0.0294, 0.0268, 0.0297, 0.0341], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:21:01,008 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:21:42,922 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0880, 2.7303, 2.5407, 4.3929, 3.3870, 3.9722, 1.6263, 2.9162], device='cuda:3'), covar=tensor([0.1346, 0.0779, 0.1294, 0.0170, 0.0196, 0.0494, 0.1677, 0.0916], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0195, 0.0192, 0.0200, 0.0213, 0.0205, 0.0194], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:21:44,756 INFO [train.py:904] (3/8) Epoch 27, batch 150, loss[loss=0.1724, simple_loss=0.2742, pruned_loss=0.03528, over 16739.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.263, pruned_loss=0.04365, over 1758612.00 frames. ], batch size: 62, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:57,549 INFO [optim.py:368] (3/8) 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 07:22:23,031 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 07:22:53,632 INFO [train.py:904] (3/8) Epoch 27, batch 200, loss[loss=0.1746, simple_loss=0.2551, pruned_loss=0.04711, over 16639.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2629, pruned_loss=0.04442, over 2097657.58 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:24:00,835 INFO [train.py:904] (3/8) Epoch 27, batch 250, loss[loss=0.1655, simple_loss=0.2495, pruned_loss=0.04078, over 16796.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2599, pruned_loss=0.04324, over 2370685.11 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:24:14,023 INFO [optim.py:368] (3/8) 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 07:24:26,850 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 07:24:50,572 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264189.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:25:10,466 INFO [train.py:904] (3/8) Epoch 27, batch 300, loss[loss=0.1833, simple_loss=0.2551, pruned_loss=0.05577, over 16677.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2551, pruned_loss=0.04087, over 2588678.87 frames. ], batch size: 134, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:25:52,971 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:26:11,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7870, 4.2656, 3.1276, 2.3267, 2.6560, 2.5985, 4.6452, 3.4760], device='cuda:3'), covar=tensor([0.3042, 0.0624, 0.1779, 0.3153, 0.3020, 0.2217, 0.0354, 0.1584], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0271, 0.0310, 0.0321, 0.0298, 0.0272, 0.0301, 0.0346], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:26:19,419 INFO [train.py:904] (3/8) Epoch 27, batch 350, loss[loss=0.1492, simple_loss=0.2344, pruned_loss=0.03198, over 17198.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2533, pruned_loss=0.03953, over 2750838.31 frames. ], batch size: 44, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:26:34,331 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.070e+02 2.475e+02 2.951e+02 5.112e+02, threshold=4.951e+02, percent-clipped=0.0 2023-05-02 07:26:45,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4093, 2.3666, 2.3642, 4.2925, 2.2940, 2.7068, 2.4394, 2.5476], device='cuda:3'), covar=tensor([0.1418, 0.3930, 0.3392, 0.0564, 0.4449, 0.2738, 0.3805, 0.3770], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0467, 0.0383, 0.0331, 0.0443, 0.0531, 0.0438, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:27:03,081 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3258, 4.0484, 4.4659, 2.4811, 4.6857, 4.7880, 3.5322, 3.6320], device='cuda:3'), covar=tensor([0.0654, 0.0273, 0.0241, 0.1170, 0.0090, 0.0187, 0.0448, 0.0424], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0110, 0.0098, 0.0139, 0.0085, 0.0128, 0.0129, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 07:27:17,377 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:27,043 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3014, 5.2479, 4.9424, 4.5713, 5.0820, 2.0407, 4.8280, 4.8426], device='cuda:3'), covar=tensor([0.0102, 0.0091, 0.0269, 0.0378, 0.0111, 0.2785, 0.0153, 0.0257], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0168, 0.0204, 0.0178, 0.0181, 0.0214, 0.0194, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:27:27,689 INFO [train.py:904] (3/8) Epoch 27, batch 400, loss[loss=0.1684, simple_loss=0.2529, pruned_loss=0.04197, over 17235.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2518, pruned_loss=0.03917, over 2880214.59 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:27:44,538 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264316.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:59,280 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7534, 2.7899, 2.4109, 2.6281, 3.0645, 2.8933, 3.2991, 3.2563], device='cuda:3'), covar=tensor([0.0162, 0.0489, 0.0583, 0.0500, 0.0341, 0.0408, 0.0313, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0243, 0.0233, 0.0233, 0.0243, 0.0242, 0.0240, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:28:33,908 INFO [train.py:904] (3/8) Epoch 27, batch 450, loss[loss=0.1461, simple_loss=0.235, pruned_loss=0.02858, over 17206.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2504, pruned_loss=0.03905, over 2975043.92 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:28:48,387 INFO [optim.py:368] (3/8) 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 07:29:03,240 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-05-02 07:29:44,911 INFO [train.py:904] (3/8) Epoch 27, batch 500, loss[loss=0.1478, simple_loss=0.2466, pruned_loss=0.02449, over 17242.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2498, pruned_loss=0.0384, over 3060294.16 frames. ], batch size: 52, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:30:40,868 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5806, 5.5745, 5.4745, 4.9534, 5.0775, 5.4627, 5.4194, 5.1437], device='cuda:3'), covar=tensor([0.0580, 0.0564, 0.0292, 0.0328, 0.1047, 0.0509, 0.0245, 0.0763], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0456, 0.0356, 0.0358, 0.0355, 0.0413, 0.0245, 0.0428], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:30:51,105 INFO [train.py:904] (3/8) Epoch 27, batch 550, loss[loss=0.1916, simple_loss=0.2676, pruned_loss=0.05778, over 16713.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2494, pruned_loss=0.03828, over 3117851.94 frames. ], batch size: 134, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:31:04,228 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.059e+02 2.347e+02 2.787e+02 4.742e+02, threshold=4.693e+02, percent-clipped=0.0 2023-05-02 07:31:26,697 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264479.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:31:41,321 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264489.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:31:55,520 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264500.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:31:59,347 INFO [train.py:904] (3/8) Epoch 27, batch 600, loss[loss=0.1381, simple_loss=0.2173, pruned_loss=0.02948, over 15939.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2488, pruned_loss=0.03824, over 3158271.58 frames. ], batch size: 35, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:32:29,861 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 07:32:46,462 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264537.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:32:50,477 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:33:08,799 INFO [train.py:904] (3/8) Epoch 27, batch 650, loss[loss=0.1722, simple_loss=0.2507, pruned_loss=0.04687, over 16890.00 frames. ], tot_loss[loss=0.161, simple_loss=0.247, pruned_loss=0.03751, over 3194174.56 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:33:18,819 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264561.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:33:20,707 INFO [optim.py:368] (3/8) 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 07:33:58,956 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:34:15,997 INFO [train.py:904] (3/8) Epoch 27, batch 700, loss[loss=0.1447, simple_loss=0.2268, pruned_loss=0.0313, over 16783.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2468, pruned_loss=0.03767, over 3214918.93 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:34:31,887 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-02 07:34:34,278 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264616.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:34:40,629 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9989, 5.0422, 5.4786, 5.4524, 5.4396, 5.1126, 5.0621, 4.9453], device='cuda:3'), covar=tensor([0.0363, 0.0637, 0.0426, 0.0436, 0.0557, 0.0465, 0.0962, 0.0493], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0484, 0.0468, 0.0430, 0.0518, 0.0496, 0.0568, 0.0394], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 07:34:41,742 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8315, 3.8872, 4.1225, 4.1126, 4.1306, 3.8913, 3.9375, 3.9342], device='cuda:3'), covar=tensor([0.0439, 0.0840, 0.0538, 0.0460, 0.0613, 0.0605, 0.0819, 0.0606], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0484, 0.0468, 0.0430, 0.0518, 0.0496, 0.0568, 0.0394], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 07:34:42,766 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4013, 4.4344, 4.5629, 4.3524, 4.4533, 5.0178, 4.5197, 4.1756], device='cuda:3'), covar=tensor([0.1700, 0.2258, 0.2925, 0.2455, 0.2953, 0.1170, 0.1850, 0.2861], device='cuda:3'), in_proj_covar=tensor([0.0419, 0.0619, 0.0688, 0.0508, 0.0672, 0.0712, 0.0535, 0.0676], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:35:22,118 INFO [train.py:904] (3/8) Epoch 27, batch 750, loss[loss=0.1446, simple_loss=0.2312, pruned_loss=0.02903, over 16481.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2469, pruned_loss=0.03768, over 3241615.00 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:35:35,422 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.184e+02 2.458e+02 2.836e+02 5.785e+02, threshold=4.916e+02, percent-clipped=2.0 2023-05-02 07:35:37,451 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264664.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:35:53,482 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-05-02 07:36:29,180 INFO [train.py:904] (3/8) Epoch 27, batch 800, loss[loss=0.1684, simple_loss=0.2453, pruned_loss=0.04573, over 16853.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2477, pruned_loss=0.03779, over 3267404.64 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:36,919 INFO [train.py:904] (3/8) Epoch 27, batch 850, loss[loss=0.1857, simple_loss=0.2804, pruned_loss=0.04544, over 17072.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2469, pruned_loss=0.03747, over 3268491.75 frames. ], batch size: 55, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:51,806 INFO [optim.py:368] (3/8) 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 07:37:53,965 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7477, 4.7064, 4.6592, 4.2798, 4.3201, 4.6545, 4.4938, 4.4029], device='cuda:3'), covar=tensor([0.0679, 0.0880, 0.0395, 0.0404, 0.0991, 0.0582, 0.0485, 0.0784], device='cuda:3'), in_proj_covar=tensor([0.0314, 0.0465, 0.0363, 0.0366, 0.0362, 0.0422, 0.0250, 0.0437], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:38:44,877 INFO [train.py:904] (3/8) Epoch 27, batch 900, loss[loss=0.1481, simple_loss=0.2426, pruned_loss=0.02677, over 16681.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2457, pruned_loss=0.03673, over 3286048.36 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:38:47,145 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:02,529 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9620, 4.0116, 2.9455, 4.7991, 3.4702, 4.7846, 2.8490, 3.5352], device='cuda:3'), covar=tensor([0.0342, 0.0440, 0.1377, 0.0345, 0.0710, 0.0445, 0.1436, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0182, 0.0198, 0.0174, 0.0182, 0.0222, 0.0208, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:39:28,075 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 950, loss[loss=0.1573, simple_loss=0.2403, pruned_loss=0.03719, over 16736.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2455, pruned_loss=0.03656, over 3295956.47 frames. ], batch size: 83, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:39:55,586 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264856.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:40:04,148 INFO [optim.py:368] (3/8) 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 07:40:07,742 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264865.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:34,592 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6120, 3.2104, 3.6130, 2.0609, 3.7206, 3.7165, 3.0752, 2.8160], device='cuda:3'), covar=tensor([0.0796, 0.0316, 0.0233, 0.1178, 0.0122, 0.0262, 0.0456, 0.0497], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0140, 0.0086, 0.0131, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:40:41,673 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264890.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:57,191 INFO [train.py:904] (3/8) Epoch 27, batch 1000, loss[loss=0.1688, simple_loss=0.245, pruned_loss=0.04634, over 16525.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2451, pruned_loss=0.03679, over 3308126.14 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:41:46,841 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:42:07,407 INFO [train.py:904] (3/8) Epoch 27, batch 1050, loss[loss=0.1735, simple_loss=0.2552, pruned_loss=0.04592, over 16677.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2446, pruned_loss=0.03648, over 3310612.37 frames. ], batch size: 76, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:42:19,712 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.065e+02 2.389e+02 2.901e+02 6.221e+02, threshold=4.777e+02, percent-clipped=3.0 2023-05-02 07:42:22,671 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 2023-05-02 07:42:44,607 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 07:43:16,444 INFO [train.py:904] (3/8) Epoch 27, batch 1100, loss[loss=0.1602, simple_loss=0.2507, pruned_loss=0.03484, over 16707.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2446, pruned_loss=0.03685, over 3313994.73 frames. ], batch size: 62, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:43:52,934 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 07:44:10,174 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5831, 3.6458, 3.3622, 3.0155, 3.2327, 3.5178, 3.3088, 3.3848], device='cuda:3'), covar=tensor([0.0564, 0.0655, 0.0315, 0.0313, 0.0566, 0.0489, 0.1418, 0.0497], device='cuda:3'), in_proj_covar=tensor([0.0318, 0.0472, 0.0368, 0.0371, 0.0368, 0.0429, 0.0253, 0.0442], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:44:16,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6231, 2.5509, 1.9527, 2.7004, 2.1995, 2.7552, 2.1959, 2.4084], device='cuda:3'), covar=tensor([0.0334, 0.0390, 0.1278, 0.0318, 0.0639, 0.0505, 0.1231, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0182, 0.0197, 0.0174, 0.0181, 0.0222, 0.0207, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:44:25,472 INFO [train.py:904] (3/8) Epoch 27, batch 1150, loss[loss=0.1584, simple_loss=0.2325, pruned_loss=0.04211, over 16759.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2446, pruned_loss=0.03677, over 3311654.11 frames. ], batch size: 124, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:44:39,263 INFO [optim.py:368] (3/8) 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 07:45:34,347 INFO [train.py:904] (3/8) Epoch 27, batch 1200, loss[loss=0.156, simple_loss=0.2488, pruned_loss=0.03165, over 17194.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2441, pruned_loss=0.03625, over 3311722.48 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:00,651 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8197, 4.1306, 2.9688, 4.7333, 3.3444, 4.6499, 3.0838, 3.5613], device='cuda:3'), covar=tensor([0.0396, 0.0416, 0.1434, 0.0276, 0.0796, 0.0610, 0.1306, 0.0694], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0175, 0.0182, 0.0223, 0.0208, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:46:19,365 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 1250, loss[loss=0.1596, simple_loss=0.2562, pruned_loss=0.0315, over 17123.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2441, pruned_loss=0.03643, over 3312887.63 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:48,316 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265156.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:46:53,475 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 07:47:08,018 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 07:47:13,195 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8877, 2.2454, 2.4537, 3.1012, 2.2678, 2.4053, 2.4159, 2.3548], device='cuda:3'), covar=tensor([0.1463, 0.3396, 0.2992, 0.0885, 0.4296, 0.2498, 0.3392, 0.3672], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0473, 0.0387, 0.0337, 0.0447, 0.0540, 0.0444, 0.0552], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:47:25,262 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265183.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:47:37,872 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8959, 4.4526, 3.0702, 2.4689, 2.9588, 2.5344, 4.8032, 3.6968], device='cuda:3'), covar=tensor([0.3100, 0.0660, 0.2024, 0.2858, 0.2887, 0.2411, 0.0378, 0.1432], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0274, 0.0311, 0.0325, 0.0302, 0.0274, 0.0303, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:47:49,592 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 07:47:50,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1059, 5.4443, 5.2210, 5.2091, 4.9764, 4.9025, 4.8446, 5.5586], device='cuda:3'), covar=tensor([0.1411, 0.0914, 0.1097, 0.0917, 0.0857, 0.1036, 0.1446, 0.0863], device='cuda:3'), in_proj_covar=tensor([0.0722, 0.0869, 0.0711, 0.0670, 0.0553, 0.0550, 0.0737, 0.0682], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:47:53,012 INFO [train.py:904] (3/8) Epoch 27, batch 1300, loss[loss=0.1553, simple_loss=0.2529, pruned_loss=0.02886, over 17299.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2447, pruned_loss=0.03671, over 3311532.41 frames. ], batch size: 52, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:47:54,314 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265204.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:48:31,754 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6762, 4.3771, 4.3614, 4.8374, 5.0273, 4.5880, 4.9431, 5.0166], device='cuda:3'), covar=tensor([0.2129, 0.1632, 0.2744, 0.1164, 0.0886, 0.1445, 0.1352, 0.1185], device='cuda:3'), in_proj_covar=tensor([0.0683, 0.0841, 0.0971, 0.0850, 0.0645, 0.0670, 0.0708, 0.0821], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 07:48:48,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8464, 4.3584, 3.0602, 2.4232, 2.6651, 2.6271, 4.6699, 3.5856], device='cuda:3'), covar=tensor([0.2979, 0.0606, 0.1891, 0.2969, 0.3082, 0.2226, 0.0371, 0.1507], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0326, 0.0303, 0.0275, 0.0304, 0.0351], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:49:00,603 INFO [train.py:904] (3/8) Epoch 27, batch 1350, loss[loss=0.1579, simple_loss=0.2525, pruned_loss=0.03167, over 17042.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.245, pruned_loss=0.03626, over 3319027.20 frames. ], batch size: 53, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:49:14,441 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.174e+02 2.461e+02 3.019e+02 8.065e+02, threshold=4.923e+02, percent-clipped=2.0 2023-05-02 07:49:35,233 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:50:07,766 INFO [train.py:904] (3/8) Epoch 27, batch 1400, loss[loss=0.1435, simple_loss=0.2244, pruned_loss=0.03133, over 16845.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2454, pruned_loss=0.03692, over 3321991.12 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:50:56,845 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:51:15,553 INFO [train.py:904] (3/8) Epoch 27, batch 1450, loss[loss=0.1421, simple_loss=0.2375, pruned_loss=0.02333, over 17247.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2441, pruned_loss=0.03589, over 3318054.97 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:51:29,669 INFO [optim.py:368] (3/8) 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 07:51:30,570 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-02 07:52:19,277 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 07:52:24,303 INFO [train.py:904] (3/8) Epoch 27, batch 1500, loss[loss=0.16, simple_loss=0.259, pruned_loss=0.03051, over 17075.00 frames. ], tot_loss[loss=0.158, simple_loss=0.244, pruned_loss=0.036, over 3321544.91 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:36,985 INFO [train.py:904] (3/8) Epoch 27, batch 1550, loss[loss=0.1804, simple_loss=0.2542, pruned_loss=0.05328, over 16906.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2454, pruned_loss=0.03747, over 3315769.49 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:45,899 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265460.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:53:49,832 INFO [optim.py:368] (3/8) 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 07:54:43,792 INFO [train.py:904] (3/8) Epoch 27, batch 1600, loss[loss=0.1557, simple_loss=0.2519, pruned_loss=0.02976, over 17106.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2472, pruned_loss=0.03801, over 3315639.79 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:54:51,418 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:10,326 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 07:55:21,295 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:51,085 INFO [train.py:904] (3/8) Epoch 27, batch 1650, loss[loss=0.1794, simple_loss=0.2747, pruned_loss=0.04202, over 17038.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.248, pruned_loss=0.0385, over 3319514.71 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:56:04,414 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.682e+02 2.309e+02 2.667e+02 3.273e+02 6.043e+02, threshold=5.334e+02, percent-clipped=4.0 2023-05-02 07:56:43,892 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:56:49,363 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6040, 4.6836, 4.8717, 4.6759, 4.7518, 5.3178, 4.7933, 4.4773], device='cuda:3'), covar=tensor([0.1575, 0.2354, 0.2475, 0.2120, 0.2601, 0.1138, 0.1818, 0.2604], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0637, 0.0705, 0.0523, 0.0693, 0.0730, 0.0549, 0.0696], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:57:00,225 INFO [train.py:904] (3/8) Epoch 27, batch 1700, loss[loss=0.1464, simple_loss=0.2265, pruned_loss=0.0332, over 16946.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2495, pruned_loss=0.03904, over 3322822.60 frames. ], batch size: 41, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:57:02,163 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-02 07:57:16,009 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:34,322 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0580, 3.0538, 1.7422, 3.2198, 2.3810, 3.2350, 1.9558, 2.5449], device='cuda:3'), covar=tensor([0.0372, 0.0420, 0.1963, 0.0377, 0.0868, 0.0664, 0.1822, 0.0822], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0175, 0.0182, 0.0223, 0.0207, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 07:57:42,660 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265634.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:58:07,543 INFO [train.py:904] (3/8) Epoch 27, batch 1750, loss[loss=0.1553, simple_loss=0.2332, pruned_loss=0.03872, over 16703.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2503, pruned_loss=0.03902, over 3317573.98 frames. ], batch size: 89, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:58:22,124 INFO [optim.py:368] (3/8) 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 07:58:38,370 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265675.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:58:39,504 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3870, 5.3875, 5.2924, 4.8325, 4.8302, 5.3129, 5.2714, 4.9080], device='cuda:3'), covar=tensor([0.0712, 0.0777, 0.0398, 0.0448, 0.1286, 0.0587, 0.0373, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0477, 0.0371, 0.0374, 0.0371, 0.0432, 0.0255, 0.0447], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 07:59:15,654 INFO [train.py:904] (3/8) Epoch 27, batch 1800, loss[loss=0.15, simple_loss=0.2428, pruned_loss=0.0286, over 17181.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2511, pruned_loss=0.0389, over 3319891.32 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:59:32,420 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:00:23,854 INFO [train.py:904] (3/8) Epoch 27, batch 1850, loss[loss=0.1516, simple_loss=0.2429, pruned_loss=0.03018, over 17229.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2519, pruned_loss=0.03902, over 3317857.38 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:00:37,844 INFO [optim.py:368] (3/8) 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 08:00:55,751 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:01:33,280 INFO [train.py:904] (3/8) Epoch 27, batch 1900, loss[loss=0.1525, simple_loss=0.2479, pruned_loss=0.02861, over 17224.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2516, pruned_loss=0.0383, over 3320881.41 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:01:42,984 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0283, 3.1348, 3.3100, 2.1669, 2.9132, 2.2403, 3.5323, 3.5042], device='cuda:3'), covar=tensor([0.0234, 0.1011, 0.0623, 0.2011, 0.0882, 0.1145, 0.0540, 0.0889], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0157, 0.0148, 0.0133, 0.0147, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 08:02:13,050 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5966, 3.1408, 3.6204, 1.9232, 3.6996, 3.7283, 3.0511, 2.7685], device='cuda:3'), covar=tensor([0.0775, 0.0338, 0.0216, 0.1237, 0.0124, 0.0227, 0.0420, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0131, 0.0130, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:02:42,098 INFO [train.py:904] (3/8) Epoch 27, batch 1950, loss[loss=0.1814, simple_loss=0.264, pruned_loss=0.04935, over 16841.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2514, pruned_loss=0.03805, over 3316125.56 frames. ], batch size: 116, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:54,906 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.085e+02 2.462e+02 3.209e+02 6.071e+02, threshold=4.924e+02, percent-clipped=6.0 2023-05-02 08:03:01,108 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 08:03:08,250 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:45,782 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 08:03:50,711 INFO [train.py:904] (3/8) Epoch 27, batch 2000, loss[loss=0.1524, simple_loss=0.2335, pruned_loss=0.03571, over 16518.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2517, pruned_loss=0.03813, over 3309735.26 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:04:10,492 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:33,657 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265933.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:34,774 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:01,213 INFO [train.py:904] (3/8) Epoch 27, batch 2050, loss[loss=0.1426, simple_loss=0.2339, pruned_loss=0.0257, over 16993.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2521, pruned_loss=0.03807, over 3295374.22 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:05:14,204 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.179e+02 2.495e+02 3.022e+02 6.334e+02, threshold=4.990e+02, percent-clipped=2.0 2023-05-02 08:05:25,357 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265970.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 08:05:36,758 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265982.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:44,534 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6483, 3.6226, 2.8243, 2.1932, 2.3307, 2.3704, 3.7209, 3.1299], device='cuda:3'), covar=tensor([0.2725, 0.0634, 0.1759, 0.3166, 0.2832, 0.2243, 0.0536, 0.1671], device='cuda:3'), in_proj_covar=tensor([0.0333, 0.0276, 0.0312, 0.0325, 0.0304, 0.0275, 0.0304, 0.0351], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 08:06:14,060 INFO [train.py:904] (3/8) Epoch 27, batch 2100, loss[loss=0.146, simple_loss=0.2368, pruned_loss=0.02758, over 16800.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2526, pruned_loss=0.03841, over 3311471.47 frames. ], batch size: 39, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:06:43,923 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 08:06:48,596 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6039, 3.7171, 2.1505, 4.3404, 2.8246, 4.2271, 2.1792, 2.9844], device='cuda:3'), covar=tensor([0.0398, 0.0435, 0.2119, 0.0438, 0.0986, 0.0517, 0.2136, 0.0923], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0183, 0.0199, 0.0176, 0.0183, 0.0224, 0.0208, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:07:22,090 INFO [train.py:904] (3/8) Epoch 27, batch 2150, loss[loss=0.1953, simple_loss=0.2677, pruned_loss=0.06148, over 16689.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2537, pruned_loss=0.03938, over 3312467.95 frames. ], batch size: 89, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:07:37,031 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.561e+02 2.274e+02 2.676e+02 3.152e+02 5.314e+02, threshold=5.352e+02, percent-clipped=2.0 2023-05-02 08:07:43,182 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.46 vs. limit=2.0 2023-05-02 08:07:46,308 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:08:13,873 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7202, 3.3955, 3.8345, 2.0324, 3.9040, 3.8787, 3.1479, 2.8936], device='cuda:3'), covar=tensor([0.0730, 0.0278, 0.0185, 0.1176, 0.0123, 0.0220, 0.0425, 0.0468], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0087, 0.0131, 0.0131, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:08:32,966 INFO [train.py:904] (3/8) Epoch 27, batch 2200, loss[loss=0.1866, simple_loss=0.2652, pruned_loss=0.05403, over 16873.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2543, pruned_loss=0.0399, over 3304035.49 frames. ], batch size: 116, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:08:52,015 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1517, 2.3346, 2.7891, 3.1122, 2.9462, 3.6539, 2.4770, 3.6191], device='cuda:3'), covar=tensor([0.0303, 0.0523, 0.0367, 0.0374, 0.0392, 0.0211, 0.0533, 0.0171], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0200, 0.0188, 0.0194, 0.0209, 0.0168, 0.0206, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 08:09:41,316 INFO [train.py:904] (3/8) Epoch 27, batch 2250, loss[loss=0.17, simple_loss=0.2522, pruned_loss=0.04384, over 16712.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2555, pruned_loss=0.04011, over 3310805.25 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:45,632 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6543, 2.7917, 2.8714, 4.5148, 2.7087, 3.0826, 2.7827, 2.9490], device='cuda:3'), covar=tensor([0.1268, 0.3261, 0.2699, 0.0559, 0.3620, 0.2329, 0.3162, 0.3213], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0474, 0.0387, 0.0338, 0.0446, 0.0542, 0.0444, 0.0553], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:09:56,594 INFO [optim.py:368] (3/8) 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 08:10:17,098 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3777, 2.6257, 2.0594, 2.3638, 2.9429, 2.7110, 3.1351, 3.1548], device='cuda:3'), covar=tensor([0.0249, 0.0478, 0.0666, 0.0539, 0.0319, 0.0426, 0.0278, 0.0317], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0249, 0.0238, 0.0238, 0.0250, 0.0249, 0.0248, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:10:24,556 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9565, 4.7234, 4.9702, 5.1676, 5.3914, 4.7148, 5.3798, 5.3924], device='cuda:3'), covar=tensor([0.2085, 0.1474, 0.1906, 0.0879, 0.0594, 0.1045, 0.0575, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0690, 0.0847, 0.0981, 0.0858, 0.0651, 0.0680, 0.0713, 0.0828], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:10:27,375 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:10:49,432 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3049, 4.3202, 4.6228, 4.6035, 4.6632, 4.3600, 4.3617, 4.2650], device='cuda:3'), covar=tensor([0.0398, 0.0724, 0.0432, 0.0442, 0.0541, 0.0499, 0.0879, 0.0738], device='cuda:3'), in_proj_covar=tensor([0.0438, 0.0495, 0.0476, 0.0440, 0.0528, 0.0506, 0.0579, 0.0403], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 08:10:51,416 INFO [train.py:904] (3/8) Epoch 27, batch 2300, loss[loss=0.1609, simple_loss=0.2559, pruned_loss=0.033, over 17259.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.255, pruned_loss=0.03971, over 3319054.14 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:11:06,837 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:23,537 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266226.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:11:23,869 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-05-02 08:11:25,873 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 27, batch 2350, loss[loss=0.1542, simple_loss=0.2478, pruned_loss=0.03032, over 17180.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2551, pruned_loss=0.03945, over 3327983.29 frames. ], batch size: 46, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:12:14,159 INFO [optim.py:368] (3/8) 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 08:12:22,261 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266270.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:12:24,901 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8775, 3.0754, 2.6315, 5.0591, 4.0945, 4.4006, 1.5757, 3.1489], device='cuda:3'), covar=tensor([0.1400, 0.0764, 0.1346, 0.0196, 0.0226, 0.0438, 0.1751, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0208, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:12:27,931 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:30,345 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:46,359 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266287.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 08:13:07,763 INFO [train.py:904] (3/8) Epoch 27, batch 2400, loss[loss=0.1685, simple_loss=0.2507, pruned_loss=0.04317, over 16517.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2553, pruned_loss=0.04015, over 3327040.81 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:13:28,958 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:07,671 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 2450, loss[loss=0.1435, simple_loss=0.236, pruned_loss=0.02549, over 17231.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2559, pruned_loss=0.03994, over 3327168.45 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:14:31,083 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:26,445 INFO [train.py:904] (3/8) Epoch 27, batch 2500, loss[loss=0.1494, simple_loss=0.229, pruned_loss=0.03487, over 16812.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2547, pruned_loss=0.03911, over 3332289.06 frames. ], batch size: 39, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:15:31,474 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9661, 4.9920, 5.3733, 5.3279, 5.3841, 5.0465, 4.9440, 4.7918], device='cuda:3'), covar=tensor([0.0374, 0.0596, 0.0408, 0.0481, 0.0494, 0.0482, 0.1120, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0440, 0.0496, 0.0478, 0.0441, 0.0530, 0.0508, 0.0583, 0.0405], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 08:15:31,605 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 27, batch 2550, loss[loss=0.1471, simple_loss=0.2375, pruned_loss=0.02835, over 15966.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2547, pruned_loss=0.03955, over 3327440.09 frames. ], batch size: 35, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:16:51,124 INFO [optim.py:368] (3/8) 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 08:16:51,500 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 2600, loss[loss=0.1908, simple_loss=0.274, pruned_loss=0.05387, over 16487.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2543, pruned_loss=0.03931, over 3335400.09 frames. ], batch size: 146, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:17:48,582 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6628, 6.0421, 5.7990, 5.9188, 5.5067, 5.3662, 5.4498, 6.2203], device='cuda:3'), covar=tensor([0.1571, 0.0964, 0.1093, 0.0867, 0.0877, 0.0729, 0.1407, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0724, 0.0876, 0.0713, 0.0674, 0.0555, 0.0550, 0.0740, 0.0686], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:18:07,458 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 08:18:17,126 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266525.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:20,978 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:55,445 INFO [train.py:904] (3/8) Epoch 27, batch 2650, loss[loss=0.186, simple_loss=0.2675, pruned_loss=0.05223, over 16798.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2553, pruned_loss=0.03915, over 3331541.55 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:19:08,879 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7212, 2.5808, 2.5087, 4.1193, 3.3379, 4.0685, 1.7238, 2.9225], device='cuda:3'), covar=tensor([0.1453, 0.0770, 0.1220, 0.0168, 0.0141, 0.0377, 0.1580, 0.0834], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0208, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:19:11,569 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:21,033 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5856, 3.7811, 3.9398, 2.6949, 3.5723, 4.0112, 3.6805, 2.2865], device='cuda:3'), covar=tensor([0.0530, 0.0201, 0.0065, 0.0436, 0.0124, 0.0095, 0.0108, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0136, 0.0103, 0.0116, 0.0099, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 08:19:24,575 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:29,420 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266582.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:20:05,798 INFO [train.py:904] (3/8) Epoch 27, batch 2700, loss[loss=0.1728, simple_loss=0.2696, pruned_loss=0.03803, over 17023.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2551, pruned_loss=0.03877, over 3331448.30 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:20:31,907 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:21:15,403 INFO [train.py:904] (3/8) Epoch 27, batch 2750, loss[loss=0.1775, simple_loss=0.2546, pruned_loss=0.05016, over 16768.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2556, pruned_loss=0.03856, over 3332078.44 frames. ], batch size: 124, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:21:29,196 INFO [optim.py:368] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 27, batch 2800, loss[loss=0.1699, simple_loss=0.2708, pruned_loss=0.03453, over 17023.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2555, pruned_loss=0.03824, over 3326850.98 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:22:25,082 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266704.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:32,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3040, 3.9872, 4.5428, 2.5445, 4.7122, 4.8276, 3.5022, 3.7767], device='cuda:3'), covar=tensor([0.0697, 0.0273, 0.0234, 0.1093, 0.0076, 0.0198, 0.0437, 0.0392], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0132, 0.0130, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:22:38,169 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7919, 3.5632, 4.0593, 2.1476, 4.1538, 4.2367, 3.1867, 3.2440], device='cuda:3'), covar=tensor([0.0795, 0.0294, 0.0223, 0.1166, 0.0108, 0.0250, 0.0440, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0132, 0.0131, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:23:15,450 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 2850, loss[loss=0.1755, simple_loss=0.2669, pruned_loss=0.04202, over 16527.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2547, pruned_loss=0.03816, over 3322254.20 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:23:48,195 INFO [optim.py:368] (3/8) 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 08:23:49,877 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:24:00,016 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 08:24:13,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6788, 3.4977, 2.8656, 2.2950, 2.2783, 2.3568, 3.6598, 3.0795], device='cuda:3'), covar=tensor([0.2767, 0.0668, 0.1736, 0.3281, 0.3121, 0.2243, 0.0625, 0.1795], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0277, 0.0314, 0.0327, 0.0306, 0.0276, 0.0305, 0.0354], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 08:24:40,541 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6109, 3.4957, 3.8422, 2.0569, 3.9003, 3.9844, 3.2395, 3.0156], device='cuda:3'), covar=tensor([0.0877, 0.0282, 0.0192, 0.1303, 0.0120, 0.0245, 0.0403, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0086, 0.0132, 0.0131, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:24:41,243 INFO [train.py:904] (3/8) Epoch 27, batch 2900, loss[loss=0.1778, simple_loss=0.2531, pruned_loss=0.05127, over 16878.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2538, pruned_loss=0.03823, over 3322484.29 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:25:04,247 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 2950, loss[loss=0.144, simple_loss=0.2322, pruned_loss=0.02788, over 16857.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.253, pruned_loss=0.03885, over 3317443.95 frames. ], batch size: 42, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:03,562 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2200, 4.2477, 4.4203, 4.0887, 4.2476, 4.7829, 4.2617, 4.0184], device='cuda:3'), covar=tensor([0.1984, 0.2470, 0.2937, 0.2806, 0.3035, 0.1485, 0.2059, 0.2847], device='cuda:3'), in_proj_covar=tensor([0.0437, 0.0644, 0.0714, 0.0529, 0.0704, 0.0739, 0.0555, 0.0705], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 08:26:04,413 INFO [optim.py:368] (3/8) 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 08:26:14,832 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:26:27,847 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-02 08:26:29,706 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266882.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:26:46,389 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 08:26:58,265 INFO [train.py:904] (3/8) Epoch 27, batch 3000, loss[loss=0.1744, simple_loss=0.2503, pruned_loss=0.04924, over 16701.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2538, pruned_loss=0.03973, over 3316806.54 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:58,265 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 08:27:07,045 INFO [train.py:938] (3/8) Epoch 27, validation: loss=0.1336, simple_loss=0.2386, pruned_loss=0.01429, over 944034.00 frames. 2023-05-02 08:27:07,046 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 08:27:27,842 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:30,412 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266930.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:28:16,681 INFO [train.py:904] (3/8) Epoch 27, batch 3050, loss[loss=0.1639, simple_loss=0.2621, pruned_loss=0.03282, over 17270.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2534, pruned_loss=0.03951, over 3326438.99 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:28:30,545 INFO [optim.py:368] (3/8) 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 08:28:55,901 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266981.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:24,924 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267002.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:25,706 INFO [train.py:904] (3/8) Epoch 27, batch 3100, loss[loss=0.1418, simple_loss=0.2368, pruned_loss=0.02338, over 17225.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2533, pruned_loss=0.03962, over 3329977.32 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:11,853 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1055, 5.7104, 5.7856, 5.4714, 5.6794, 6.1538, 5.6324, 5.3673], device='cuda:3'), covar=tensor([0.0917, 0.1842, 0.2717, 0.2032, 0.2559, 0.1072, 0.1478, 0.2237], device='cuda:3'), in_proj_covar=tensor([0.0436, 0.0644, 0.0713, 0.0528, 0.0704, 0.0738, 0.0554, 0.0703], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 08:30:11,862 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:33,232 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 3150, loss[loss=0.1671, simple_loss=0.2613, pruned_loss=0.03644, over 17036.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2531, pruned_loss=0.03955, over 3330932.38 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:45,724 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.137e+02 2.489e+02 2.954e+02 4.893e+02, threshold=4.979e+02, percent-clipped=1.0 2023-05-02 08:31:09,452 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-05-02 08:31:15,345 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:37,582 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267098.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:38,161 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 08:31:43,510 INFO [train.py:904] (3/8) Epoch 27, batch 3200, loss[loss=0.1491, simple_loss=0.2386, pruned_loss=0.02982, over 17201.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2518, pruned_loss=0.03882, over 3336053.79 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:31:44,390 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 08:32:07,960 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267120.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:13,457 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-05-02 08:32:26,639 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7932, 5.0002, 5.1313, 4.9050, 4.9992, 5.5646, 5.0358, 4.7576], device='cuda:3'), covar=tensor([0.1450, 0.2081, 0.2976, 0.2400, 0.2644, 0.1161, 0.1929, 0.2595], device='cuda:3'), in_proj_covar=tensor([0.0437, 0.0646, 0.0714, 0.0529, 0.0706, 0.0740, 0.0555, 0.0704], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 08:32:39,077 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:50,799 INFO [train.py:904] (3/8) Epoch 27, batch 3250, loss[loss=0.1428, simple_loss=0.242, pruned_loss=0.02179, over 17097.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2524, pruned_loss=0.03941, over 3322660.88 frames. ], batch size: 47, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:33:01,457 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:07,329 INFO [optim.py:368] (3/8) 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 08:33:12,826 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267168.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:19,039 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.52 vs. limit=5.0 2023-05-02 08:34:00,067 INFO [train.py:904] (3/8) Epoch 27, batch 3300, loss[loss=0.1902, simple_loss=0.274, pruned_loss=0.05317, over 16837.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.253, pruned_loss=0.03933, over 3332781.54 frames. ], batch size: 116, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:35:07,428 INFO [train.py:904] (3/8) Epoch 27, batch 3350, loss[loss=0.1532, simple_loss=0.2378, pruned_loss=0.03429, over 16272.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03896, over 3333680.80 frames. ], batch size: 36, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:35:22,678 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.912e+02 2.198e+02 2.588e+02 3.700e+02, threshold=4.395e+02, percent-clipped=0.0 2023-05-02 08:35:38,146 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:36:14,882 INFO [train.py:904] (3/8) Epoch 27, batch 3400, loss[loss=0.1502, simple_loss=0.2334, pruned_loss=0.03347, over 16854.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.254, pruned_loss=0.03929, over 3330180.25 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:36:32,680 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3385, 3.3705, 3.5457, 2.4517, 3.2340, 3.6474, 3.3404, 2.1061], device='cuda:3'), covar=tensor([0.0538, 0.0157, 0.0076, 0.0431, 0.0147, 0.0126, 0.0129, 0.0533], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0134], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 08:36:58,199 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267335.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:22,625 INFO [train.py:904] (3/8) Epoch 27, batch 3450, loss[loss=0.1732, simple_loss=0.2519, pruned_loss=0.04725, over 15568.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2525, pruned_loss=0.03891, over 3326304.57 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:37:31,745 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267360.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:36,675 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.041e+02 2.356e+02 2.660e+02 5.077e+02, threshold=4.713e+02, percent-clipped=1.0 2023-05-02 08:38:03,484 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:38:30,913 INFO [train.py:904] (3/8) Epoch 27, batch 3500, loss[loss=0.1588, simple_loss=0.2584, pruned_loss=0.02959, over 17129.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2505, pruned_loss=0.03792, over 3327828.21 frames. ], batch size: 48, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:38:39,646 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267408.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:19,826 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 3550, loss[loss=0.1502, simple_loss=0.2477, pruned_loss=0.02632, over 16999.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2494, pruned_loss=0.03769, over 3325639.91 frames. ], batch size: 50, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:39:41,935 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:43,212 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2542, 4.9566, 5.2219, 5.4319, 5.6375, 4.9673, 5.5816, 5.6089], device='cuda:3'), covar=tensor([0.1772, 0.1332, 0.1780, 0.0797, 0.0526, 0.0903, 0.0567, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0701, 0.0864, 0.1004, 0.0873, 0.0664, 0.0694, 0.0725, 0.0843], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:39:53,962 INFO [optim.py:368] (3/8) 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 08:40:06,756 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4090, 2.4027, 2.4703, 4.3524, 2.3440, 2.8158, 2.4554, 2.5534], device='cuda:3'), covar=tensor([0.1468, 0.4041, 0.3324, 0.0555, 0.4471, 0.2771, 0.3999, 0.3764], device='cuda:3'), in_proj_covar=tensor([0.0424, 0.0476, 0.0388, 0.0340, 0.0448, 0.0544, 0.0447, 0.0556], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:40:11,804 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1064, 3.0285, 3.1290, 2.2013, 2.9852, 3.2465, 3.0217, 1.9324], device='cuda:3'), covar=tensor([0.0532, 0.0139, 0.0100, 0.0445, 0.0156, 0.0154, 0.0138, 0.0561], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 08:40:36,722 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:40:48,164 INFO [train.py:904] (3/8) Epoch 27, batch 3600, loss[loss=0.136, simple_loss=0.22, pruned_loss=0.02599, over 17217.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2486, pruned_loss=0.03748, over 3318359.47 frames. ], batch size: 43, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:41:25,789 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 08:42:00,665 INFO [train.py:904] (3/8) Epoch 27, batch 3650, loss[loss=0.1642, simple_loss=0.2381, pruned_loss=0.04516, over 16742.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2475, pruned_loss=0.03799, over 3311238.20 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:03,479 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:16,688 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.009e+02 2.392e+02 2.845e+02 4.698e+02, threshold=4.785e+02, percent-clipped=1.0 2023-05-02 08:42:36,673 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:57,130 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-05-02 08:43:06,273 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6178, 2.5431, 1.9906, 2.6912, 2.1715, 2.7679, 2.1916, 2.3763], device='cuda:3'), covar=tensor([0.0347, 0.0404, 0.1298, 0.0271, 0.0682, 0.0439, 0.1231, 0.0678], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0177, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:43:14,273 INFO [train.py:904] (3/8) Epoch 27, batch 3700, loss[loss=0.1755, simple_loss=0.2543, pruned_loss=0.04834, over 15539.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2467, pruned_loss=0.03921, over 3288873.42 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:43:45,314 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267624.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:44:00,993 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8273, 2.7441, 2.6074, 4.2553, 3.5959, 4.1775, 1.6408, 3.0153], device='cuda:3'), covar=tensor([0.1400, 0.0714, 0.1185, 0.0193, 0.0194, 0.0391, 0.1614, 0.0864], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0202, 0.0207, 0.0219, 0.0208, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:44:27,016 INFO [train.py:904] (3/8) Epoch 27, batch 3750, loss[loss=0.1959, simple_loss=0.2907, pruned_loss=0.05055, over 17245.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2482, pruned_loss=0.04104, over 3268415.67 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:44:28,621 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3780, 4.2423, 4.4216, 4.5583, 4.6553, 4.2518, 4.4736, 4.6511], device='cuda:3'), covar=tensor([0.1804, 0.1268, 0.1329, 0.0744, 0.0681, 0.1192, 0.2911, 0.1056], device='cuda:3'), in_proj_covar=tensor([0.0699, 0.0859, 0.0999, 0.0869, 0.0661, 0.0690, 0.0721, 0.0838], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 08:44:42,769 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.122e+02 2.445e+02 2.969e+02 5.138e+02, threshold=4.890e+02, percent-clipped=1.0 2023-05-02 08:45:40,197 INFO [train.py:904] (3/8) Epoch 27, batch 3800, loss[loss=0.1611, simple_loss=0.2389, pruned_loss=0.04168, over 16418.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2492, pruned_loss=0.04189, over 3271750.28 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:45:41,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0374, 5.1157, 5.4397, 5.3953, 5.4718, 5.1012, 5.0233, 4.8876], device='cuda:3'), covar=tensor([0.0336, 0.0506, 0.0353, 0.0451, 0.0443, 0.0398, 0.0992, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0448, 0.0506, 0.0486, 0.0449, 0.0536, 0.0515, 0.0592, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 08:45:46,112 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8669, 4.7156, 4.5403, 3.1005, 4.0066, 4.5716, 3.9882, 2.6577], device='cuda:3'), covar=tensor([0.0495, 0.0032, 0.0044, 0.0389, 0.0100, 0.0078, 0.0098, 0.0444], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0090, 0.0092, 0.0137, 0.0103, 0.0117, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 08:46:31,911 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:46:52,912 INFO [train.py:904] (3/8) Epoch 27, batch 3850, loss[loss=0.1719, simple_loss=0.2522, pruned_loss=0.04583, over 16691.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2502, pruned_loss=0.0428, over 3271761.48 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:54,247 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:47:08,875 INFO [optim.py:368] (3/8) 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 08:47:09,281 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0812, 4.1206, 3.9787, 3.7282, 3.7758, 4.0799, 3.7230, 3.9243], device='cuda:3'), covar=tensor([0.0633, 0.0765, 0.0316, 0.0279, 0.0641, 0.0504, 0.1213, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0482, 0.0374, 0.0380, 0.0375, 0.0436, 0.0256, 0.0451], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 08:47:39,701 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:03,541 INFO [train.py:904] (3/8) Epoch 27, batch 3900, loss[loss=0.1659, simple_loss=0.2459, pruned_loss=0.04292, over 16443.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2495, pruned_loss=0.04335, over 3275683.24 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:48:10,192 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7882, 3.8280, 2.5160, 4.2119, 3.0281, 4.1928, 2.5921, 3.1691], device='cuda:3'), covar=tensor([0.0283, 0.0407, 0.1536, 0.0289, 0.0758, 0.0611, 0.1500, 0.0780], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0183, 0.0198, 0.0177, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 08:49:11,728 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267850.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:49:12,079 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 08:49:14,858 INFO [train.py:904] (3/8) Epoch 27, batch 3950, loss[loss=0.1605, simple_loss=0.2485, pruned_loss=0.0363, over 16496.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2491, pruned_loss=0.04413, over 3285839.82 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:49:32,326 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.233e+02 2.530e+02 3.152e+02 5.510e+02, threshold=5.059e+02, percent-clipped=1.0 2023-05-02 08:50:11,197 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-05-02 08:50:28,770 INFO [train.py:904] (3/8) Epoch 27, batch 4000, loss[loss=0.1723, simple_loss=0.2535, pruned_loss=0.0456, over 15301.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2491, pruned_loss=0.04434, over 3275896.09 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:42,554 INFO [train.py:904] (3/8) Epoch 27, batch 4050, loss[loss=0.1782, simple_loss=0.2648, pruned_loss=0.04586, over 16331.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2501, pruned_loss=0.04363, over 3272720.36 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:58,298 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.841e+02 2.143e+02 2.539e+02 4.386e+02, threshold=4.286e+02, percent-clipped=0.0 2023-05-02 08:52:59,882 INFO [train.py:904] (3/8) Epoch 27, batch 4100, loss[loss=0.1936, simple_loss=0.2853, pruned_loss=0.05097, over 16870.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2514, pruned_loss=0.0429, over 3266070.29 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:54:16,165 INFO [train.py:904] (3/8) Epoch 27, batch 4150, loss[loss=0.2498, simple_loss=0.3223, pruned_loss=0.08863, over 11701.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2585, pruned_loss=0.04513, over 3245956.46 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:54:33,172 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.031e+02 2.466e+02 3.151e+02 6.245e+02, threshold=4.932e+02, percent-clipped=8.0 2023-05-02 08:54:39,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7524, 1.8758, 2.4082, 2.7460, 2.6885, 3.1467, 2.0261, 3.0789], device='cuda:3'), covar=tensor([0.0291, 0.0624, 0.0353, 0.0375, 0.0357, 0.0209, 0.0631, 0.0179], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0199, 0.0188, 0.0195, 0.0209, 0.0168, 0.0206, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 08:55:32,138 INFO [train.py:904] (3/8) Epoch 27, batch 4200, loss[loss=0.2063, simple_loss=0.3003, pruned_loss=0.05616, over 17202.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2652, pruned_loss=0.04632, over 3224034.13 frames. ], batch size: 44, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:56:01,776 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 08:56:40,965 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268150.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:56:44,662 INFO [train.py:904] (3/8) Epoch 27, batch 4250, loss[loss=0.1729, simple_loss=0.2782, pruned_loss=0.03382, over 16879.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2685, pruned_loss=0.04622, over 3198541.27 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:57:00,874 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:57:57,374 INFO [train.py:904] (3/8) Epoch 27, batch 4300, loss[loss=0.1921, simple_loss=0.2872, pruned_loss=0.04848, over 15388.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2697, pruned_loss=0.04549, over 3200076.18 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:58:10,847 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268212.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:58:26,698 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-05-02 08:59:12,218 INFO [train.py:904] (3/8) Epoch 27, batch 4350, loss[loss=0.2048, simple_loss=0.292, pruned_loss=0.0588, over 16423.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.273, pruned_loss=0.04673, over 3160501.77 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:59:27,930 INFO [optim.py:368] (3/8) 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 08:59:43,198 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268273.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 08:59:45,607 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 4400, loss[loss=0.1892, simple_loss=0.2843, pruned_loss=0.04708, over 16691.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2754, pruned_loss=0.04789, over 3176036.56 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:03,458 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8579, 3.4953, 3.9800, 2.0523, 4.1414, 4.1692, 3.1747, 3.2809], device='cuda:3'), covar=tensor([0.0728, 0.0317, 0.0225, 0.1171, 0.0084, 0.0137, 0.0439, 0.0412], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0101, 0.0140, 0.0087, 0.0131, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 09:01:14,917 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268336.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:01:20,833 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9524, 5.3363, 5.5446, 5.2721, 5.2786, 5.8737, 5.3620, 5.0659], device='cuda:3'), covar=tensor([0.0891, 0.1694, 0.2113, 0.1678, 0.2371, 0.0879, 0.1301, 0.2151], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0634, 0.0695, 0.0517, 0.0689, 0.0725, 0.0541, 0.0690], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 09:01:37,963 INFO [train.py:904] (3/8) Epoch 27, batch 4450, loss[loss=0.2005, simple_loss=0.2828, pruned_loss=0.05917, over 12145.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2783, pruned_loss=0.04909, over 3178748.21 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:55,122 INFO [optim.py:368] (3/8) 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 09:02:40,602 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4893, 4.3675, 4.5636, 4.6828, 4.8123, 4.3723, 4.8047, 4.8525], device='cuda:3'), covar=tensor([0.1581, 0.1047, 0.1204, 0.0572, 0.0445, 0.1001, 0.0536, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0674, 0.0831, 0.0965, 0.0843, 0.0640, 0.0672, 0.0698, 0.0811], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:02:50,810 INFO [train.py:904] (3/8) Epoch 27, batch 4500, loss[loss=0.1755, simple_loss=0.2741, pruned_loss=0.03847, over 16890.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2788, pruned_loss=0.04975, over 3173099.31 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:03:42,697 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6704, 4.9016, 5.0877, 4.7698, 4.8582, 5.4540, 4.9166, 4.6229], device='cuda:3'), covar=tensor([0.1206, 0.1988, 0.2191, 0.2077, 0.2693, 0.0999, 0.1523, 0.2572], device='cuda:3'), in_proj_covar=tensor([0.0432, 0.0638, 0.0700, 0.0520, 0.0694, 0.0731, 0.0546, 0.0696], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 09:04:03,670 INFO [train.py:904] (3/8) Epoch 27, batch 4550, loss[loss=0.2152, simple_loss=0.298, pruned_loss=0.0662, over 17002.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2805, pruned_loss=0.05138, over 3183613.66 frames. ], batch size: 41, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:04:20,725 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.908e+02 2.230e+02 2.489e+02 1.268e+03, threshold=4.461e+02, percent-clipped=3.0 2023-05-02 09:04:42,358 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9447, 4.9691, 5.3218, 5.2650, 5.3368, 4.9664, 4.9258, 4.7092], device='cuda:3'), covar=tensor([0.0289, 0.0439, 0.0313, 0.0371, 0.0470, 0.0372, 0.0970, 0.0508], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0486, 0.0469, 0.0432, 0.0517, 0.0496, 0.0571, 0.0395], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 09:05:15,491 INFO [train.py:904] (3/8) Epoch 27, batch 4600, loss[loss=0.2102, simple_loss=0.2883, pruned_loss=0.06604, over 11759.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2821, pruned_loss=0.05157, over 3197304.31 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:05:58,120 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-02 09:06:23,484 INFO [train.py:904] (3/8) Epoch 27, batch 4650, loss[loss=0.1809, simple_loss=0.2695, pruned_loss=0.04612, over 16503.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2816, pruned_loss=0.05216, over 3192260.82 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:40,831 INFO [optim.py:368] (3/8) 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 09:06:45,260 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268568.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:07:32,852 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 4700, loss[loss=0.1969, simple_loss=0.2788, pruned_loss=0.0575, over 16557.00 frames. ], tot_loss[loss=0.1904, simple_loss=0.2787, pruned_loss=0.05107, over 3200118.84 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:08:15,778 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268631.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:08:45,864 INFO [train.py:904] (3/8) Epoch 27, batch 4750, loss[loss=0.1624, simple_loss=0.2566, pruned_loss=0.03406, over 16717.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2745, pruned_loss=0.04912, over 3199194.75 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:09:00,264 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 1.855e+02 2.157e+02 2.478e+02 7.312e+02, threshold=4.313e+02, percent-clipped=3.0 2023-05-02 09:09:59,302 INFO [train.py:904] (3/8) Epoch 27, batch 4800, loss[loss=0.1731, simple_loss=0.266, pruned_loss=0.04011, over 16720.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2708, pruned_loss=0.04676, over 3205062.59 frames. ], batch size: 76, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:14,170 INFO [train.py:904] (3/8) Epoch 27, batch 4850, loss[loss=0.1856, simple_loss=0.2746, pruned_loss=0.04834, over 16695.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2713, pruned_loss=0.04586, over 3188676.93 frames. ], batch size: 76, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:27,462 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3047, 2.4704, 2.4088, 4.1036, 2.3479, 2.7648, 2.4925, 2.6287], device='cuda:3'), covar=tensor([0.1407, 0.3474, 0.2916, 0.0559, 0.3916, 0.2590, 0.3616, 0.2995], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0470, 0.0382, 0.0334, 0.0444, 0.0538, 0.0440, 0.0551], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:11:31,499 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.812e+02 2.075e+02 2.444e+02 6.308e+02, threshold=4.151e+02, percent-clipped=1.0 2023-05-02 09:12:27,702 INFO [train.py:904] (3/8) Epoch 27, batch 4900, loss[loss=0.1665, simple_loss=0.2565, pruned_loss=0.03826, over 16527.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2697, pruned_loss=0.04439, over 3183013.01 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:21,916 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0770, 2.4165, 2.6074, 1.9218, 2.7173, 2.7647, 2.5094, 2.3675], device='cuda:3'), covar=tensor([0.0743, 0.0299, 0.0212, 0.1015, 0.0137, 0.0257, 0.0449, 0.0477], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0141, 0.0087, 0.0132, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 09:13:37,966 INFO [train.py:904] (3/8) Epoch 27, batch 4950, loss[loss=0.1839, simple_loss=0.2779, pruned_loss=0.04499, over 16380.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2689, pruned_loss=0.04359, over 3194523.14 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:54,418 INFO [optim.py:368] (3/8) 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 09:13:58,983 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268868.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:14:05,807 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6200, 4.6334, 4.9067, 4.8601, 4.8986, 4.6256, 4.5621, 4.4690], device='cuda:3'), covar=tensor([0.0292, 0.0446, 0.0344, 0.0418, 0.0522, 0.0353, 0.0937, 0.0472], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0482, 0.0466, 0.0429, 0.0513, 0.0493, 0.0568, 0.0393], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 09:14:28,266 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0250, 3.0816, 1.9465, 3.2860, 2.3439, 3.3271, 2.0133, 2.4584], device='cuda:3'), covar=tensor([0.0328, 0.0415, 0.1662, 0.0175, 0.0879, 0.0473, 0.1617, 0.0913], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0180, 0.0196, 0.0171, 0.0179, 0.0219, 0.0203, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 09:14:36,296 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:14:50,662 INFO [train.py:904] (3/8) Epoch 27, batch 5000, loss[loss=0.1764, simple_loss=0.2659, pruned_loss=0.0434, over 16527.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2713, pruned_loss=0.04413, over 3194402.33 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:15:09,313 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268916.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:15:29,354 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9085, 5.1420, 5.3623, 5.1107, 5.1875, 5.7291, 5.1496, 4.8358], device='cuda:3'), covar=tensor([0.0879, 0.1696, 0.1941, 0.1817, 0.2188, 0.0875, 0.1343, 0.2209], device='cuda:3'), in_proj_covar=tensor([0.0426, 0.0629, 0.0688, 0.0514, 0.0686, 0.0719, 0.0538, 0.0684], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 09:15:29,470 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:00,154 INFO [train.py:904] (3/8) Epoch 27, batch 5050, loss[loss=0.1763, simple_loss=0.2729, pruned_loss=0.03985, over 16661.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2723, pruned_loss=0.04408, over 3218326.49 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:16:03,797 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:17,236 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 1.970e+02 2.360e+02 2.844e+02 4.584e+02, threshold=4.719e+02, percent-clipped=0.0 2023-05-02 09:16:37,071 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:54,219 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 5100, loss[loss=0.1579, simple_loss=0.2494, pruned_loss=0.03323, over 17277.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2711, pruned_loss=0.04355, over 3209512.13 frames. ], batch size: 52, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:17:18,487 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7837, 5.0721, 4.8696, 4.9184, 4.6538, 4.5966, 4.5083, 5.1557], device='cuda:3'), covar=tensor([0.1322, 0.0868, 0.0976, 0.0762, 0.0804, 0.1119, 0.1177, 0.0814], device='cuda:3'), in_proj_covar=tensor([0.0708, 0.0860, 0.0701, 0.0660, 0.0545, 0.0542, 0.0724, 0.0675], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:17:31,914 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-02 09:18:22,308 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 5150, loss[loss=0.1716, simple_loss=0.2556, pruned_loss=0.04383, over 16627.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.27, pruned_loss=0.04237, over 3200133.25 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:18:41,484 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 1.963e+02 2.152e+02 2.508e+02 3.926e+02, threshold=4.304e+02, percent-clipped=0.0 2023-05-02 09:19:36,079 INFO [train.py:904] (3/8) Epoch 27, batch 5200, loss[loss=0.1669, simple_loss=0.2642, pruned_loss=0.0348, over 16745.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2681, pruned_loss=0.04162, over 3216575.41 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:20:47,202 INFO [train.py:904] (3/8) Epoch 27, batch 5250, loss[loss=0.1643, simple_loss=0.2529, pruned_loss=0.03786, over 16515.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2661, pruned_loss=0.04159, over 3217424.88 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:21:04,388 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 1.922e+02 2.287e+02 2.667e+02 4.356e+02, threshold=4.574e+02, percent-clipped=2.0 2023-05-02 09:21:19,820 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5209, 5.8160, 5.5489, 5.6630, 5.3520, 5.2797, 5.2251, 5.9324], device='cuda:3'), covar=tensor([0.1248, 0.0804, 0.0940, 0.0701, 0.0771, 0.0593, 0.1151, 0.0807], device='cuda:3'), in_proj_covar=tensor([0.0711, 0.0862, 0.0702, 0.0661, 0.0547, 0.0543, 0.0726, 0.0677], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:21:28,272 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3770, 4.4104, 4.2953, 2.5611, 3.7056, 4.3137, 3.6670, 2.4959], device='cuda:3'), covar=tensor([0.0643, 0.0034, 0.0040, 0.0472, 0.0111, 0.0094, 0.0119, 0.0477], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0089, 0.0090, 0.0136, 0.0101, 0.0114, 0.0098, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 09:22:00,557 INFO [train.py:904] (3/8) Epoch 27, batch 5300, loss[loss=0.1611, simple_loss=0.2535, pruned_loss=0.03434, over 16406.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2627, pruned_loss=0.04087, over 3214960.82 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:22:23,042 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8713, 4.2350, 3.1147, 2.5177, 2.8115, 2.7581, 4.5818, 3.6298], device='cuda:3'), covar=tensor([0.2745, 0.0573, 0.1792, 0.2753, 0.2651, 0.1936, 0.0413, 0.1232], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0326, 0.0305, 0.0274, 0.0304, 0.0351], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 09:22:55,475 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:08,881 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 5350, loss[loss=0.1771, simple_loss=0.2778, pruned_loss=0.03824, over 16170.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2615, pruned_loss=0.04036, over 3217306.70 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:23:18,792 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269257.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:29,779 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.050e+02 2.380e+02 2.740e+02 5.067e+02, threshold=4.761e+02, percent-clipped=1.0 2023-05-02 09:23:33,795 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0999, 5.1012, 4.9743, 4.5528, 4.6129, 5.0189, 4.9442, 4.7852], device='cuda:3'), covar=tensor([0.0624, 0.0734, 0.0311, 0.0314, 0.1050, 0.0622, 0.0304, 0.0621], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0464, 0.0360, 0.0365, 0.0360, 0.0420, 0.0246, 0.0430], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:23:43,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4704, 4.3593, 4.2982, 2.7729, 3.6951, 4.2860, 3.6790, 2.4241], device='cuda:3'), covar=tensor([0.0602, 0.0043, 0.0048, 0.0427, 0.0121, 0.0096, 0.0108, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0135, 0.0101, 0.0114, 0.0098, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 09:24:22,285 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-02 09:24:24,484 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:24:25,134 INFO [train.py:904] (3/8) Epoch 27, batch 5400, loss[loss=0.1782, simple_loss=0.274, pruned_loss=0.04123, over 16835.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2637, pruned_loss=0.04071, over 3217203.25 frames. ], batch size: 116, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:24:28,350 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:05,547 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-05-02 09:25:31,313 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:40,803 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 5450, loss[loss=0.1842, simple_loss=0.2687, pruned_loss=0.04988, over 12136.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2663, pruned_loss=0.04196, over 3202120.79 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:26:01,148 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 27, batch 5500, loss[loss=0.2403, simple_loss=0.3097, pruned_loss=0.08549, over 11780.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2732, pruned_loss=0.04578, over 3177404.20 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:27:06,092 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8592, 3.7835, 3.9081, 4.0096, 4.0993, 3.7072, 4.0519, 4.1168], device='cuda:3'), covar=tensor([0.1549, 0.1069, 0.1225, 0.0649, 0.0596, 0.1745, 0.0863, 0.0736], device='cuda:3'), in_proj_covar=tensor([0.0665, 0.0814, 0.0947, 0.0827, 0.0627, 0.0659, 0.0685, 0.0800], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:27:16,662 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:00,079 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:19,423 INFO [train.py:904] (3/8) Epoch 27, batch 5550, loss[loss=0.1953, simple_loss=0.2884, pruned_loss=0.05116, over 16681.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2802, pruned_loss=0.05052, over 3141689.51 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:28:26,906 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 09:28:38,496 INFO [optim.py:368] (3/8) 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 09:29:37,428 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269502.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:29:38,094 INFO [train.py:904] (3/8) Epoch 27, batch 5600, loss[loss=0.2384, simple_loss=0.3254, pruned_loss=0.07573, over 16234.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2853, pruned_loss=0.05504, over 3089876.31 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:30:56,885 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:31:01,865 INFO [train.py:904] (3/8) Epoch 27, batch 5650, loss[loss=0.2013, simple_loss=0.2932, pruned_loss=0.05472, over 17033.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2897, pruned_loss=0.05898, over 3060526.90 frames. ], batch size: 53, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:31:09,370 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8939, 4.1261, 3.9669, 3.9888, 3.7472, 3.7677, 3.8099, 4.1232], device='cuda:3'), covar=tensor([0.1077, 0.0926, 0.1146, 0.0914, 0.0803, 0.1595, 0.0992, 0.1020], device='cuda:3'), in_proj_covar=tensor([0.0703, 0.0854, 0.0696, 0.0656, 0.0541, 0.0536, 0.0717, 0.0669], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:31:20,024 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.426e+02 3.478e+02 4.295e+02 5.160e+02 1.255e+03, threshold=8.591e+02, percent-clipped=5.0 2023-05-02 09:32:11,889 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:13,057 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:20,978 INFO [train.py:904] (3/8) Epoch 27, batch 5700, loss[loss=0.2438, simple_loss=0.3125, pruned_loss=0.08757, over 11551.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2913, pruned_loss=0.06052, over 3052591.98 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:31,921 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269647.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:33:41,286 INFO [train.py:904] (3/8) Epoch 27, batch 5750, loss[loss=0.2217, simple_loss=0.3077, pruned_loss=0.06788, over 16777.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2946, pruned_loss=0.06267, over 3025761.36 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:47,045 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0597, 2.2178, 2.2668, 3.6688, 2.0917, 2.5359, 2.3127, 2.3858], device='cuda:3'), covar=tensor([0.1471, 0.3599, 0.3060, 0.0604, 0.4192, 0.2464, 0.3575, 0.3362], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0469, 0.0380, 0.0333, 0.0442, 0.0536, 0.0439, 0.0548], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:33:59,182 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.911e+02 3.444e+02 4.109e+02 8.393e+02, threshold=6.889e+02, percent-clipped=0.0 2023-05-02 09:34:50,303 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:35:02,093 INFO [train.py:904] (3/8) Epoch 27, batch 5800, loss[loss=0.1995, simple_loss=0.2886, pruned_loss=0.05516, over 16672.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2941, pruned_loss=0.06084, over 3047876.36 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:35:11,633 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269708.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:36:21,258 INFO [train.py:904] (3/8) Epoch 27, batch 5850, loss[loss=0.2294, simple_loss=0.3069, pruned_loss=0.07596, over 15458.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2924, pruned_loss=0.05989, over 3044096.17 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:36:40,939 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 3.085e+02 3.568e+02 4.510e+02 7.349e+02, threshold=7.136e+02, percent-clipped=2.0 2023-05-02 09:36:53,609 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269797.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:37:45,792 INFO [train.py:904] (3/8) Epoch 27, batch 5900, loss[loss=0.201, simple_loss=0.2858, pruned_loss=0.05808, over 16238.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.292, pruned_loss=0.05964, over 3050198.82 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:38:38,110 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269834.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:38:40,685 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:39:06,822 INFO [train.py:904] (3/8) Epoch 27, batch 5950, loss[loss=0.1988, simple_loss=0.2935, pruned_loss=0.05205, over 16675.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2926, pruned_loss=0.05889, over 3019893.17 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:39:27,600 INFO [optim.py:368] (3/8) 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 09:39:33,618 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 09:40:05,716 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2300, 5.5492, 5.2952, 5.3245, 5.0418, 5.0253, 4.9592, 5.6742], device='cuda:3'), covar=tensor([0.1246, 0.0837, 0.0939, 0.0879, 0.0813, 0.0814, 0.1211, 0.0823], device='cuda:3'), in_proj_covar=tensor([0.0700, 0.0850, 0.0692, 0.0654, 0.0536, 0.0535, 0.0713, 0.0665], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:40:17,501 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:17,561 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:24,918 INFO [train.py:904] (3/8) Epoch 27, batch 6000, loss[loss=0.1872, simple_loss=0.2788, pruned_loss=0.04782, over 16228.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2916, pruned_loss=0.058, over 3044341.87 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:40:24,918 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 09:40:35,106 INFO [train.py:938] (3/8) Epoch 27, validation: loss=0.148, simple_loss=0.2603, pruned_loss=0.01783, over 944034.00 frames. 2023-05-02 09:40:35,107 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 09:41:37,529 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 6050, loss[loss=0.2292, simple_loss=0.2985, pruned_loss=0.07993, over 11646.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.29, pruned_loss=0.05737, over 3053038.89 frames. ], batch size: 246, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:42:12,241 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.707e+02 3.068e+02 3.739e+02 6.756e+02, threshold=6.136e+02, percent-clipped=2.0 2023-05-02 09:42:17,184 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2764, 2.4426, 2.4466, 4.0102, 2.3076, 2.8314, 2.4691, 2.5666], device='cuda:3'), covar=tensor([0.1547, 0.3648, 0.3089, 0.0575, 0.4155, 0.2408, 0.3677, 0.3334], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0468, 0.0379, 0.0333, 0.0442, 0.0535, 0.0439, 0.0547], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:43:05,452 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8411, 2.7108, 2.8589, 2.1249, 2.6983, 2.1617, 2.7489, 2.9029], device='cuda:3'), covar=tensor([0.0258, 0.0908, 0.0502, 0.1856, 0.0803, 0.0930, 0.0569, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 09:43:13,684 INFO [train.py:904] (3/8) Epoch 27, batch 6100, loss[loss=0.1803, simple_loss=0.2754, pruned_loss=0.0426, over 16340.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2894, pruned_loss=0.05608, over 3073236.10 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:43:22,351 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270008.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:15,112 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.96 vs. limit=5.0 2023-05-02 09:44:33,191 INFO [train.py:904] (3/8) Epoch 27, batch 6150, loss[loss=0.2043, simple_loss=0.2965, pruned_loss=0.05609, over 16258.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2872, pruned_loss=0.05568, over 3068834.71 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:44:37,806 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270056.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:53,892 INFO [optim.py:368] (3/8) 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 09:45:19,025 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 09:45:42,204 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270097.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:45:51,130 INFO [train.py:904] (3/8) Epoch 27, batch 6200, loss[loss=0.2028, simple_loss=0.2873, pruned_loss=0.05916, over 16725.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2858, pruned_loss=0.05537, over 3078865.14 frames. ], batch size: 134, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:46:11,104 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3422, 2.4002, 2.3950, 4.1007, 2.2898, 2.6904, 2.4736, 2.5389], device='cuda:3'), covar=tensor([0.1369, 0.3647, 0.3002, 0.0560, 0.4223, 0.2550, 0.3527, 0.3399], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0468, 0.0380, 0.0333, 0.0442, 0.0536, 0.0439, 0.0548], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:46:33,639 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270129.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:46:57,925 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:47:09,706 INFO [train.py:904] (3/8) Epoch 27, batch 6250, loss[loss=0.1821, simple_loss=0.2708, pruned_loss=0.04666, over 12111.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.285, pruned_loss=0.05477, over 3093984.63 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:47:29,204 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.710e+02 3.440e+02 4.218e+02 8.566e+02, threshold=6.880e+02, percent-clipped=4.0 2023-05-02 09:48:05,603 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4761, 2.2751, 1.9393, 2.0992, 2.5597, 2.2363, 2.3021, 2.6757], device='cuda:3'), covar=tensor([0.0240, 0.0412, 0.0552, 0.0492, 0.0281, 0.0409, 0.0250, 0.0313], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0240, 0.0230, 0.0231, 0.0242, 0.0240, 0.0239, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:48:08,658 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 6300, loss[loss=0.1594, simple_loss=0.2597, pruned_loss=0.02953, over 16810.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.285, pruned_loss=0.05397, over 3115814.59 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:49:45,973 INFO [train.py:904] (3/8) Epoch 27, batch 6350, loss[loss=0.2248, simple_loss=0.2927, pruned_loss=0.07846, over 11144.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2849, pruned_loss=0.0542, over 3131534.65 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:50:05,620 INFO [optim.py:368] (3/8) 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 09:50:28,115 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270280.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:51:02,483 INFO [train.py:904] (3/8) Epoch 27, batch 6400, loss[loss=0.1892, simple_loss=0.2748, pruned_loss=0.05183, over 17111.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2861, pruned_loss=0.05624, over 3102369.12 frames. ], batch size: 49, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:01,043 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4226, 5.5189, 5.2704, 4.8074, 4.6357, 5.3978, 5.3391, 4.9353], device='cuda:3'), covar=tensor([0.1057, 0.1441, 0.0547, 0.0681, 0.1689, 0.0914, 0.0602, 0.1435], device='cuda:3'), in_proj_covar=tensor([0.0304, 0.0459, 0.0356, 0.0360, 0.0356, 0.0413, 0.0244, 0.0428], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 09:52:01,107 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270341.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:52:19,089 INFO [train.py:904] (3/8) Epoch 27, batch 6450, loss[loss=0.1743, simple_loss=0.2634, pruned_loss=0.04259, over 16264.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2856, pruned_loss=0.0552, over 3109513.06 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:39,046 INFO [optim.py:368] (3/8) 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 09:53:26,112 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 09:53:27,521 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:53:37,254 INFO [train.py:904] (3/8) Epoch 27, batch 6500, loss[loss=0.2043, simple_loss=0.2867, pruned_loss=0.06092, over 16166.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2842, pruned_loss=0.0548, over 3108359.62 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:54:17,010 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270429.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:54:33,516 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270439.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:54:41,580 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 09:54:58,115 INFO [train.py:904] (3/8) Epoch 27, batch 6550, loss[loss=0.2152, simple_loss=0.2892, pruned_loss=0.07054, over 11570.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2866, pruned_loss=0.05557, over 3094029.30 frames. ], batch size: 250, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:55:04,759 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:15,574 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 09:55:17,871 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 2.679e+02 3.213e+02 3.936e+02 7.742e+02, threshold=6.427e+02, percent-clipped=6.0 2023-05-02 09:55:34,672 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:57,733 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:56:10,584 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:56:14,134 INFO [train.py:904] (3/8) Epoch 27, batch 6600, loss[loss=0.1905, simple_loss=0.2844, pruned_loss=0.04829, over 16500.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2887, pruned_loss=0.05626, over 3081140.52 frames. ], batch size: 68, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:09,703 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 6650, loss[loss=0.2051, simple_loss=0.2886, pruned_loss=0.06075, over 15268.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2885, pruned_loss=0.05698, over 3084955.27 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:50,322 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.544e+02 2.739e+02 3.273e+02 3.972e+02 7.700e+02, threshold=6.545e+02, percent-clipped=3.0 2023-05-02 09:58:46,097 INFO [train.py:904] (3/8) Epoch 27, batch 6700, loss[loss=0.1734, simple_loss=0.2624, pruned_loss=0.04218, over 16760.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2878, pruned_loss=0.0571, over 3090898.32 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:58:52,127 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9817, 5.2975, 5.6282, 5.5427, 5.6157, 5.2956, 4.9647, 5.0112], device='cuda:3'), covar=tensor([0.0622, 0.0724, 0.0463, 0.0640, 0.0606, 0.0559, 0.1435, 0.0583], device='cuda:3'), in_proj_covar=tensor([0.0426, 0.0477, 0.0464, 0.0424, 0.0510, 0.0489, 0.0561, 0.0390], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 09:58:56,453 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1776, 3.0029, 3.3769, 1.7914, 3.4717, 3.5637, 2.7834, 2.6051], device='cuda:3'), covar=tensor([0.0972, 0.0363, 0.0246, 0.1346, 0.0116, 0.0212, 0.0494, 0.0575], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0111, 0.0102, 0.0140, 0.0086, 0.0131, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 09:59:01,606 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:59:35,301 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270636.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:59:53,486 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2443, 5.2728, 5.6887, 5.6381, 5.7177, 5.3630, 5.2605, 5.0458], device='cuda:3'), covar=tensor([0.0316, 0.0568, 0.0324, 0.0367, 0.0484, 0.0358, 0.1010, 0.0473], device='cuda:3'), in_proj_covar=tensor([0.0426, 0.0477, 0.0464, 0.0425, 0.0511, 0.0489, 0.0562, 0.0391], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 10:00:01,086 INFO [train.py:904] (3/8) Epoch 27, batch 6750, loss[loss=0.1843, simple_loss=0.2836, pruned_loss=0.04248, over 16174.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2871, pruned_loss=0.05696, over 3102639.07 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:00:20,167 INFO [optim.py:368] (3/8) 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 10:00:23,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4127, 1.7168, 2.1190, 2.3346, 2.4120, 2.6302, 1.8553, 2.5825], device='cuda:3'), covar=tensor([0.0255, 0.0559, 0.0339, 0.0369, 0.0351, 0.0230, 0.0615, 0.0169], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0190, 0.0205, 0.0164, 0.0202, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:00:32,938 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:00:52,236 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6180, 4.4502, 4.2950, 3.0389, 3.8023, 4.3233, 3.8085, 2.4825], device='cuda:3'), covar=tensor([0.0541, 0.0044, 0.0051, 0.0364, 0.0113, 0.0102, 0.0096, 0.0461], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0090, 0.0090, 0.0136, 0.0101, 0.0115, 0.0098, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 10:01:15,210 INFO [train.py:904] (3/8) Epoch 27, batch 6800, loss[loss=0.1859, simple_loss=0.2841, pruned_loss=0.04379, over 16840.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2874, pruned_loss=0.05714, over 3095028.74 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:01:53,377 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270726.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:33,309 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:34,157 INFO [train.py:904] (3/8) Epoch 27, batch 6850, loss[loss=0.2049, simple_loss=0.3105, pruned_loss=0.04961, over 16776.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2886, pruned_loss=0.05776, over 3092602.42 frames. ], batch size: 39, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:02:53,221 INFO [optim.py:368] (3/8) 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 10:02:56,913 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 10:03:23,981 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3435, 3.4584, 3.5900, 3.5714, 3.5867, 3.4419, 3.4449, 3.5093], device='cuda:3'), covar=tensor([0.0431, 0.0784, 0.0506, 0.0462, 0.0550, 0.0555, 0.0788, 0.0535], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0480, 0.0465, 0.0427, 0.0512, 0.0491, 0.0564, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 10:03:25,244 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:37,582 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 6900, loss[loss=0.2355, simple_loss=0.3042, pruned_loss=0.0834, over 11693.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2907, pruned_loss=0.05753, over 3077924.84 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:07,698 INFO [train.py:904] (3/8) Epoch 27, batch 6950, loss[loss=0.2554, simple_loss=0.3178, pruned_loss=0.0965, over 11095.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2918, pruned_loss=0.05899, over 3057288.49 frames. ], batch size: 250, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:28,445 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.915e+02 3.513e+02 4.432e+02 8.204e+02, threshold=7.026e+02, percent-clipped=5.0 2023-05-02 10:06:23,846 INFO [train.py:904] (3/8) Epoch 27, batch 7000, loss[loss=0.1725, simple_loss=0.2738, pruned_loss=0.03557, over 17037.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2913, pruned_loss=0.05766, over 3070150.88 frames. ], batch size: 50, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:06:46,547 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:06:55,897 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6748, 4.6710, 4.5199, 3.6555, 4.5952, 1.7170, 4.3517, 4.1159], device='cuda:3'), covar=tensor([0.0142, 0.0133, 0.0229, 0.0406, 0.0122, 0.2966, 0.0191, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0172, 0.0211, 0.0185, 0.0186, 0.0215, 0.0199, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:07:15,581 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270936.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 10:07:40,699 INFO [train.py:904] (3/8) Epoch 27, batch 7050, loss[loss=0.1845, simple_loss=0.2801, pruned_loss=0.04443, over 16735.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2923, pruned_loss=0.05746, over 3077341.31 frames. ], batch size: 39, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:08:01,040 INFO [optim.py:368] (3/8) 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 10:08:05,077 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:18,540 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:24,810 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3408, 2.9120, 2.6905, 2.3089, 2.2653, 2.3585, 2.9002, 2.8312], device='cuda:3'), covar=tensor([0.2603, 0.0743, 0.1648, 0.2561, 0.2315, 0.2110, 0.0532, 0.1403], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0275, 0.0312, 0.0327, 0.0305, 0.0275, 0.0304, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:08:27,529 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270984.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:08:38,060 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 10:08:58,235 INFO [train.py:904] (3/8) Epoch 27, batch 7100, loss[loss=0.1891, simple_loss=0.2846, pruned_loss=0.04674, over 16339.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2908, pruned_loss=0.05747, over 3055595.77 frames. ], batch size: 35, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:15,327 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 7150, loss[loss=0.1916, simple_loss=0.2824, pruned_loss=0.05041, over 15358.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2889, pruned_loss=0.05689, over 3074647.27 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:37,964 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.616e+02 3.372e+02 3.961e+02 8.125e+02, threshold=6.744e+02, percent-clipped=3.0 2023-05-02 10:10:59,667 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:05,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6237, 3.7428, 2.7413, 2.2531, 2.6725, 2.4727, 4.0654, 3.3800], device='cuda:3'), covar=tensor([0.3342, 0.0800, 0.2266, 0.3113, 0.2860, 0.2307, 0.0595, 0.1390], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0275, 0.0313, 0.0328, 0.0306, 0.0276, 0.0305, 0.0351], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:11:05,466 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-05-02 10:11:18,923 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:26,213 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8006, 3.9335, 4.1316, 4.1002, 4.1072, 3.8972, 3.8937, 3.9056], device='cuda:3'), covar=tensor([0.0383, 0.0575, 0.0403, 0.0411, 0.0471, 0.0464, 0.0843, 0.0547], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0479, 0.0465, 0.0426, 0.0512, 0.0491, 0.0563, 0.0392], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 10:11:28,068 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 7200, loss[loss=0.1638, simple_loss=0.26, pruned_loss=0.03375, over 16811.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2866, pruned_loss=0.05532, over 3072788.98 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:12:37,365 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:12:53,491 INFO [train.py:904] (3/8) Epoch 27, batch 7250, loss[loss=0.179, simple_loss=0.2617, pruned_loss=0.04818, over 15454.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2842, pruned_loss=0.05413, over 3067204.00 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:13:16,037 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.666e+02 3.085e+02 3.699e+02 8.603e+02, threshold=6.169e+02, percent-clipped=3.0 2023-05-02 10:14:10,772 INFO [train.py:904] (3/8) Epoch 27, batch 7300, loss[loss=0.1875, simple_loss=0.282, pruned_loss=0.04649, over 15311.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2835, pruned_loss=0.05415, over 3066945.98 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:29,743 INFO [train.py:904] (3/8) Epoch 27, batch 7350, loss[loss=0.1902, simple_loss=0.2768, pruned_loss=0.0518, over 17031.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2851, pruned_loss=0.05549, over 3028994.28 frames. ], batch size: 53, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:42,223 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 10:15:54,815 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:01,218 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:14,369 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3332, 2.9065, 2.6029, 2.2567, 2.2335, 2.2859, 2.9458, 2.8192], device='cuda:3'), covar=tensor([0.2620, 0.0790, 0.1762, 0.2621, 0.2367, 0.2251, 0.0572, 0.1410], device='cuda:3'), in_proj_covar=tensor([0.0336, 0.0276, 0.0313, 0.0328, 0.0306, 0.0276, 0.0306, 0.0352], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:16:42,410 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0610, 2.2947, 2.3062, 2.7576, 1.7777, 3.1642, 1.9051, 2.6601], device='cuda:3'), covar=tensor([0.1190, 0.0701, 0.1183, 0.0219, 0.0124, 0.0440, 0.1557, 0.0787], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0179, 0.0200, 0.0200, 0.0207, 0.0218, 0.0209, 0.0198], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:16:48,737 INFO [train.py:904] (3/8) Epoch 27, batch 7400, loss[loss=0.1863, simple_loss=0.2828, pruned_loss=0.04485, over 16881.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2867, pruned_loss=0.05667, over 3019528.08 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:17:10,655 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:18,862 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:59,791 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 10:18:06,473 INFO [train.py:904] (3/8) Epoch 27, batch 7450, loss[loss=0.2194, simple_loss=0.3066, pruned_loss=0.06611, over 16450.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2878, pruned_loss=0.0577, over 3024189.90 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:18:30,891 INFO [optim.py:368] (3/8) 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 10:18:32,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8669, 2.1665, 2.4140, 3.1002, 2.1989, 2.3431, 2.3164, 2.2479], device='cuda:3'), covar=tensor([0.1463, 0.3455, 0.2776, 0.0840, 0.4287, 0.2735, 0.3398, 0.3507], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0471, 0.0381, 0.0334, 0.0445, 0.0539, 0.0442, 0.0549], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:18:36,034 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6549, 2.6039, 1.8959, 2.7272, 2.0484, 2.7781, 2.1675, 2.3920], device='cuda:3'), covar=tensor([0.0325, 0.0409, 0.1301, 0.0364, 0.0708, 0.0565, 0.1202, 0.0643], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0182, 0.0197, 0.0172, 0.0180, 0.0221, 0.0204, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:18:45,214 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:47,824 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2775, 2.0601, 1.7563, 1.7780, 2.2801, 1.9659, 1.8671, 2.3756], device='cuda:3'), covar=tensor([0.0286, 0.0454, 0.0602, 0.0573, 0.0302, 0.0411, 0.0222, 0.0293], device='cuda:3'), in_proj_covar=tensor([0.0224, 0.0240, 0.0229, 0.0231, 0.0241, 0.0239, 0.0238, 0.0239], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:18:55,898 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 7500, loss[loss=0.182, simple_loss=0.2732, pruned_loss=0.04544, over 16730.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.288, pruned_loss=0.05762, over 3016890.51 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:20:08,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0389, 2.4518, 2.5718, 1.9718, 2.6728, 2.7715, 2.3948, 2.3689], device='cuda:3'), covar=tensor([0.0713, 0.0294, 0.0241, 0.0924, 0.0148, 0.0317, 0.0504, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0138, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:20:08,142 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:14,060 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:22,929 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271436.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:49,668 INFO [train.py:904] (3/8) Epoch 27, batch 7550, loss[loss=0.2011, simple_loss=0.2844, pruned_loss=0.05888, over 17130.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2872, pruned_loss=0.05801, over 2991761.34 frames. ], batch size: 48, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:20:57,814 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1788, 2.9766, 3.3043, 1.7559, 3.3661, 3.4541, 2.7420, 2.5507], device='cuda:3'), covar=tensor([0.0884, 0.0327, 0.0211, 0.1267, 0.0115, 0.0206, 0.0471, 0.0564], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0138, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:21:11,191 INFO [optim.py:368] (3/8) 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 10:21:24,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3792, 3.3471, 3.4146, 3.4812, 3.5149, 3.2855, 3.4848, 3.5502], device='cuda:3'), covar=tensor([0.1268, 0.0931, 0.0986, 0.0632, 0.0670, 0.2173, 0.1142, 0.0902], device='cuda:3'), in_proj_covar=tensor([0.0656, 0.0801, 0.0932, 0.0813, 0.0621, 0.0651, 0.0681, 0.0794], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:21:40,978 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:22:05,495 INFO [train.py:904] (3/8) Epoch 27, batch 7600, loss[loss=0.185, simple_loss=0.277, pruned_loss=0.04648, over 16991.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2867, pruned_loss=0.05768, over 3005328.10 frames. ], batch size: 41, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:22,937 INFO [train.py:904] (3/8) Epoch 27, batch 7650, loss[loss=0.1903, simple_loss=0.2866, pruned_loss=0.04698, over 16496.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2867, pruned_loss=0.05733, over 3029208.46 frames. ], batch size: 75, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:45,488 INFO [optim.py:368] (3/8) 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 10:23:55,698 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:24:24,001 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7769, 2.8099, 2.4915, 4.9064, 3.7472, 4.1734, 1.6279, 3.0649], device='cuda:3'), covar=tensor([0.1429, 0.0819, 0.1382, 0.0180, 0.0364, 0.0435, 0.1757, 0.0865], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0201, 0.0208, 0.0219, 0.0210, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:24:32,613 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 10:24:43,694 INFO [train.py:904] (3/8) Epoch 27, batch 7700, loss[loss=0.186, simple_loss=0.2695, pruned_loss=0.05121, over 16652.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2862, pruned_loss=0.05731, over 3048579.02 frames. ], batch size: 134, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:24:53,374 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 10:25:06,109 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:25:12,496 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:26:02,414 INFO [train.py:904] (3/8) Epoch 27, batch 7750, loss[loss=0.2011, simple_loss=0.2837, pruned_loss=0.05931, over 16201.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2862, pruned_loss=0.05699, over 3052447.16 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:26:24,500 INFO [optim.py:368] (3/8) 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 10:27:20,107 INFO [train.py:904] (3/8) Epoch 27, batch 7800, loss[loss=0.1926, simple_loss=0.273, pruned_loss=0.05615, over 16927.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2865, pruned_loss=0.05747, over 3057740.03 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:27:31,197 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9748, 5.3661, 5.5502, 5.2681, 5.3456, 5.8751, 5.3912, 5.1112], device='cuda:3'), covar=tensor([0.0997, 0.1895, 0.2394, 0.2132, 0.2353, 0.1040, 0.1509, 0.2480], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0633, 0.0696, 0.0516, 0.0687, 0.0725, 0.0545, 0.0694], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:27:57,975 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8011, 1.4176, 1.7225, 1.6713, 1.7961, 1.9024, 1.6448, 1.8533], device='cuda:3'), covar=tensor([0.0268, 0.0414, 0.0223, 0.0344, 0.0297, 0.0209, 0.0456, 0.0157], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0197, 0.0184, 0.0189, 0.0204, 0.0164, 0.0201, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:28:04,768 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271731.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:28:36,665 INFO [train.py:904] (3/8) Epoch 27, batch 7850, loss[loss=0.242, simple_loss=0.3188, pruned_loss=0.08264, over 11324.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.288, pruned_loss=0.05728, over 3067478.75 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:28:57,992 INFO [optim.py:368] (3/8) 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 10:29:17,322 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 10:29:20,880 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:29:48,846 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1699, 1.5721, 1.9410, 2.0784, 2.1980, 2.4046, 1.7500, 2.3507], device='cuda:3'), covar=tensor([0.0273, 0.0545, 0.0315, 0.0395, 0.0331, 0.0219, 0.0582, 0.0175], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0197, 0.0184, 0.0189, 0.0204, 0.0164, 0.0201, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:29:52,543 INFO [train.py:904] (3/8) Epoch 27, batch 7900, loss[loss=0.1925, simple_loss=0.2826, pruned_loss=0.05119, over 16737.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2876, pruned_loss=0.05734, over 3075683.00 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:30:21,892 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4058, 1.7107, 2.0990, 2.3626, 2.4228, 2.7033, 1.9133, 2.6247], device='cuda:3'), covar=tensor([0.0264, 0.0533, 0.0349, 0.0366, 0.0355, 0.0215, 0.0556, 0.0173], device='cuda:3'), in_proj_covar=tensor([0.0195, 0.0196, 0.0183, 0.0188, 0.0204, 0.0163, 0.0200, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:31:12,172 INFO [train.py:904] (3/8) Epoch 27, batch 7950, loss[loss=0.216, simple_loss=0.3075, pruned_loss=0.06227, over 16877.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2878, pruned_loss=0.05759, over 3091058.50 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:15,296 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:31:23,211 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2585, 3.5828, 3.4905, 2.1477, 3.0511, 2.4536, 3.6528, 3.8331], device='cuda:3'), covar=tensor([0.0280, 0.0805, 0.0655, 0.2227, 0.0929, 0.1044, 0.0612, 0.0910], device='cuda:3'), in_proj_covar=tensor([0.0159, 0.0170, 0.0169, 0.0155, 0.0148, 0.0132, 0.0145, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:31:36,730 INFO [optim.py:368] (3/8) 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 10:32:19,174 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6480, 3.1952, 3.1705, 2.0412, 2.8313, 2.2141, 3.1083, 3.4514], device='cuda:3'), covar=tensor([0.0472, 0.0904, 0.0730, 0.2322, 0.1030, 0.1145, 0.1006, 0.0993], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0133, 0.0146, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 10:32:29,598 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 10:32:29,919 INFO [train.py:904] (3/8) Epoch 27, batch 8000, loss[loss=0.1782, simple_loss=0.2688, pruned_loss=0.04378, over 16892.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2888, pruned_loss=0.05839, over 3094690.85 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:32:49,756 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271915.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:32:52,898 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:19,510 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:43,813 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0422, 4.8794, 5.0906, 5.2506, 5.4359, 4.8333, 5.4411, 5.4561], device='cuda:3'), covar=tensor([0.2127, 0.1335, 0.1653, 0.0751, 0.0645, 0.0965, 0.0659, 0.0788], device='cuda:3'), in_proj_covar=tensor([0.0656, 0.0798, 0.0931, 0.0809, 0.0623, 0.0648, 0.0680, 0.0792], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:33:47,184 INFO [train.py:904] (3/8) Epoch 27, batch 8050, loss[loss=0.2052, simple_loss=0.2806, pruned_loss=0.06489, over 11734.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2885, pruned_loss=0.05749, over 3100082.41 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:33:51,588 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4160, 4.5273, 4.7020, 4.4884, 4.5820, 5.0500, 4.5748, 4.2881], device='cuda:3'), covar=tensor([0.1470, 0.1971, 0.1928, 0.1940, 0.2329, 0.1044, 0.1627, 0.2502], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0630, 0.0692, 0.0513, 0.0684, 0.0721, 0.0543, 0.0689], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:34:05,653 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:34:09,132 INFO [optim.py:368] (3/8) 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 10:34:16,642 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6027, 1.7882, 2.2341, 2.5467, 2.5319, 2.9807, 1.8927, 2.9013], device='cuda:3'), covar=tensor([0.0249, 0.0557, 0.0346, 0.0379, 0.0351, 0.0179, 0.0588, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0190, 0.0206, 0.0165, 0.0202, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:34:18,542 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0312, 2.3761, 2.3343, 2.7792, 1.8974, 3.2113, 1.8158, 2.7059], device='cuda:3'), covar=tensor([0.1193, 0.0631, 0.1058, 0.0216, 0.0129, 0.0383, 0.1530, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0201, 0.0208, 0.0218, 0.0209, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 10:34:38,715 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 10:34:49,658 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:35:04,670 INFO [train.py:904] (3/8) Epoch 27, batch 8100, loss[loss=0.227, simple_loss=0.2962, pruned_loss=0.07891, over 11445.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2875, pruned_loss=0.05697, over 3091936.61 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:35:06,839 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 10:35:45,687 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:36:17,761 INFO [train.py:904] (3/8) Epoch 27, batch 8150, loss[loss=0.1804, simple_loss=0.2701, pruned_loss=0.04533, over 16315.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2848, pruned_loss=0.05588, over 3093141.49 frames. ], batch size: 35, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:36:39,765 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:01,607 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:32,132 INFO [train.py:904] (3/8) Epoch 27, batch 8200, loss[loss=0.1901, simple_loss=0.2871, pruned_loss=0.04659, over 15349.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2819, pruned_loss=0.05478, over 3102490.77 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:38:12,503 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:16,441 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:34,468 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5913, 3.6449, 3.3895, 3.0384, 3.2226, 3.5357, 3.3357, 3.3426], device='cuda:3'), covar=tensor([0.0553, 0.0609, 0.0328, 0.0289, 0.0570, 0.0456, 0.1538, 0.0499], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0454, 0.0353, 0.0354, 0.0352, 0.0407, 0.0243, 0.0422], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:38:53,589 INFO [train.py:904] (3/8) Epoch 27, batch 8250, loss[loss=0.1465, simple_loss=0.2367, pruned_loss=0.02817, over 11931.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2811, pruned_loss=0.05249, over 3075495.79 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:39:09,979 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1483, 1.5831, 1.9757, 2.0788, 2.1986, 2.3255, 1.7973, 2.2661], device='cuda:3'), covar=tensor([0.0293, 0.0559, 0.0317, 0.0427, 0.0399, 0.0253, 0.0555, 0.0200], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0190, 0.0205, 0.0164, 0.0201, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 10:39:19,037 INFO [optim.py:368] (3/8) 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 10:39:40,162 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272181.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:39:50,803 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272188.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:39:52,266 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8926, 4.5009, 3.3451, 2.5245, 2.8163, 2.7719, 4.8878, 3.7497], device='cuda:3'), covar=tensor([0.3068, 0.0440, 0.1691, 0.2870, 0.3007, 0.2146, 0.0295, 0.1310], device='cuda:3'), in_proj_covar=tensor([0.0334, 0.0273, 0.0311, 0.0325, 0.0305, 0.0274, 0.0303, 0.0348], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:40:14,329 INFO [train.py:904] (3/8) Epoch 27, batch 8300, loss[loss=0.1726, simple_loss=0.2722, pruned_loss=0.03651, over 16470.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2782, pruned_loss=0.04947, over 3055979.89 frames. ], batch size: 75, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:40:26,347 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:46,124 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:15,136 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272242.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 10:41:22,868 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272247.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:31,412 INFO [train.py:904] (3/8) Epoch 27, batch 8350, loss[loss=0.1918, simple_loss=0.2862, pruned_loss=0.04869, over 16226.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2775, pruned_loss=0.04779, over 3048718.95 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:41:54,864 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.144e+02 2.517e+02 3.037e+02 5.148e+02, threshold=5.034e+02, percent-clipped=0.0 2023-05-02 10:42:20,306 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272284.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:27,559 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6387, 3.7032, 2.8509, 2.2634, 2.3099, 2.4866, 3.9364, 3.2597], device='cuda:3'), covar=tensor([0.2907, 0.0576, 0.1831, 0.3081, 0.2946, 0.2177, 0.0418, 0.1321], device='cuda:3'), in_proj_covar=tensor([0.0331, 0.0271, 0.0309, 0.0322, 0.0302, 0.0272, 0.0301, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:42:30,777 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 8400, loss[loss=0.1476, simple_loss=0.2405, pruned_loss=0.02734, over 12147.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2742, pruned_loss=0.04544, over 3044396.08 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:42:59,284 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:59,556 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-02 10:44:09,889 INFO [train.py:904] (3/8) Epoch 27, batch 8450, loss[loss=0.1729, simple_loss=0.2692, pruned_loss=0.03827, over 15348.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2726, pruned_loss=0.04358, over 3063717.41 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:44:11,702 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5402, 3.4995, 2.7749, 2.2039, 2.1994, 2.3932, 3.7075, 3.0979], device='cuda:3'), covar=tensor([0.2965, 0.0617, 0.1840, 0.3131, 0.2969, 0.2292, 0.0422, 0.1436], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0270, 0.0307, 0.0321, 0.0300, 0.0271, 0.0299, 0.0343], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:44:34,140 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 2.148e+02 2.573e+02 3.182e+02 7.232e+02, threshold=5.146e+02, percent-clipped=1.0 2023-05-02 10:44:42,774 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272373.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:45:32,397 INFO [train.py:904] (3/8) Epoch 27, batch 8500, loss[loss=0.1537, simple_loss=0.2475, pruned_loss=0.02998, over 11931.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2691, pruned_loss=0.04162, over 3053357.29 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:46:24,066 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:46:57,066 INFO [train.py:904] (3/8) Epoch 27, batch 8550, loss[loss=0.1881, simple_loss=0.2794, pruned_loss=0.04836, over 16786.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2669, pruned_loss=0.04089, over 3037960.32 frames. ], batch size: 124, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:47:25,000 INFO [optim.py:368] (3/8) 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 10:47:55,109 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272483.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:15,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0887, 2.5183, 2.6631, 1.9141, 2.7727, 2.8474, 2.5502, 2.4983], device='cuda:3'), covar=tensor([0.0710, 0.0271, 0.0254, 0.0998, 0.0145, 0.0238, 0.0446, 0.0447], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0108, 0.0099, 0.0135, 0.0084, 0.0127, 0.0127, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 10:48:35,725 INFO [train.py:904] (3/8) Epoch 27, batch 8600, loss[loss=0.1489, simple_loss=0.2539, pruned_loss=0.02199, over 16578.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2667, pruned_loss=0.03992, over 3026918.21 frames. ], batch size: 57, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:48:50,697 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272510.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:55,726 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-05-02 10:49:16,261 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 10:49:43,697 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272537.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 10:50:13,934 INFO [train.py:904] (3/8) Epoch 27, batch 8650, loss[loss=0.1637, simple_loss=0.2626, pruned_loss=0.03238, over 15284.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2648, pruned_loss=0.03894, over 3011257.67 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:50:26,982 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:50:46,999 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 10:50:48,990 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.181e+02 2.564e+02 3.046e+02 5.835e+02, threshold=5.129e+02, percent-clipped=3.0 2023-05-02 10:51:15,178 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:51:36,584 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:51:44,320 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3538, 3.5843, 3.5861, 2.5543, 3.2682, 3.5930, 3.4359, 2.0674], device='cuda:3'), covar=tensor([0.0498, 0.0070, 0.0056, 0.0393, 0.0126, 0.0100, 0.0085, 0.0536], device='cuda:3'), in_proj_covar=tensor([0.0134, 0.0087, 0.0088, 0.0133, 0.0098, 0.0111, 0.0095, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 10:52:00,243 INFO [train.py:904] (3/8) Epoch 27, batch 8700, loss[loss=0.1617, simple_loss=0.2569, pruned_loss=0.03331, over 16682.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2626, pruned_loss=0.03773, over 3032341.19 frames. ], batch size: 83, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:52:01,200 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:21,585 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:53:06,582 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272638.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:53:35,412 INFO [train.py:904] (3/8) Epoch 27, batch 8750, loss[loss=0.1759, simple_loss=0.2719, pruned_loss=0.03994, over 16514.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2625, pruned_loss=0.03727, over 3035266.00 frames. ], batch size: 68, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 10:54:15,877 INFO [optim.py:368] (3/8) 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 10:54:28,882 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:55:27,068 INFO [train.py:904] (3/8) Epoch 27, batch 8800, loss[loss=0.1825, simple_loss=0.2724, pruned_loss=0.0463, over 12326.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2608, pruned_loss=0.03611, over 3048810.03 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:56:23,109 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 8850, loss[loss=0.1472, simple_loss=0.2428, pruned_loss=0.02578, over 12588.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2634, pruned_loss=0.03555, over 3037968.35 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:57:25,519 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-05-02 10:57:46,547 INFO [optim.py:368] (3/8) 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 10:58:17,186 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272783.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:58:25,745 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-05-02 10:58:57,762 INFO [train.py:904] (3/8) Epoch 27, batch 8900, loss[loss=0.1765, simple_loss=0.2745, pruned_loss=0.03929, over 16239.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2645, pruned_loss=0.03491, over 3055758.07 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:00:01,533 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:00:20,809 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272837.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 11:01:00,397 INFO [train.py:904] (3/8) Epoch 27, batch 8950, loss[loss=0.182, simple_loss=0.2704, pruned_loss=0.04682, over 12361.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2645, pruned_loss=0.03538, over 3063536.63 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:01:35,337 INFO [optim.py:368] (3/8) 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 11:01:37,113 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6889, 4.9939, 4.8253, 4.8123, 4.5483, 4.4969, 4.4356, 5.0800], device='cuda:3'), covar=tensor([0.1210, 0.0894, 0.0897, 0.0836, 0.0787, 0.1156, 0.1240, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0692, 0.0838, 0.0688, 0.0643, 0.0527, 0.0529, 0.0701, 0.0656], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:01:57,014 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272885.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 11:02:49,712 INFO [train.py:904] (3/8) Epoch 27, batch 9000, loss[loss=0.1505, simple_loss=0.2441, pruned_loss=0.0284, over 16728.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2613, pruned_loss=0.03424, over 3073147.22 frames. ], batch size: 134, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:02:49,712 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 11:02:56,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2399, 2.4612, 2.8216, 3.3277, 3.0258, 3.7649, 2.7004, 3.6975], device='cuda:3'), covar=tensor([0.0267, 0.0530, 0.0412, 0.0288, 0.0379, 0.0173, 0.0495, 0.0176], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0196, 0.0183, 0.0187, 0.0204, 0.0162, 0.0199, 0.0162], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:02:59,738 INFO [train.py:938] (3/8) Epoch 27, validation: loss=0.1436, simple_loss=0.2474, pruned_loss=0.01989, over 944034.00 frames. 2023-05-02 11:02:59,739 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 11:03:00,808 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:03:51,479 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:02,250 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4782, 4.5768, 4.6684, 4.5070, 4.6127, 5.0668, 4.6378, 4.3611], device='cuda:3'), covar=tensor([0.1336, 0.1960, 0.2094, 0.1914, 0.2197, 0.0891, 0.1487, 0.2232], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0613, 0.0672, 0.0501, 0.0665, 0.0706, 0.0530, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 11:04:42,534 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:44,607 INFO [train.py:904] (3/8) Epoch 27, batch 9050, loss[loss=0.1692, simple_loss=0.2607, pruned_loss=0.03885, over 12753.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2626, pruned_loss=0.03477, over 3075912.65 frames. ], batch size: 250, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:05:18,776 INFO [optim.py:368] (3/8) 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 11:05:19,380 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:05:21,236 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8520, 3.9526, 4.0243, 3.8032, 3.9888, 4.3518, 4.0227, 3.6771], device='cuda:3'), covar=tensor([0.2058, 0.2150, 0.2116, 0.2244, 0.2439, 0.1436, 0.1661, 0.2562], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0613, 0.0671, 0.0500, 0.0665, 0.0706, 0.0530, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 11:06:32,229 INFO [train.py:904] (3/8) Epoch 27, batch 9100, loss[loss=0.1786, simple_loss=0.2745, pruned_loss=0.04141, over 16961.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2618, pruned_loss=0.03521, over 3078501.58 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:07:36,101 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273029.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:08:32,211 INFO [train.py:904] (3/8) Epoch 27, batch 9150, loss[loss=0.1477, simple_loss=0.2475, pruned_loss=0.02395, over 16910.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2618, pruned_loss=0.03498, over 3058646.06 frames. ], batch size: 96, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:09:08,126 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273077.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:09:40,791 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 11:10:18,353 INFO [train.py:904] (3/8) Epoch 27, batch 9200, loss[loss=0.1491, simple_loss=0.244, pruned_loss=0.02705, over 16568.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.258, pruned_loss=0.03413, over 3070470.78 frames. ], batch size: 75, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:11:45,576 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4879, 3.4220, 3.5296, 3.5798, 3.6372, 3.3416, 3.6111, 3.6765], device='cuda:3'), covar=tensor([0.1176, 0.0884, 0.0974, 0.0605, 0.0618, 0.2241, 0.0837, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0640, 0.0781, 0.0907, 0.0793, 0.0609, 0.0634, 0.0665, 0.0775], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:11:53,368 INFO [train.py:904] (3/8) Epoch 27, batch 9250, loss[loss=0.1388, simple_loss=0.2278, pruned_loss=0.02491, over 12020.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2578, pruned_loss=0.0344, over 3048918.34 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:11:58,726 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5096, 4.5179, 4.8662, 4.8374, 4.8341, 4.5798, 4.5520, 4.5510], device='cuda:3'), covar=tensor([0.0650, 0.1401, 0.0920, 0.1220, 0.1179, 0.1274, 0.1362, 0.0598], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0468, 0.0454, 0.0418, 0.0501, 0.0479, 0.0549, 0.0385], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 11:11:58,752 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5130, 4.3442, 4.5101, 4.6719, 4.8416, 4.4099, 4.8178, 4.8421], device='cuda:3'), covar=tensor([0.1854, 0.1333, 0.1702, 0.0822, 0.0549, 0.1118, 0.0600, 0.0768], device='cuda:3'), in_proj_covar=tensor([0.0640, 0.0782, 0.0907, 0.0793, 0.0609, 0.0634, 0.0665, 0.0776], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:11:58,811 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4916, 3.9899, 3.9663, 2.7557, 3.4980, 3.9995, 3.6794, 2.4901], device='cuda:3'), covar=tensor([0.0489, 0.0043, 0.0046, 0.0399, 0.0111, 0.0076, 0.0078, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0087, 0.0088, 0.0133, 0.0099, 0.0112, 0.0095, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 11:12:25,531 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.225e+02 2.595e+02 3.229e+02 5.341e+02, threshold=5.191e+02, percent-clipped=1.0 2023-05-02 11:13:43,821 INFO [train.py:904] (3/8) Epoch 27, batch 9300, loss[loss=0.1405, simple_loss=0.2372, pruned_loss=0.02193, over 16424.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2559, pruned_loss=0.03382, over 3026564.40 frames. ], batch size: 68, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:15:28,949 INFO [train.py:904] (3/8) Epoch 27, batch 9350, loss[loss=0.1686, simple_loss=0.2615, pruned_loss=0.03786, over 16967.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.256, pruned_loss=0.0338, over 3032764.92 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:16:02,980 INFO [optim.py:368] (3/8) 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 11:16:03,980 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:16:53,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6621, 3.7105, 3.5153, 3.2255, 3.3453, 3.6218, 3.4061, 3.4365], device='cuda:3'), covar=tensor([0.0569, 0.0598, 0.0325, 0.0276, 0.0514, 0.0496, 0.1316, 0.0494], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0447, 0.0349, 0.0351, 0.0345, 0.0404, 0.0241, 0.0418], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:17:10,294 INFO [train.py:904] (3/8) Epoch 27, batch 9400, loss[loss=0.1755, simple_loss=0.2778, pruned_loss=0.03664, over 16386.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2559, pruned_loss=0.0332, over 3045347.29 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:17:32,689 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2648, 4.3905, 4.5020, 4.2674, 4.4264, 4.8589, 4.4352, 4.1330], device='cuda:3'), covar=tensor([0.1654, 0.1988, 0.2443, 0.2153, 0.2265, 0.0971, 0.1556, 0.2341], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0606, 0.0666, 0.0494, 0.0657, 0.0699, 0.0525, 0.0657], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 11:17:39,075 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:17:41,531 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 11:18:52,055 INFO [train.py:904] (3/8) Epoch 27, batch 9450, loss[loss=0.1722, simple_loss=0.2658, pruned_loss=0.03933, over 15412.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2576, pruned_loss=0.03359, over 3032661.33 frames. ], batch size: 192, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:19:21,966 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.197e+02 2.626e+02 3.206e+02 9.861e+02, threshold=5.251e+02, percent-clipped=1.0 2023-05-02 11:20:32,362 INFO [train.py:904] (3/8) Epoch 27, batch 9500, loss[loss=0.1618, simple_loss=0.2554, pruned_loss=0.03407, over 15345.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2572, pruned_loss=0.03353, over 3035573.82 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:20:49,216 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273409.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:21:38,390 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:21:40,832 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6256, 3.6911, 3.4873, 3.1362, 3.3124, 3.5715, 3.3771, 3.4078], device='cuda:3'), covar=tensor([0.0575, 0.0587, 0.0301, 0.0292, 0.0514, 0.0471, 0.1566, 0.0476], device='cuda:3'), in_proj_covar=tensor([0.0295, 0.0443, 0.0346, 0.0348, 0.0343, 0.0399, 0.0238, 0.0413], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:22:18,627 INFO [train.py:904] (3/8) Epoch 27, batch 9550, loss[loss=0.149, simple_loss=0.2504, pruned_loss=0.02376, over 16753.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2565, pruned_loss=0.03329, over 3042934.79 frames. ], batch size: 83, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:22:51,234 INFO [optim.py:368] (3/8) 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 11:22:53,703 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:46,477 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:57,854 INFO [train.py:904] (3/8) Epoch 27, batch 9600, loss[loss=0.1908, simple_loss=0.2921, pruned_loss=0.04472, over 16143.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2576, pruned_loss=0.03375, over 3039594.84 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:24:01,865 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6154, 3.5702, 3.5400, 2.8369, 3.4698, 1.9998, 3.3013, 2.9223], device='cuda:3'), covar=tensor([0.0153, 0.0136, 0.0196, 0.0230, 0.0112, 0.2702, 0.0139, 0.0318], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0167, 0.0203, 0.0176, 0.0180, 0.0210, 0.0192, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:25:41,458 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-05-02 11:25:44,211 INFO [train.py:904] (3/8) Epoch 27, batch 9650, loss[loss=0.1589, simple_loss=0.2515, pruned_loss=0.0331, over 16555.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2593, pruned_loss=0.03429, over 3030433.66 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:26:23,245 INFO [optim.py:368] (3/8) 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 11:27:30,808 INFO [train.py:904] (3/8) Epoch 27, batch 9700, loss[loss=0.2007, simple_loss=0.2764, pruned_loss=0.0625, over 12388.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2591, pruned_loss=0.0342, over 3062366.09 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:12,471 INFO [train.py:904] (3/8) Epoch 27, batch 9750, loss[loss=0.1668, simple_loss=0.2665, pruned_loss=0.03356, over 16518.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2579, pruned_loss=0.03423, over 3062160.84 frames. ], batch size: 68, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:42,115 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 2.115e+02 2.683e+02 3.309e+02 9.068e+02, threshold=5.365e+02, percent-clipped=4.0 2023-05-02 11:30:51,288 INFO [train.py:904] (3/8) Epoch 27, batch 9800, loss[loss=0.1586, simple_loss=0.2627, pruned_loss=0.02723, over 17211.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2578, pruned_loss=0.03336, over 3056908.26 frames. ], batch size: 46, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:31:16,873 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-05-02 11:32:36,073 INFO [train.py:904] (3/8) Epoch 27, batch 9850, loss[loss=0.166, simple_loss=0.2663, pruned_loss=0.03287, over 16635.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2587, pruned_loss=0.03309, over 3046771.53 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:32:43,940 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 11:33:00,285 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:33:11,081 INFO [optim.py:368] (3/8) 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 11:33:36,808 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.09 vs. limit=5.0 2023-05-02 11:34:02,560 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 9900, loss[loss=0.1617, simple_loss=0.2643, pruned_loss=0.02952, over 15470.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2592, pruned_loss=0.03315, over 3057790.88 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:35:12,353 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6002, 4.7405, 4.8579, 4.6481, 4.7192, 5.2015, 4.8117, 4.5392], device='cuda:3'), covar=tensor([0.1316, 0.1939, 0.2084, 0.2182, 0.2623, 0.1002, 0.1495, 0.2345], device='cuda:3'), in_proj_covar=tensor([0.0402, 0.0601, 0.0658, 0.0488, 0.0651, 0.0692, 0.0518, 0.0649], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 11:36:11,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4109, 3.4799, 3.6764, 3.6704, 3.6764, 3.4981, 3.5293, 3.5780], device='cuda:3'), covar=tensor([0.0569, 0.1016, 0.0703, 0.0630, 0.0683, 0.0849, 0.0958, 0.0695], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0463, 0.0452, 0.0414, 0.0498, 0.0476, 0.0544, 0.0381], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 11:36:22,574 INFO [train.py:904] (3/8) Epoch 27, batch 9950, loss[loss=0.1657, simple_loss=0.2623, pruned_loss=0.03455, over 16311.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2608, pruned_loss=0.03351, over 3033789.67 frames. ], batch size: 35, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:37:03,257 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 27, batch 10000, loss[loss=0.1499, simple_loss=0.2533, pruned_loss=0.02323, over 16675.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.26, pruned_loss=0.0331, over 3049442.51 frames. ], batch size: 89, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:40:06,630 INFO [train.py:904] (3/8) Epoch 27, batch 10050, loss[loss=0.1534, simple_loss=0.2538, pruned_loss=0.0265, over 16935.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2603, pruned_loss=0.0331, over 3052911.41 frames. ], batch size: 96, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:40:24,789 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2578, 4.3296, 4.4609, 4.2102, 4.3627, 4.8263, 4.3793, 4.0735], device='cuda:3'), covar=tensor([0.1633, 0.2065, 0.2315, 0.2442, 0.2538, 0.1041, 0.1801, 0.2779], device='cuda:3'), in_proj_covar=tensor([0.0404, 0.0604, 0.0663, 0.0492, 0.0654, 0.0697, 0.0523, 0.0654], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 11:40:39,176 INFO [optim.py:368] (3/8) 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 11:41:39,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0934, 3.2562, 3.7549, 2.2038, 3.0760, 2.3529, 3.5571, 3.4488], device='cuda:3'), covar=tensor([0.0252, 0.0850, 0.0499, 0.2116, 0.0800, 0.1024, 0.0670, 0.1093], device='cuda:3'), in_proj_covar=tensor([0.0155, 0.0162, 0.0163, 0.0152, 0.0143, 0.0128, 0.0140, 0.0173], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 11:41:40,583 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 27, batch 10100, loss[loss=0.1631, simple_loss=0.2587, pruned_loss=0.03369, over 15291.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.261, pruned_loss=0.03342, over 3059741.83 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:42:12,010 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-05-02 11:43:27,165 INFO [train.py:904] (3/8) Epoch 28, batch 0, loss[loss=0.1973, simple_loss=0.2711, pruned_loss=0.06178, over 16820.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2711, pruned_loss=0.06178, over 16820.00 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 8.0 2023-05-02 11:43:27,165 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 11:43:34,595 INFO [train.py:938] (3/8) Epoch 28, validation: loss=0.1434, simple_loss=0.2464, pruned_loss=0.02022, over 944034.00 frames. 2023-05-02 11:43:34,596 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 11:43:48,030 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274062.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:43:52,225 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:01,205 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.305e+02 2.657e+02 3.293e+02 5.505e+02, threshold=5.314e+02, percent-clipped=2.0 2023-05-02 11:44:13,583 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9085, 2.5497, 2.4740, 4.0844, 3.1839, 3.9608, 1.5872, 2.9693], device='cuda:3'), covar=tensor([0.1413, 0.0756, 0.1305, 0.0152, 0.0118, 0.0400, 0.1669, 0.0844], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0176, 0.0197, 0.0194, 0.0199, 0.0213, 0.0206, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 11:44:27,478 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 50, loss[loss=0.1623, simple_loss=0.2439, pruned_loss=0.04034, over 16733.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2651, pruned_loss=0.04286, over 753213.28 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:44:59,065 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274113.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:24,193 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:34,056 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 100, loss[loss=0.198, simple_loss=0.274, pruned_loss=0.06101, over 16832.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2604, pruned_loss=0.04227, over 1322957.75 frames. ], batch size: 116, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:46:20,295 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.302e+02 2.829e+02 3.561e+02 6.259e+02, threshold=5.657e+02, percent-clipped=1.0 2023-05-02 11:46:23,013 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9442, 5.0260, 5.4416, 5.4075, 5.4226, 5.0473, 4.9556, 4.8210], device='cuda:3'), covar=tensor([0.0442, 0.0546, 0.0366, 0.0420, 0.0529, 0.0485, 0.1119, 0.0495], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0468, 0.0456, 0.0419, 0.0502, 0.0481, 0.0549, 0.0385], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 11:46:27,163 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0179, 2.9938, 2.6678, 4.9952, 4.0647, 4.4350, 1.8755, 3.2299], device='cuda:3'), covar=tensor([0.1389, 0.0814, 0.1370, 0.0222, 0.0246, 0.0411, 0.1657, 0.0813], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0177, 0.0199, 0.0196, 0.0201, 0.0215, 0.0208, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 11:46:48,042 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:47:00,098 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0185, 4.7493, 5.0170, 5.2156, 5.4073, 4.7118, 5.3933, 5.4163], device='cuda:3'), covar=tensor([0.1999, 0.1431, 0.1861, 0.0834, 0.0590, 0.1134, 0.0583, 0.0667], device='cuda:3'), in_proj_covar=tensor([0.0642, 0.0784, 0.0906, 0.0798, 0.0611, 0.0635, 0.0668, 0.0778], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:47:02,023 INFO [train.py:904] (3/8) Epoch 28, batch 150, loss[loss=0.1458, simple_loss=0.2443, pruned_loss=0.02367, over 17226.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2583, pruned_loss=0.04064, over 1758096.05 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:47:17,173 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 11:47:29,046 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8953, 4.6974, 4.8519, 5.1082, 5.2901, 4.7487, 5.2808, 5.2947], device='cuda:3'), covar=tensor([0.2298, 0.1638, 0.2301, 0.1085, 0.0882, 0.0965, 0.0875, 0.0835], device='cuda:3'), in_proj_covar=tensor([0.0642, 0.0783, 0.0906, 0.0797, 0.0610, 0.0635, 0.0667, 0.0777], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:48:08,591 INFO [train.py:904] (3/8) Epoch 28, batch 200, loss[loss=0.1867, simple_loss=0.283, pruned_loss=0.04519, over 16715.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2583, pruned_loss=0.04114, over 2108113.78 frames. ], batch size: 62, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:48:30,271 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:48:33,860 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.228e+02 2.594e+02 3.067e+02 5.865e+02, threshold=5.188e+02, percent-clipped=1.0 2023-05-02 11:48:50,996 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 11:49:11,515 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5354, 2.4300, 2.5088, 4.4343, 2.4332, 2.7939, 2.5128, 2.6053], device='cuda:3'), covar=tensor([0.1350, 0.3920, 0.3257, 0.0533, 0.4348, 0.2797, 0.3891, 0.3836], device='cuda:3'), in_proj_covar=tensor([0.0415, 0.0468, 0.0382, 0.0331, 0.0442, 0.0532, 0.0440, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:49:16,127 INFO [train.py:904] (3/8) Epoch 28, batch 250, loss[loss=0.1463, simple_loss=0.2371, pruned_loss=0.02777, over 17200.00 frames. ], tot_loss[loss=0.17, simple_loss=0.257, pruned_loss=0.04153, over 2376861.30 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:49:16,554 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274303.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:49:52,020 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274330.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:05,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2384, 4.6048, 2.9873, 2.7487, 2.9842, 2.6452, 4.8974, 3.5863], device='cuda:3'), covar=tensor([0.2815, 0.0667, 0.2119, 0.2866, 0.3005, 0.2465, 0.0458, 0.1638], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0271, 0.0308, 0.0321, 0.0298, 0.0272, 0.0299, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 11:50:23,021 INFO [train.py:904] (3/8) Epoch 28, batch 300, loss[loss=0.1506, simple_loss=0.2336, pruned_loss=0.03381, over 16268.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2544, pruned_loss=0.04042, over 2579441.67 frames. ], batch size: 36, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:50:29,899 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:39,314 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:49,961 INFO [optim.py:368] (3/8) 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 11:51:02,894 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 11:51:31,008 INFO [train.py:904] (3/8) Epoch 28, batch 350, loss[loss=0.1323, simple_loss=0.2188, pruned_loss=0.02293, over 16847.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2522, pruned_loss=0.03966, over 2732886.07 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:51:31,753 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-02 11:52:37,185 INFO [train.py:904] (3/8) Epoch 28, batch 400, loss[loss=0.1492, simple_loss=0.2462, pruned_loss=0.0261, over 17192.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2508, pruned_loss=0.03948, over 2872444.43 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:52:46,813 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 11:53:03,891 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.265e+02 2.649e+02 3.069e+02 5.270e+02, threshold=5.298e+02, percent-clipped=3.0 2023-05-02 11:53:04,356 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0401, 5.1345, 5.5455, 5.4899, 5.5286, 5.1898, 5.1061, 4.9312], device='cuda:3'), covar=tensor([0.0394, 0.0552, 0.0370, 0.0440, 0.0430, 0.0475, 0.0938, 0.0464], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0480, 0.0466, 0.0429, 0.0513, 0.0492, 0.0562, 0.0395], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 11:53:24,165 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:53:44,148 INFO [train.py:904] (3/8) Epoch 28, batch 450, loss[loss=0.1623, simple_loss=0.2438, pruned_loss=0.04043, over 12412.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2502, pruned_loss=0.03946, over 2975430.24 frames. ], batch size: 247, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:54:35,063 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:54:53,415 INFO [train.py:904] (3/8) Epoch 28, batch 500, loss[loss=0.1769, simple_loss=0.2558, pruned_loss=0.04903, over 16464.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2495, pruned_loss=0.03872, over 3063196.11 frames. ], batch size: 75, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:55:21,738 INFO [optim.py:368] (3/8) 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 11:55:58,854 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 550, loss[loss=0.1529, simple_loss=0.2329, pruned_loss=0.03641, over 16759.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2489, pruned_loss=0.03794, over 3112875.27 frames. ], batch size: 83, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:56:30,880 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274625.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:02,646 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 11:57:09,863 INFO [train.py:904] (3/8) Epoch 28, batch 600, loss[loss=0.1421, simple_loss=0.2303, pruned_loss=0.02694, over 16771.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2477, pruned_loss=0.03797, over 3158661.66 frames. ], batch size: 39, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:57:15,245 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:17,484 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:36,633 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.130e+02 2.496e+02 2.954e+02 2.616e+03, threshold=4.992e+02, percent-clipped=3.0 2023-05-02 11:58:16,066 INFO [train.py:904] (3/8) Epoch 28, batch 650, loss[loss=0.1671, simple_loss=0.2411, pruned_loss=0.04654, over 16855.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2471, pruned_loss=0.03779, over 3197135.18 frames. ], batch size: 96, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:58:18,638 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:58:42,113 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3177, 5.2866, 5.0518, 4.4585, 5.0967, 1.8295, 4.8892, 4.9707], device='cuda:3'), covar=tensor([0.0109, 0.0121, 0.0248, 0.0455, 0.0135, 0.2963, 0.0162, 0.0241], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0170, 0.0207, 0.0179, 0.0184, 0.0214, 0.0196, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 11:59:25,786 INFO [train.py:904] (3/8) Epoch 28, batch 700, loss[loss=0.1672, simple_loss=0.2489, pruned_loss=0.04274, over 16285.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2468, pruned_loss=0.03789, over 3217606.72 frames. ], batch size: 165, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:59:54,342 INFO [optim.py:368] (3/8) 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 12:00:13,256 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:00:14,402 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2094, 4.0308, 4.2860, 4.3954, 4.4754, 4.0822, 4.3376, 4.4697], device='cuda:3'), covar=tensor([0.1632, 0.1311, 0.1311, 0.0731, 0.0641, 0.1344, 0.2225, 0.0773], device='cuda:3'), in_proj_covar=tensor([0.0665, 0.0813, 0.0940, 0.0824, 0.0630, 0.0656, 0.0690, 0.0803], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:00:30,105 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2651, 5.9545, 6.0594, 5.7269, 5.9142, 6.4033, 5.8529, 5.5312], device='cuda:3'), covar=tensor([0.0896, 0.1946, 0.2297, 0.2336, 0.2719, 0.1019, 0.1701, 0.2391], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0638, 0.0699, 0.0517, 0.0690, 0.0727, 0.0549, 0.0684], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 12:00:34,709 INFO [train.py:904] (3/8) Epoch 28, batch 750, loss[loss=0.1664, simple_loss=0.2573, pruned_loss=0.03777, over 17217.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2475, pruned_loss=0.03805, over 3239314.02 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 12:01:19,359 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 800, loss[loss=0.1744, simple_loss=0.2625, pruned_loss=0.04319, over 16580.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2479, pruned_loss=0.03831, over 3252272.58 frames. ], batch size: 62, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:02:15,002 INFO [optim.py:368] (3/8) 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 12:02:43,383 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:45,550 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 12:02:46,201 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 850, loss[loss=0.1551, simple_loss=0.2475, pruned_loss=0.03133, over 17143.00 frames. ], tot_loss[loss=0.161, simple_loss=0.247, pruned_loss=0.03753, over 3270860.68 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:03:26,691 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274925.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:03:50,036 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8250, 3.8065, 3.8207, 3.6566, 3.7993, 4.2810, 3.8393, 3.5248], device='cuda:3'), covar=tensor([0.2043, 0.2316, 0.2640, 0.2602, 0.3079, 0.1824, 0.1913, 0.2948], device='cuda:3'), in_proj_covar=tensor([0.0424, 0.0637, 0.0700, 0.0517, 0.0690, 0.0726, 0.0548, 0.0685], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 12:04:05,739 INFO [train.py:904] (3/8) Epoch 28, batch 900, loss[loss=0.1589, simple_loss=0.2352, pruned_loss=0.04132, over 16828.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2461, pruned_loss=0.03699, over 3283165.92 frames. ], batch size: 83, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:04:09,215 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:14,509 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274959.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:19,048 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.48 vs. limit=5.0 2023-05-02 12:04:20,103 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9807, 2.1646, 2.6190, 2.9871, 2.7785, 3.4121, 2.4477, 3.4617], device='cuda:3'), covar=tensor([0.0297, 0.0542, 0.0388, 0.0366, 0.0427, 0.0251, 0.0550, 0.0205], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0200, 0.0189, 0.0194, 0.0209, 0.0168, 0.0206, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 12:04:21,292 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9838, 2.1356, 2.5339, 2.8609, 2.7463, 3.1270, 2.2874, 3.2147], device='cuda:3'), covar=tensor([0.0266, 0.0566, 0.0403, 0.0388, 0.0411, 0.0354, 0.0593, 0.0207], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0200, 0.0189, 0.0194, 0.0209, 0.0168, 0.0206, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 12:04:33,594 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:04:47,134 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-02 12:05:14,559 INFO [train.py:904] (3/8) Epoch 28, batch 950, loss[loss=0.1594, simple_loss=0.2374, pruned_loss=0.04068, over 16807.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2456, pruned_loss=0.03686, over 3294361.98 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:05:20,983 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275007.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:06:23,168 INFO [train.py:904] (3/8) Epoch 28, batch 1000, loss[loss=0.1547, simple_loss=0.232, pruned_loss=0.03868, over 16424.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2447, pruned_loss=0.0366, over 3307904.52 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:06:28,148 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0874, 3.0905, 1.7490, 3.2451, 2.4285, 3.2721, 1.9351, 2.6315], device='cuda:3'), covar=tensor([0.0356, 0.0448, 0.2006, 0.0402, 0.0894, 0.0624, 0.2008, 0.0808], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0172, 0.0179, 0.0220, 0.0206, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:06:52,343 INFO [optim.py:368] (3/8) 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 12:07:31,782 INFO [train.py:904] (3/8) Epoch 28, batch 1050, loss[loss=0.1673, simple_loss=0.2435, pruned_loss=0.04553, over 16825.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2447, pruned_loss=0.03659, over 3307900.81 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:09,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5276, 3.5495, 4.1311, 2.2128, 3.2692, 2.5736, 3.9857, 3.7795], device='cuda:3'), covar=tensor([0.0253, 0.1124, 0.0496, 0.2167, 0.0860, 0.1029, 0.0571, 0.1160], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0157, 0.0148, 0.0133, 0.0147, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 12:08:39,620 INFO [train.py:904] (3/8) Epoch 28, batch 1100, loss[loss=0.1593, simple_loss=0.2356, pruned_loss=0.04147, over 16445.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2443, pruned_loss=0.03635, over 3314613.93 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:47,424 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:07,966 INFO [optim.py:368] (3/8) 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 12:09:39,579 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:49,907 INFO [train.py:904] (3/8) Epoch 28, batch 1150, loss[loss=0.1558, simple_loss=0.2494, pruned_loss=0.03108, over 17169.00 frames. ], tot_loss[loss=0.157, simple_loss=0.2431, pruned_loss=0.03546, over 3323465.22 frames. ], batch size: 46, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:09:52,293 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 12:10:10,243 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8921, 4.8905, 4.7229, 4.1840, 4.8263, 2.0054, 4.5678, 4.4197], device='cuda:3'), covar=tensor([0.0140, 0.0113, 0.0199, 0.0348, 0.0108, 0.2688, 0.0146, 0.0234], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0174, 0.0210, 0.0182, 0.0187, 0.0217, 0.0200, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:10:11,465 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:30,483 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6267, 2.3963, 1.9472, 2.2470, 2.7559, 2.5135, 2.6482, 2.8385], device='cuda:3'), covar=tensor([0.0279, 0.0518, 0.0642, 0.0508, 0.0305, 0.0410, 0.0257, 0.0339], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0250, 0.0238, 0.0238, 0.0251, 0.0248, 0.0247, 0.0247], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:10:44,954 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:53,439 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:57,365 INFO [train.py:904] (3/8) Epoch 28, batch 1200, loss[loss=0.1621, simple_loss=0.2476, pruned_loss=0.03833, over 16847.00 frames. ], tot_loss[loss=0.1562, simple_loss=0.2423, pruned_loss=0.03499, over 3320551.66 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:11:20,308 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7292, 4.1597, 4.4035, 2.2637, 4.6321, 4.8447, 3.5326, 3.4579], device='cuda:3'), covar=tensor([0.1151, 0.0214, 0.0249, 0.1397, 0.0084, 0.0128, 0.0444, 0.0576], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0112, 0.0103, 0.0142, 0.0087, 0.0133, 0.0132, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:11:21,633 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 12:11:26,948 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 1250, loss[loss=0.1672, simple_loss=0.2462, pruned_loss=0.04416, over 16676.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2433, pruned_loss=0.03563, over 3318390.89 frames. ], batch size: 89, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:12:12,672 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:08,418 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275347.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:13:17,201 INFO [train.py:904] (3/8) Epoch 28, batch 1300, loss[loss=0.1253, simple_loss=0.2078, pruned_loss=0.02138, over 16820.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2432, pruned_loss=0.03548, over 3326057.94 frames. ], batch size: 39, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:13:39,139 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 12:14:02,914 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275385.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:14:26,877 INFO [train.py:904] (3/8) Epoch 28, batch 1350, loss[loss=0.1852, simple_loss=0.2621, pruned_loss=0.05419, over 16681.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2431, pruned_loss=0.03533, over 3319770.70 frames. ], batch size: 89, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:14:35,334 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275408.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 12:15:28,640 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 1400, loss[loss=0.1457, simple_loss=0.2363, pruned_loss=0.02752, over 16692.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2439, pruned_loss=0.03558, over 3326987.60 frames. ], batch size: 62, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:16:06,674 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 1450, loss[loss=0.1593, simple_loss=0.252, pruned_loss=0.03332, over 16622.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.2436, pruned_loss=0.03586, over 3323217.67 frames. ], batch size: 62, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:17:03,271 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:05,699 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6697, 1.9038, 2.2608, 2.5108, 2.5891, 2.5786, 2.0395, 2.7610], device='cuda:3'), covar=tensor([0.0220, 0.0584, 0.0388, 0.0375, 0.0391, 0.0435, 0.0591, 0.0243], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0200, 0.0189, 0.0194, 0.0209, 0.0168, 0.0205, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 12:17:53,998 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:57,211 INFO [train.py:904] (3/8) Epoch 28, batch 1500, loss[loss=0.1722, simple_loss=0.2656, pruned_loss=0.03947, over 16610.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2433, pruned_loss=0.03596, over 3324855.44 frames. ], batch size: 57, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:18:26,333 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.124e+02 2.433e+02 3.002e+02 9.704e+02, threshold=4.866e+02, percent-clipped=1.0 2023-05-02 12:19:01,050 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 1550, loss[loss=0.1518, simple_loss=0.241, pruned_loss=0.03126, over 16874.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2446, pruned_loss=0.03685, over 3329816.92 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:19:18,180 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 12:19:31,513 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1844, 3.8881, 4.4007, 2.2490, 4.6032, 4.7111, 3.4755, 3.5707], device='cuda:3'), covar=tensor([0.0735, 0.0312, 0.0265, 0.1223, 0.0097, 0.0191, 0.0467, 0.0441], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0111, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:20:18,406 INFO [train.py:904] (3/8) Epoch 28, batch 1600, loss[loss=0.2286, simple_loss=0.3033, pruned_loss=0.077, over 12257.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2456, pruned_loss=0.03719, over 3328043.08 frames. ], batch size: 248, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:20:33,167 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275663.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:20:48,273 INFO [optim.py:368] (3/8) 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 12:21:10,158 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1219, 3.3359, 3.4256, 2.3306, 3.1292, 3.5527, 3.2540, 2.1664], device='cuda:3'), covar=tensor([0.0576, 0.0161, 0.0075, 0.0447, 0.0145, 0.0111, 0.0119, 0.0502], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0103, 0.0115, 0.0099, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 12:21:28,984 INFO [train.py:904] (3/8) Epoch 28, batch 1650, loss[loss=0.172, simple_loss=0.2662, pruned_loss=0.03886, over 16700.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2475, pruned_loss=0.03805, over 3310158.41 frames. ], batch size: 57, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:21:29,287 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275703.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 12:22:21,191 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275741.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:22:37,918 INFO [train.py:904] (3/8) Epoch 28, batch 1700, loss[loss=0.1554, simple_loss=0.2363, pruned_loss=0.03726, over 16896.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.249, pruned_loss=0.03813, over 3309137.90 frames. ], batch size: 116, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:23:08,711 INFO [optim.py:368] (3/8) 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 12:23:15,120 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9856, 4.7338, 5.0344, 5.1993, 5.3815, 4.7553, 5.3927, 5.4000], device='cuda:3'), covar=tensor([0.1940, 0.1403, 0.1772, 0.0813, 0.0553, 0.0960, 0.0612, 0.0711], device='cuda:3'), in_proj_covar=tensor([0.0692, 0.0842, 0.0978, 0.0855, 0.0652, 0.0683, 0.0718, 0.0831], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:23:49,162 INFO [train.py:904] (3/8) Epoch 28, batch 1750, loss[loss=0.1723, simple_loss=0.2699, pruned_loss=0.03736, over 17072.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2507, pruned_loss=0.03829, over 3309991.91 frames. ], batch size: 55, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:24:04,500 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5135, 4.4423, 4.4338, 4.1142, 4.1762, 4.4597, 4.2213, 4.2106], device='cuda:3'), covar=tensor([0.0651, 0.0802, 0.0354, 0.0344, 0.0822, 0.0546, 0.0617, 0.0724], device='cuda:3'), in_proj_covar=tensor([0.0319, 0.0479, 0.0373, 0.0376, 0.0370, 0.0430, 0.0256, 0.0447], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 12:24:05,544 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 1800, loss[loss=0.1765, simple_loss=0.2677, pruned_loss=0.04266, over 16697.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2514, pruned_loss=0.03819, over 3317523.98 frames. ], batch size: 89, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:25:08,980 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275860.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:11,866 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275862.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:30,537 INFO [optim.py:368] (3/8) 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 12:26:08,320 INFO [train.py:904] (3/8) Epoch 28, batch 1850, loss[loss=0.1475, simple_loss=0.2445, pruned_loss=0.0253, over 17205.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2516, pruned_loss=0.03766, over 3324043.35 frames. ], batch size: 46, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:26:33,811 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275921.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:26:48,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9534, 3.6095, 4.1612, 2.1391, 4.3050, 4.3532, 3.3032, 3.3122], device='cuda:3'), covar=tensor([0.0715, 0.0319, 0.0235, 0.1213, 0.0090, 0.0233, 0.0440, 0.0449], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:27:05,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3642, 4.6718, 4.5111, 4.5266, 4.2688, 4.1933, 4.2243, 4.7346], device='cuda:3'), covar=tensor([0.1226, 0.0859, 0.0913, 0.0805, 0.0807, 0.1625, 0.1198, 0.0869], device='cuda:3'), in_proj_covar=tensor([0.0724, 0.0880, 0.0716, 0.0678, 0.0556, 0.0555, 0.0740, 0.0686], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:27:16,668 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275952.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:17,461 INFO [train.py:904] (3/8) Epoch 28, batch 1900, loss[loss=0.1595, simple_loss=0.2573, pruned_loss=0.03088, over 17272.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.251, pruned_loss=0.03698, over 3330235.58 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:27:22,873 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 12:27:32,186 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275963.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:39,831 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7605, 3.5744, 3.8990, 2.0111, 3.9789, 4.0094, 3.4000, 2.9756], device='cuda:3'), covar=tensor([0.0820, 0.0262, 0.0209, 0.1347, 0.0119, 0.0233, 0.0366, 0.0503], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0139, 0.0086, 0.0131, 0.0129, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:27:43,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8257, 4.6196, 4.8866, 5.0676, 5.2537, 4.6603, 5.2467, 5.2549], device='cuda:3'), covar=tensor([0.2207, 0.1413, 0.1865, 0.0830, 0.0578, 0.1109, 0.0657, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0693, 0.0843, 0.0981, 0.0858, 0.0652, 0.0685, 0.0719, 0.0832], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:27:47,686 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 1950, loss[loss=0.174, simple_loss=0.2582, pruned_loss=0.04484, over 16659.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2511, pruned_loss=0.0369, over 3319981.24 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:28:30,822 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:43,437 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:24,688 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:37,514 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 2000, loss[loss=0.1689, simple_loss=0.2477, pruned_loss=0.04502, over 16746.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2507, pruned_loss=0.03681, over 3319526.83 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:11,361 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.104e+02 2.511e+02 2.919e+02 7.567e+02, threshold=5.023e+02, percent-clipped=3.0 2023-05-02 12:30:31,277 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 2050, loss[loss=0.187, simple_loss=0.2637, pruned_loss=0.05518, over 16823.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2512, pruned_loss=0.03738, over 3316122.71 frames. ], batch size: 116, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:54,857 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 12:31:09,065 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-02 12:31:21,926 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9687, 5.2983, 5.1206, 5.1169, 4.8952, 4.7732, 4.7796, 5.4314], device='cuda:3'), covar=tensor([0.1351, 0.0941, 0.0995, 0.0877, 0.0820, 0.1119, 0.1228, 0.0886], device='cuda:3'), in_proj_covar=tensor([0.0727, 0.0889, 0.0723, 0.0683, 0.0560, 0.0557, 0.0745, 0.0691], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:31:26,614 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2923, 2.4161, 2.4032, 4.0214, 2.3174, 2.7540, 2.4596, 2.5215], device='cuda:3'), covar=tensor([0.1504, 0.3529, 0.3235, 0.0666, 0.4266, 0.2512, 0.3650, 0.3776], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0475, 0.0388, 0.0338, 0.0447, 0.0545, 0.0447, 0.0556], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:31:31,840 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9368, 4.4243, 4.4484, 3.2429, 3.6618, 4.3784, 3.9805, 2.7775], device='cuda:3'), covar=tensor([0.0491, 0.0086, 0.0052, 0.0378, 0.0166, 0.0111, 0.0101, 0.0478], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0092, 0.0092, 0.0137, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 12:32:00,357 INFO [train.py:904] (3/8) Epoch 28, batch 2100, loss[loss=0.1538, simple_loss=0.2528, pruned_loss=0.02743, over 17053.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2526, pruned_loss=0.03831, over 3316132.62 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:32:30,698 INFO [optim.py:368] (3/8) 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 12:33:09,363 INFO [train.py:904] (3/8) Epoch 28, batch 2150, loss[loss=0.1468, simple_loss=0.2279, pruned_loss=0.03285, over 16735.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2534, pruned_loss=0.03849, over 3318321.55 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:33:09,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8947, 4.1102, 3.1124, 2.3717, 2.6627, 2.6050, 4.2735, 3.4501], device='cuda:3'), covar=tensor([0.2635, 0.0489, 0.1719, 0.3129, 0.2889, 0.2167, 0.0447, 0.1597], device='cuda:3'), in_proj_covar=tensor([0.0336, 0.0277, 0.0313, 0.0328, 0.0306, 0.0279, 0.0306, 0.0355], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 12:33:28,003 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276216.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:34:18,058 INFO [train.py:904] (3/8) Epoch 28, batch 2200, loss[loss=0.1752, simple_loss=0.2505, pruned_loss=0.04994, over 16930.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2541, pruned_loss=0.03888, over 3318451.38 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:34:38,415 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 12:34:50,518 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 2250, loss[loss=0.1613, simple_loss=0.2591, pruned_loss=0.03175, over 17140.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2558, pruned_loss=0.03992, over 3309054.03 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:35:35,332 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:48,025 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:36:37,083 INFO [train.py:904] (3/8) Epoch 28, batch 2300, loss[loss=0.1585, simple_loss=0.2437, pruned_loss=0.03668, over 15453.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2563, pruned_loss=0.03977, over 3298554.70 frames. ], batch size: 190, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:36:38,737 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5439, 3.7026, 3.7732, 2.6796, 3.5685, 3.9184, 3.6754, 2.0212], device='cuda:3'), covar=tensor([0.0631, 0.0294, 0.0105, 0.0501, 0.0147, 0.0158, 0.0147, 0.0739], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 12:37:01,193 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 12:37:08,704 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.043e+02 2.409e+02 2.878e+02 4.638e+02, threshold=4.818e+02, percent-clipped=0.0 2023-05-02 12:37:12,883 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 2350, loss[loss=0.1751, simple_loss=0.2667, pruned_loss=0.04176, over 17022.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2558, pruned_loss=0.03949, over 3301676.96 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:38:54,360 INFO [train.py:904] (3/8) Epoch 28, batch 2400, loss[loss=0.1673, simple_loss=0.2655, pruned_loss=0.03454, over 17118.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2561, pruned_loss=0.03996, over 3309920.73 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:39:26,376 INFO [optim.py:368] (3/8) 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 12:39:39,802 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0677, 5.0218, 4.9330, 4.4971, 4.6498, 4.9754, 4.8554, 4.6404], device='cuda:3'), covar=tensor([0.0563, 0.0603, 0.0308, 0.0351, 0.0961, 0.0471, 0.0406, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0322, 0.0482, 0.0375, 0.0378, 0.0371, 0.0432, 0.0257, 0.0450], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 12:39:46,069 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 12:40:04,310 INFO [train.py:904] (3/8) Epoch 28, batch 2450, loss[loss=0.1778, simple_loss=0.2724, pruned_loss=0.0416, over 15834.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2556, pruned_loss=0.03923, over 3314843.35 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:40:23,639 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276516.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:40:37,911 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1457, 5.4311, 5.2052, 5.2368, 4.9590, 4.8776, 4.8990, 5.5483], device='cuda:3'), covar=tensor([0.1223, 0.0933, 0.1096, 0.0900, 0.0833, 0.1037, 0.1245, 0.0908], device='cuda:3'), in_proj_covar=tensor([0.0729, 0.0889, 0.0724, 0.0685, 0.0560, 0.0557, 0.0746, 0.0691], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:41:11,020 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5317, 3.8402, 4.0278, 3.9800, 4.0227, 3.8316, 3.5834, 3.8546], device='cuda:3'), covar=tensor([0.0716, 0.0868, 0.0648, 0.0750, 0.0756, 0.0743, 0.1456, 0.0698], device='cuda:3'), in_proj_covar=tensor([0.0444, 0.0503, 0.0483, 0.0446, 0.0531, 0.0511, 0.0587, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 12:41:13,985 INFO [train.py:904] (3/8) Epoch 28, batch 2500, loss[loss=0.1748, simple_loss=0.2609, pruned_loss=0.0443, over 16509.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2558, pruned_loss=0.03969, over 3310926.56 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:41:24,532 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0529, 2.1636, 2.5636, 2.9283, 2.8502, 3.0947, 2.2355, 3.2567], device='cuda:3'), covar=tensor([0.0216, 0.0527, 0.0409, 0.0315, 0.0368, 0.0274, 0.0595, 0.0182], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0201, 0.0189, 0.0195, 0.0210, 0.0168, 0.0205, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 12:41:30,321 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.004e+02 2.344e+02 2.885e+02 4.375e+02, threshold=4.688e+02, percent-clipped=0.0 2023-05-02 12:42:24,131 INFO [train.py:904] (3/8) Epoch 28, batch 2550, loss[loss=0.162, simple_loss=0.2595, pruned_loss=0.03223, over 17060.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2553, pruned_loss=0.03912, over 3307177.43 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:42:31,780 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:36,038 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276611.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:52,497 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 12:43:33,173 INFO [train.py:904] (3/8) Epoch 28, batch 2600, loss[loss=0.1921, simple_loss=0.2752, pruned_loss=0.05453, over 16511.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2548, pruned_loss=0.03866, over 3316575.04 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:43:37,974 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276656.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:55,712 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276672.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:02,598 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276673.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:05,672 INFO [optim.py:368] (3/8) 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 12:44:29,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8479, 1.9730, 2.4811, 2.8431, 2.7620, 3.3358, 2.2388, 3.2785], device='cuda:3'), covar=tensor([0.0323, 0.0638, 0.0421, 0.0411, 0.0425, 0.0248, 0.0624, 0.0205], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0202, 0.0189, 0.0196, 0.0211, 0.0169, 0.0206, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 12:44:43,681 INFO [train.py:904] (3/8) Epoch 28, batch 2650, loss[loss=0.1827, simple_loss=0.2641, pruned_loss=0.05065, over 16450.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.03923, over 3313785.96 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:45:09,284 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 12:45:19,143 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:45:34,561 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8806, 4.6421, 4.9127, 5.0918, 5.2596, 4.6520, 5.3218, 5.2818], device='cuda:3'), covar=tensor([0.2169, 0.1501, 0.1893, 0.0887, 0.0637, 0.1077, 0.0676, 0.0712], device='cuda:3'), in_proj_covar=tensor([0.0700, 0.0851, 0.0987, 0.0866, 0.0659, 0.0692, 0.0726, 0.0838], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:45:36,424 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1692, 3.8460, 4.3229, 2.1478, 4.4742, 4.6107, 3.3208, 3.5830], device='cuda:3'), covar=tensor([0.0658, 0.0271, 0.0238, 0.1200, 0.0108, 0.0199, 0.0456, 0.0418], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:45:49,057 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9265, 2.1165, 2.2102, 3.5950, 2.1212, 2.3901, 2.1710, 2.2232], device='cuda:3'), covar=tensor([0.1804, 0.4078, 0.3520, 0.0805, 0.4329, 0.2915, 0.4347, 0.3487], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0478, 0.0391, 0.0341, 0.0449, 0.0549, 0.0451, 0.0559], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:45:50,738 INFO [train.py:904] (3/8) Epoch 28, batch 2700, loss[loss=0.1582, simple_loss=0.2533, pruned_loss=0.03157, over 16801.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.257, pruned_loss=0.03897, over 3308497.70 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:46:18,163 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7988, 4.8115, 5.1584, 5.1296, 5.2165, 4.8771, 4.8331, 4.6643], device='cuda:3'), covar=tensor([0.0364, 0.0642, 0.0431, 0.0501, 0.0536, 0.0467, 0.1032, 0.0547], device='cuda:3'), in_proj_covar=tensor([0.0446, 0.0504, 0.0485, 0.0448, 0.0535, 0.0513, 0.0589, 0.0409], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 12:46:23,524 INFO [optim.py:368] (3/8) 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 12:46:35,126 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 12:46:42,594 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:47:00,576 INFO [train.py:904] (3/8) Epoch 28, batch 2750, loss[loss=0.1801, simple_loss=0.263, pruned_loss=0.04864, over 16864.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2581, pruned_loss=0.039, over 3318808.02 frames. ], batch size: 96, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:47:03,239 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 12:47:41,483 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7050, 3.8437, 2.6540, 4.5302, 2.9673, 4.4273, 2.6745, 3.1593], device='cuda:3'), covar=tensor([0.0365, 0.0469, 0.1511, 0.0329, 0.0882, 0.0573, 0.1495, 0.0806], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0185, 0.0198, 0.0177, 0.0182, 0.0225, 0.0207, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:48:07,244 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 2800, loss[loss=0.1405, simple_loss=0.2322, pruned_loss=0.02442, over 16777.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2575, pruned_loss=0.03875, over 3318469.55 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:23,823 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4817, 3.1265, 3.4834, 1.9035, 3.5573, 3.6170, 3.0256, 2.7486], device='cuda:3'), covar=tensor([0.0787, 0.0285, 0.0192, 0.1205, 0.0139, 0.0217, 0.0406, 0.0491], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0111, 0.0102, 0.0140, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:48:24,929 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276863.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:41,263 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 2850, loss[loss=0.1548, simple_loss=0.2512, pruned_loss=0.02915, over 15985.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2562, pruned_loss=0.0391, over 3306949.98 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:49:22,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0456, 2.1883, 2.3051, 3.5717, 2.1867, 2.4497, 2.2419, 2.3082], device='cuda:3'), covar=tensor([0.1706, 0.3957, 0.3346, 0.0866, 0.4199, 0.2923, 0.4052, 0.3474], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0478, 0.0390, 0.0341, 0.0448, 0.0549, 0.0450, 0.0558], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:49:50,151 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276924.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:49:57,684 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:27,309 INFO [train.py:904] (3/8) Epoch 28, batch 2900, loss[loss=0.1375, simple_loss=0.2324, pruned_loss=0.02126, over 16858.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2544, pruned_loss=0.03888, over 3315242.80 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:50:46,672 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:55,045 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:58,047 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.124e+02 2.391e+02 2.943e+02 6.733e+02, threshold=4.783e+02, percent-clipped=5.0 2023-05-02 12:51:21,979 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276992.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 12:51:36,311 INFO [train.py:904] (3/8) Epoch 28, batch 2950, loss[loss=0.2014, simple_loss=0.2863, pruned_loss=0.05823, over 16743.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2545, pruned_loss=0.03946, over 3321182.68 frames. ], batch size: 62, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:51:48,532 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:05,340 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:23,158 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4661, 3.6031, 4.1088, 2.2904, 3.2611, 2.5073, 3.9034, 3.8357], device='cuda:3'), covar=tensor([0.0287, 0.1003, 0.0424, 0.2082, 0.0819, 0.1015, 0.0612, 0.1060], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0170, 0.0158, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 12:52:45,412 INFO [train.py:904] (3/8) Epoch 28, batch 3000, loss[loss=0.1667, simple_loss=0.2498, pruned_loss=0.0418, over 16473.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2545, pruned_loss=0.04006, over 3327449.43 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:52:45,413 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 12:52:54,788 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 12:53:21,120 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277072.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:53:25,661 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.148e+02 2.585e+02 3.132e+02 1.139e+03, threshold=5.170e+02, percent-clipped=2.0 2023-05-02 12:54:02,141 INFO [train.py:904] (3/8) Epoch 28, batch 3050, loss[loss=0.1921, simple_loss=0.2704, pruned_loss=0.05691, over 16729.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2538, pruned_loss=0.04009, over 3328296.79 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:55:01,337 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 3100, loss[loss=0.1456, simple_loss=0.2466, pruned_loss=0.02224, over 17251.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2532, pruned_loss=0.03965, over 3334623.79 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:55:43,287 INFO [optim.py:368] (3/8) 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 12:55:45,352 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:00,440 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 12:56:12,054 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6452, 3.3057, 3.7039, 2.0137, 3.8149, 3.8327, 3.1954, 2.8542], device='cuda:3'), covar=tensor([0.0757, 0.0288, 0.0204, 0.1134, 0.0111, 0.0223, 0.0391, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 12:56:21,091 INFO [train.py:904] (3/8) Epoch 28, batch 3150, loss[loss=0.1806, simple_loss=0.2611, pruned_loss=0.05003, over 16358.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2515, pruned_loss=0.03911, over 3335687.80 frames. ], batch size: 165, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:56:44,544 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:44,673 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:10,168 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4763, 4.4514, 4.3978, 3.8839, 4.4310, 1.7738, 4.1919, 4.0037], device='cuda:3'), covar=tensor([0.0143, 0.0128, 0.0200, 0.0320, 0.0117, 0.2960, 0.0160, 0.0255], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0179, 0.0217, 0.0190, 0.0193, 0.0222, 0.0206, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:57:10,228 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277238.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:30,727 INFO [train.py:904] (3/8) Epoch 28, batch 3200, loss[loss=0.1634, simple_loss=0.2605, pruned_loss=0.03319, over 17125.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2521, pruned_loss=0.0391, over 3334320.75 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:57:44,246 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7288, 1.9385, 2.3370, 2.6484, 2.6490, 2.7506, 2.0903, 2.8760], device='cuda:3'), covar=tensor([0.0239, 0.0577, 0.0411, 0.0349, 0.0397, 0.0356, 0.0595, 0.0250], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0201, 0.0190, 0.0196, 0.0211, 0.0169, 0.0206, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 12:57:50,723 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277267.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:01,815 INFO [optim.py:368] (3/8) 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 12:58:07,075 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277280.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:17,846 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277287.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:58:39,066 INFO [train.py:904] (3/8) Epoch 28, batch 3250, loss[loss=0.1635, simple_loss=0.2534, pruned_loss=0.03684, over 16837.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2526, pruned_loss=0.03938, over 3331396.14 frames. ], batch size: 83, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:58:42,470 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:09,077 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:21,740 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5299, 2.3782, 2.6289, 4.3028, 2.4268, 2.6496, 2.5429, 2.5386], device='cuda:3'), covar=tensor([0.1639, 0.4403, 0.3309, 0.0710, 0.4732, 0.3126, 0.3932, 0.4368], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0480, 0.0391, 0.0342, 0.0449, 0.0551, 0.0451, 0.0561], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 12:59:47,704 INFO [train.py:904] (3/8) Epoch 28, batch 3300, loss[loss=0.142, simple_loss=0.2247, pruned_loss=0.02959, over 16929.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2531, pruned_loss=0.03944, over 3337659.13 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:00:05,291 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:06,312 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277367.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:09,339 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5909, 3.9951, 4.0182, 2.7586, 3.6041, 4.0486, 3.6384, 2.3539], device='cuda:3'), covar=tensor([0.0530, 0.0172, 0.0070, 0.0411, 0.0141, 0.0113, 0.0118, 0.0526], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 13:00:12,568 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.125e+02 2.555e+02 3.019e+02 5.327e+02, threshold=5.111e+02, percent-clipped=1.0 2023-05-02 13:00:55,903 INFO [train.py:904] (3/8) Epoch 28, batch 3350, loss[loss=0.1556, simple_loss=0.2338, pruned_loss=0.03874, over 16788.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2535, pruned_loss=0.0396, over 3337702.94 frames. ], batch size: 96, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:01:55,030 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:01:59,068 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6177, 3.3949, 3.8380, 2.0630, 3.8996, 3.9472, 3.1671, 2.8751], device='cuda:3'), covar=tensor([0.0801, 0.0279, 0.0169, 0.1200, 0.0104, 0.0187, 0.0422, 0.0522], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0111, 0.0103, 0.0139, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:01:59,126 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9053, 2.6990, 2.5587, 4.2577, 3.3397, 4.0973, 1.7517, 3.0046], device='cuda:3'), covar=tensor([0.1387, 0.0751, 0.1244, 0.0166, 0.0154, 0.0378, 0.1579, 0.0854], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0206, 0.0208, 0.0221, 0.0210, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:02:03,950 INFO [train.py:904] (3/8) Epoch 28, batch 3400, loss[loss=0.1521, simple_loss=0.2429, pruned_loss=0.03065, over 17098.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2529, pruned_loss=0.03915, over 3340419.46 frames. ], batch size: 47, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:02:34,599 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:13,659 INFO [train.py:904] (3/8) Epoch 28, batch 3450, loss[loss=0.1751, simple_loss=0.2525, pruned_loss=0.04885, over 16749.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.252, pruned_loss=0.0386, over 3342817.17 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:03:35,994 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277519.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:55,440 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:23,509 INFO [train.py:904] (3/8) Epoch 28, batch 3500, loss[loss=0.1353, simple_loss=0.2308, pruned_loss=0.01993, over 17174.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2505, pruned_loss=0.03815, over 3327374.56 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:04:25,840 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:28,245 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5594, 3.6523, 2.3211, 3.8838, 2.8465, 3.8053, 2.3320, 2.9536], device='cuda:3'), covar=tensor([0.0288, 0.0397, 0.1542, 0.0355, 0.0799, 0.0852, 0.1473, 0.0753], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0186, 0.0199, 0.0179, 0.0183, 0.0227, 0.0208, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:04:42,186 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:47,061 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:49,054 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9507, 2.5509, 2.4537, 3.8671, 3.0611, 3.9224, 1.6626, 2.7769], device='cuda:3'), covar=tensor([0.1428, 0.0874, 0.1367, 0.0218, 0.0158, 0.0513, 0.1736, 0.0969], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0206, 0.0208, 0.0221, 0.0210, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:04:54,493 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:55,301 INFO [optim.py:368] (3/8) 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 13:05:10,619 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277587.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 13:05:20,840 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2727, 5.1580, 4.9185, 3.8004, 5.1080, 1.7262, 4.6632, 4.6949], device='cuda:3'), covar=tensor([0.0162, 0.0146, 0.0303, 0.0800, 0.0146, 0.3759, 0.0244, 0.0407], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0181, 0.0219, 0.0192, 0.0195, 0.0225, 0.0209, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:05:32,630 INFO [train.py:904] (3/8) Epoch 28, batch 3550, loss[loss=0.1748, simple_loss=0.2557, pruned_loss=0.04694, over 16227.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2496, pruned_loss=0.03772, over 3308092.04 frames. ], batch size: 165, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:05:35,392 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2150, 4.5055, 4.4951, 3.2585, 3.7438, 4.4991, 4.0559, 2.8960], device='cuda:3'), covar=tensor([0.0451, 0.0069, 0.0056, 0.0367, 0.0165, 0.0109, 0.0099, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0104, 0.0116, 0.0100, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 13:05:49,417 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277615.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:11,011 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277631.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:16,856 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:42,325 INFO [train.py:904] (3/8) Epoch 28, batch 3600, loss[loss=0.1374, simple_loss=0.223, pruned_loss=0.02588, over 17037.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2481, pruned_loss=0.03767, over 3303161.50 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:06:52,860 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:01,954 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277667.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:10,839 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 13:07:14,463 INFO [optim.py:368] (3/8) 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 13:07:36,956 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 3650, loss[loss=0.1559, simple_loss=0.2421, pruned_loss=0.03481, over 11602.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2479, pruned_loss=0.03803, over 3295768.95 frames. ], batch size: 247, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:08:10,538 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-05-02 13:08:12,450 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:09:06,597 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277752.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 13:09:07,229 INFO [train.py:904] (3/8) Epoch 28, batch 3700, loss[loss=0.1712, simple_loss=0.2538, pruned_loss=0.04432, over 16745.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.247, pruned_loss=0.03909, over 3273150.80 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:09:35,429 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5832, 1.8470, 2.3025, 2.4256, 2.5383, 2.5508, 1.9040, 2.7077], device='cuda:3'), covar=tensor([0.0227, 0.0548, 0.0363, 0.0365, 0.0372, 0.0373, 0.0635, 0.0228], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0201, 0.0189, 0.0196, 0.0211, 0.0169, 0.0206, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 13:09:41,446 INFO [optim.py:368] (3/8) 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 13:10:22,496 INFO [train.py:904] (3/8) Epoch 28, batch 3750, loss[loss=0.1834, simple_loss=0.2748, pruned_loss=0.04598, over 16662.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2479, pruned_loss=0.04022, over 3251909.36 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:10:48,732 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0870, 2.1267, 2.3236, 3.7976, 2.1878, 2.4537, 2.2704, 2.2936], device='cuda:3'), covar=tensor([0.1745, 0.4109, 0.3193, 0.0730, 0.4135, 0.2747, 0.4116, 0.3258], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0482, 0.0392, 0.0344, 0.0451, 0.0553, 0.0453, 0.0563], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:11:06,993 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:11:35,739 INFO [train.py:904] (3/8) Epoch 28, batch 3800, loss[loss=0.1754, simple_loss=0.2662, pruned_loss=0.04226, over 17124.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2492, pruned_loss=0.04171, over 3250486.29 frames. ], batch size: 49, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:11:41,262 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7570, 4.1169, 2.9793, 2.4757, 2.7948, 2.7330, 4.3812, 3.6825], device='cuda:3'), covar=tensor([0.3014, 0.0560, 0.1871, 0.2738, 0.2420, 0.1934, 0.0422, 0.1112], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0279, 0.0315, 0.0330, 0.0309, 0.0280, 0.0307, 0.0358], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 13:11:47,292 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-02 13:12:08,692 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 13:12:17,666 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:48,151 INFO [train.py:904] (3/8) Epoch 28, batch 3850, loss[loss=0.1747, simple_loss=0.2524, pruned_loss=0.04852, over 16430.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2491, pruned_loss=0.0423, over 3259627.52 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:59,660 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:18,278 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:23,015 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 3900, loss[loss=0.1614, simple_loss=0.2387, pruned_loss=0.04206, over 16756.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2494, pruned_loss=0.04329, over 3265879.52 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:14:13,213 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277961.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:27,560 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:36,142 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.147e+02 2.438e+02 2.765e+02 5.958e+02, threshold=4.876e+02, percent-clipped=2.0 2023-05-02 13:15:16,801 INFO [train.py:904] (3/8) Epoch 28, batch 3950, loss[loss=0.1716, simple_loss=0.2514, pruned_loss=0.04585, over 16203.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2487, pruned_loss=0.04365, over 3276594.28 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:15:25,690 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278009.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:15:58,291 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278032.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:16:11,279 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6399, 5.6402, 5.4853, 5.1140, 5.1560, 5.5863, 5.4231, 5.2937], device='cuda:3'), covar=tensor([0.0537, 0.0410, 0.0260, 0.0281, 0.0903, 0.0360, 0.0234, 0.0542], device='cuda:3'), in_proj_covar=tensor([0.0325, 0.0487, 0.0379, 0.0381, 0.0377, 0.0438, 0.0259, 0.0454], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 13:16:19,768 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278047.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 13:16:26,865 INFO [train.py:904] (3/8) Epoch 28, batch 4000, loss[loss=0.1545, simple_loss=0.2369, pruned_loss=0.03607, over 16447.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2488, pruned_loss=0.04365, over 3286965.44 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:01,537 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 1.968e+02 2.308e+02 2.954e+02 7.046e+02, threshold=4.616e+02, percent-clipped=2.0 2023-05-02 13:17:38,528 INFO [train.py:904] (3/8) Epoch 28, batch 4050, loss[loss=0.1664, simple_loss=0.2497, pruned_loss=0.04152, over 16847.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2499, pruned_loss=0.04316, over 3288756.11 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:44,507 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9317, 2.2242, 2.1713, 3.4966, 2.1306, 2.4821, 2.2944, 2.3038], device='cuda:3'), covar=tensor([0.1605, 0.3453, 0.3138, 0.0654, 0.4040, 0.2549, 0.3696, 0.3264], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0481, 0.0390, 0.0342, 0.0449, 0.0552, 0.0452, 0.0562], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:18:12,349 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0866, 3.1333, 3.4440, 2.0202, 2.9303, 2.2650, 3.5733, 3.4467], device='cuda:3'), covar=tensor([0.0223, 0.0844, 0.0645, 0.2173, 0.0885, 0.0977, 0.0537, 0.0948], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0174, 0.0171, 0.0158, 0.0150, 0.0133, 0.0148, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 13:18:52,733 INFO [train.py:904] (3/8) Epoch 28, batch 4100, loss[loss=0.1977, simple_loss=0.2786, pruned_loss=0.05838, over 12071.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2517, pruned_loss=0.04274, over 3278499.47 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:18:56,552 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 13:19:29,061 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 4150, loss[loss=0.2022, simple_loss=0.2953, pruned_loss=0.0545, over 16202.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2582, pruned_loss=0.04502, over 3213889.53 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:20:19,256 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:20:42,372 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:14,399 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6837, 4.0234, 2.8116, 2.3387, 2.9933, 2.6586, 4.3468, 3.5799], device='cuda:3'), covar=tensor([0.3308, 0.0725, 0.2159, 0.3050, 0.2813, 0.2141, 0.0575, 0.1339], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0279, 0.0315, 0.0330, 0.0310, 0.0280, 0.0308, 0.0358], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 13:21:22,143 INFO [train.py:904] (3/8) Epoch 28, batch 4200, loss[loss=0.2035, simple_loss=0.2978, pruned_loss=0.05459, over 15402.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.265, pruned_loss=0.04721, over 3164635.35 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:21:30,670 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:54,358 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 13:22:00,399 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 13:22:37,391 INFO [train.py:904] (3/8) Epoch 28, batch 4250, loss[loss=0.1653, simple_loss=0.2567, pruned_loss=0.03693, over 16702.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.268, pruned_loss=0.04643, over 3164139.66 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:22:45,284 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 13:23:07,780 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:13,508 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:44,077 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278347.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 13:23:52,555 INFO [train.py:904] (3/8) Epoch 28, batch 4300, loss[loss=0.1812, simple_loss=0.28, pruned_loss=0.04123, over 16749.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2697, pruned_loss=0.04582, over 3167310.35 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:24:25,589 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278374.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:29,321 INFO [optim.py:368] (3/8) 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 13:24:40,805 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:51,259 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1012, 3.9999, 4.2108, 4.3080, 4.3907, 4.0016, 4.3361, 4.4432], device='cuda:3'), covar=tensor([0.1674, 0.1186, 0.1218, 0.0625, 0.0517, 0.1431, 0.0840, 0.0642], device='cuda:3'), in_proj_covar=tensor([0.0689, 0.0842, 0.0975, 0.0856, 0.0651, 0.0682, 0.0714, 0.0825], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:24:56,294 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:25:07,418 INFO [train.py:904] (3/8) Epoch 28, batch 4350, loss[loss=0.1994, simple_loss=0.2913, pruned_loss=0.05378, over 16859.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2728, pruned_loss=0.04654, over 3172616.58 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:25:55,311 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9807, 2.9439, 2.4983, 2.7674, 3.2964, 2.9068, 3.4697, 3.4418], device='cuda:3'), covar=tensor([0.0087, 0.0389, 0.0550, 0.0464, 0.0253, 0.0379, 0.0225, 0.0270], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0247, 0.0234, 0.0235, 0.0248, 0.0245, 0.0246, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:25:57,141 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 4400, loss[loss=0.1793, simple_loss=0.2788, pruned_loss=0.03989, over 16826.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2757, pruned_loss=0.04818, over 3164486.58 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:26:23,514 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8186, 3.6306, 4.1472, 2.1217, 4.4304, 4.4365, 3.1193, 3.3626], device='cuda:3'), covar=tensor([0.0813, 0.0325, 0.0231, 0.1260, 0.0067, 0.0125, 0.0487, 0.0458], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:26:58,177 INFO [optim.py:368] (3/8) 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 13:27:21,187 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 13:27:35,966 INFO [train.py:904] (3/8) Epoch 28, batch 4450, loss[loss=0.1944, simple_loss=0.2886, pruned_loss=0.05006, over 16472.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2798, pruned_loss=0.0498, over 3182640.16 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:27:47,281 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:28:49,073 INFO [train.py:904] (3/8) Epoch 28, batch 4500, loss[loss=0.1924, simple_loss=0.2807, pruned_loss=0.05203, over 16983.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2803, pruned_loss=0.05062, over 3199730.68 frames. ], batch size: 41, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:29:06,850 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3894, 3.3439, 2.1338, 3.8014, 2.5172, 3.8036, 2.2693, 2.7124], device='cuda:3'), covar=tensor([0.0301, 0.0392, 0.1653, 0.0147, 0.0897, 0.0463, 0.1540, 0.0878], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0183, 0.0197, 0.0175, 0.0182, 0.0224, 0.0205, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:29:17,480 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:29:23,510 INFO [optim.py:368] (3/8) 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 13:29:23,837 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2141, 5.4518, 5.2450, 5.2949, 4.9987, 4.8742, 4.9663, 5.5850], device='cuda:3'), covar=tensor([0.1134, 0.0886, 0.1009, 0.0890, 0.0784, 0.0896, 0.1061, 0.0836], device='cuda:3'), in_proj_covar=tensor([0.0722, 0.0883, 0.0717, 0.0677, 0.0555, 0.0551, 0.0735, 0.0686], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:30:01,298 INFO [train.py:904] (3/8) Epoch 28, batch 4550, loss[loss=0.207, simple_loss=0.2921, pruned_loss=0.06093, over 16362.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2807, pruned_loss=0.05121, over 3208108.94 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:30:36,963 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278627.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:12,169 INFO [train.py:904] (3/8) Epoch 28, batch 4600, loss[loss=0.2103, simple_loss=0.3016, pruned_loss=0.05953, over 16891.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2813, pruned_loss=0.05123, over 3214561.81 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:31:34,990 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 13:31:43,899 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:46,041 INFO [optim.py:368] (3/8) 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 13:31:49,266 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:32:14,633 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 13:32:22,022 INFO [train.py:904] (3/8) Epoch 28, batch 4650, loss[loss=0.2077, simple_loss=0.2807, pruned_loss=0.06737, over 11821.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2805, pruned_loss=0.05143, over 3223803.23 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:32:42,877 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2292, 3.4677, 3.5641, 2.1251, 3.0195, 2.4000, 3.4916, 3.7751], device='cuda:3'), covar=tensor([0.0261, 0.0841, 0.0642, 0.2270, 0.0931, 0.1046, 0.0704, 0.1061], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 13:33:00,842 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:33:22,569 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1430, 3.2142, 3.5365, 1.9885, 3.0113, 2.1998, 3.4528, 3.5750], device='cuda:3'), covar=tensor([0.0275, 0.0902, 0.0670, 0.2384, 0.0916, 0.1125, 0.0659, 0.1074], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 13:33:33,716 INFO [train.py:904] (3/8) Epoch 28, batch 4700, loss[loss=0.1645, simple_loss=0.2556, pruned_loss=0.03676, over 16508.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2775, pruned_loss=0.05006, over 3224537.22 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:33:55,652 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1534, 2.2177, 2.4542, 3.8846, 2.1506, 2.4748, 2.3728, 2.3944], device='cuda:3'), covar=tensor([0.1914, 0.4153, 0.3187, 0.0714, 0.5092, 0.3097, 0.3595, 0.4081], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0475, 0.0385, 0.0338, 0.0446, 0.0545, 0.0446, 0.0555], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:34:07,033 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 4750, loss[loss=0.159, simple_loss=0.2506, pruned_loss=0.03368, over 16809.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2723, pruned_loss=0.04764, over 3234863.33 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:35:32,924 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0403, 4.0293, 3.9716, 3.1067, 3.9156, 1.6941, 3.7415, 3.4957], device='cuda:3'), covar=tensor([0.0155, 0.0194, 0.0181, 0.0424, 0.0120, 0.3103, 0.0167, 0.0314], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0176, 0.0214, 0.0187, 0.0190, 0.0219, 0.0203, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:35:57,544 INFO [train.py:904] (3/8) Epoch 28, batch 4800, loss[loss=0.1756, simple_loss=0.2647, pruned_loss=0.04323, over 16322.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2693, pruned_loss=0.04573, over 3229106.70 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:36:18,858 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:36:27,847 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4628, 3.3915, 3.8596, 1.8277, 3.9568, 4.0124, 3.0607, 2.9639], device='cuda:3'), covar=tensor([0.0922, 0.0315, 0.0181, 0.1407, 0.0090, 0.0173, 0.0434, 0.0548], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:36:31,722 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 13:36:32,582 INFO [optim.py:368] (3/8) 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 13:37:09,560 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8793, 2.1483, 2.1995, 3.4439, 2.0682, 2.4151, 2.2492, 2.2756], device='cuda:3'), covar=tensor([0.1737, 0.3929, 0.3178, 0.0699, 0.4273, 0.2739, 0.3917, 0.3366], device='cuda:3'), in_proj_covar=tensor([0.0422, 0.0475, 0.0384, 0.0337, 0.0445, 0.0544, 0.0446, 0.0554], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:37:13,016 INFO [train.py:904] (3/8) Epoch 28, batch 4850, loss[loss=0.1827, simple_loss=0.2799, pruned_loss=0.04269, over 16733.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2693, pruned_loss=0.04453, over 3227506.10 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:38:00,087 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2001, 4.3181, 4.1152, 3.8474, 3.8055, 4.2297, 3.9250, 3.9755], device='cuda:3'), covar=tensor([0.0607, 0.0437, 0.0303, 0.0304, 0.0762, 0.0441, 0.0944, 0.0587], device='cuda:3'), in_proj_covar=tensor([0.0313, 0.0470, 0.0367, 0.0369, 0.0363, 0.0421, 0.0250, 0.0437], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:38:26,577 INFO [train.py:904] (3/8) Epoch 28, batch 4900, loss[loss=0.1979, simple_loss=0.2967, pruned_loss=0.0495, over 16915.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2686, pruned_loss=0.04334, over 3220928.27 frames. ], batch size: 109, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:00,724 INFO [optim.py:368] (3/8) 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 13:39:04,647 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:39:39,808 INFO [train.py:904] (3/8) Epoch 28, batch 4950, loss[loss=0.1669, simple_loss=0.2589, pruned_loss=0.03743, over 16876.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2687, pruned_loss=0.04294, over 3220702.50 frames. ], batch size: 42, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:45,069 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1671, 2.4459, 1.9358, 2.2510, 2.7968, 2.4199, 2.6449, 2.9457], device='cuda:3'), covar=tensor([0.0165, 0.0517, 0.0738, 0.0550, 0.0321, 0.0461, 0.0282, 0.0320], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0247, 0.0235, 0.0235, 0.0247, 0.0245, 0.0245, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:40:14,413 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:19,454 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:51,497 INFO [train.py:904] (3/8) Epoch 28, batch 5000, loss[loss=0.1574, simple_loss=0.2547, pruned_loss=0.03, over 16590.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2712, pruned_loss=0.04354, over 3222628.00 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:41:26,569 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279078.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:42:04,191 INFO [train.py:904] (3/8) Epoch 28, batch 5050, loss[loss=0.182, simple_loss=0.272, pruned_loss=0.046, over 16822.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2712, pruned_loss=0.04322, over 3232675.42 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:42:14,332 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0537, 2.2398, 2.2535, 3.6098, 2.1325, 2.5319, 2.2978, 2.3745], device='cuda:3'), covar=tensor([0.1575, 0.3740, 0.3206, 0.0650, 0.4323, 0.2770, 0.3945, 0.3299], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0478, 0.0386, 0.0339, 0.0446, 0.0546, 0.0448, 0.0556], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:42:25,620 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-05-02 13:42:29,501 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2913, 5.5693, 5.3067, 5.3996, 5.0879, 5.0279, 4.9679, 5.6325], device='cuda:3'), covar=tensor([0.1283, 0.0800, 0.1016, 0.0780, 0.0799, 0.0691, 0.1187, 0.0834], device='cuda:3'), in_proj_covar=tensor([0.0715, 0.0873, 0.0710, 0.0669, 0.0550, 0.0546, 0.0727, 0.0678], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:42:45,643 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:42:47,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7545, 2.4002, 2.5050, 4.4128, 3.0211, 3.9359, 1.5606, 2.9595], device='cuda:3'), covar=tensor([0.1511, 0.1036, 0.1401, 0.0177, 0.0249, 0.0395, 0.1925, 0.0907], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0203, 0.0207, 0.0217, 0.0209, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:43:17,183 INFO [train.py:904] (3/8) Epoch 28, batch 5100, loss[loss=0.1684, simple_loss=0.2599, pruned_loss=0.03845, over 15532.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2697, pruned_loss=0.04279, over 3224908.49 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:43:20,267 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 13:43:29,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4730, 2.6332, 2.2588, 2.4444, 2.9629, 2.5894, 2.9902, 3.1929], device='cuda:3'), covar=tensor([0.0121, 0.0479, 0.0617, 0.0487, 0.0287, 0.0436, 0.0216, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0246, 0.0235, 0.0235, 0.0247, 0.0245, 0.0244, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:43:37,525 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:52,643 INFO [optim.py:368] (3/8) 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 13:44:15,307 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:44:30,680 INFO [train.py:904] (3/8) Epoch 28, batch 5150, loss[loss=0.1755, simple_loss=0.2805, pruned_loss=0.03527, over 16295.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2695, pruned_loss=0.04179, over 3225483.43 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:44:48,998 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:45:42,233 INFO [train.py:904] (3/8) Epoch 28, batch 5200, loss[loss=0.1593, simple_loss=0.2502, pruned_loss=0.03422, over 16752.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2687, pruned_loss=0.04191, over 3217164.78 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:45:53,232 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 13:46:00,551 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5512, 1.8228, 2.1668, 2.5538, 2.5577, 2.8508, 1.9733, 2.7748], device='cuda:3'), covar=tensor([0.0255, 0.0609, 0.0387, 0.0360, 0.0367, 0.0229, 0.0616, 0.0165], device='cuda:3'), in_proj_covar=tensor([0.0201, 0.0200, 0.0188, 0.0194, 0.0210, 0.0168, 0.0205, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:46:17,346 INFO [optim.py:368] (3/8) 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 13:46:28,868 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0019, 2.4488, 2.6645, 1.8982, 2.7718, 2.8473, 2.5016, 2.3816], device='cuda:3'), covar=tensor([0.0756, 0.0288, 0.0206, 0.0959, 0.0134, 0.0219, 0.0439, 0.0448], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0110, 0.0102, 0.0138, 0.0086, 0.0130, 0.0129, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:46:53,552 INFO [train.py:904] (3/8) Epoch 28, batch 5250, loss[loss=0.1816, simple_loss=0.2587, pruned_loss=0.05222, over 12250.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2659, pruned_loss=0.04145, over 3220755.60 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:46:57,398 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6151, 3.7037, 3.4910, 3.2501, 3.3053, 3.6066, 3.3502, 3.4406], device='cuda:3'), covar=tensor([0.0606, 0.0692, 0.0323, 0.0339, 0.0659, 0.0590, 0.1361, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0470, 0.0365, 0.0368, 0.0364, 0.0422, 0.0249, 0.0435], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:47:22,284 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6846, 4.7266, 4.9740, 4.9512, 4.9758, 4.7022, 4.6538, 4.5274], device='cuda:3'), covar=tensor([0.0282, 0.0527, 0.0328, 0.0348, 0.0428, 0.0361, 0.0886, 0.0465], device='cuda:3'), in_proj_covar=tensor([0.0434, 0.0489, 0.0473, 0.0436, 0.0517, 0.0497, 0.0574, 0.0400], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 13:48:06,861 INFO [train.py:904] (3/8) Epoch 28, batch 5300, loss[loss=0.1554, simple_loss=0.2486, pruned_loss=0.03116, over 16429.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2626, pruned_loss=0.0403, over 3221047.81 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:41,217 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 1.963e+02 2.333e+02 2.759e+02 6.747e+02, threshold=4.665e+02, percent-clipped=2.0 2023-05-02 13:49:02,930 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0299, 4.1265, 3.9358, 3.6694, 3.6849, 4.0472, 3.6859, 3.8495], device='cuda:3'), covar=tensor([0.0606, 0.0551, 0.0304, 0.0313, 0.0750, 0.0498, 0.1133, 0.0538], device='cuda:3'), in_proj_covar=tensor([0.0312, 0.0470, 0.0365, 0.0369, 0.0365, 0.0423, 0.0249, 0.0435], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 13:49:20,979 INFO [train.py:904] (3/8) Epoch 28, batch 5350, loss[loss=0.1679, simple_loss=0.2565, pruned_loss=0.03961, over 16845.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2613, pruned_loss=0.03996, over 3218943.02 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:49:55,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8075, 2.3115, 2.2763, 3.4264, 2.0326, 3.5389, 1.5722, 2.7439], device='cuda:3'), covar=tensor([0.1334, 0.0867, 0.1310, 0.0172, 0.0128, 0.0352, 0.1701, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0203, 0.0207, 0.0218, 0.0210, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:50:32,504 INFO [train.py:904] (3/8) Epoch 28, batch 5400, loss[loss=0.1804, simple_loss=0.2722, pruned_loss=0.04429, over 16689.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2637, pruned_loss=0.04079, over 3231532.74 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:51:08,322 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 1.928e+02 2.246e+02 2.506e+02 7.644e+02, threshold=4.493e+02, percent-clipped=1.0 2023-05-02 13:51:24,273 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:51:50,344 INFO [train.py:904] (3/8) Epoch 28, batch 5450, loss[loss=0.1912, simple_loss=0.2841, pruned_loss=0.04914, over 16810.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2663, pruned_loss=0.04198, over 3223863.17 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:52:31,275 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279529.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:52:52,409 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 5500, loss[loss=0.2253, simple_loss=0.2971, pruned_loss=0.07671, over 11919.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2725, pruned_loss=0.04527, over 3197841.12 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:53:10,589 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7914, 3.0596, 3.2618, 1.9711, 2.8300, 2.1398, 3.3273, 3.4079], device='cuda:3'), covar=tensor([0.0285, 0.0915, 0.0656, 0.2339, 0.0947, 0.1109, 0.0656, 0.0949], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 13:53:47,383 INFO [optim.py:368] (3/8) 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 13:54:00,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3350, 3.5096, 3.8685, 2.1929, 3.2968, 2.3350, 3.8334, 3.8742], device='cuda:3'), covar=tensor([0.0217, 0.0782, 0.0522, 0.2165, 0.0744, 0.1040, 0.0491, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 13:54:09,345 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:28,765 INFO [train.py:904] (3/8) Epoch 28, batch 5550, loss[loss=0.1961, simple_loss=0.2826, pruned_loss=0.05476, over 16330.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2795, pruned_loss=0.05048, over 3149283.28 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:54:29,333 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:32,735 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 13:55:02,372 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.55 vs. limit=5.0 2023-05-02 13:55:47,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4644, 3.2048, 3.6211, 1.9002, 3.7449, 3.7665, 2.9153, 2.8519], device='cuda:3'), covar=tensor([0.0790, 0.0327, 0.0232, 0.1203, 0.0093, 0.0188, 0.0471, 0.0484], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 13:55:48,648 INFO [train.py:904] (3/8) Epoch 28, batch 5600, loss[loss=0.2139, simple_loss=0.2985, pruned_loss=0.06461, over 16845.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2849, pruned_loss=0.05524, over 3090326.34 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:56:28,890 INFO [optim.py:368] (3/8) 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 13:57:08,613 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-02 13:57:11,930 INFO [train.py:904] (3/8) Epoch 28, batch 5650, loss[loss=0.2533, simple_loss=0.3238, pruned_loss=0.0914, over 15235.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2904, pruned_loss=0.05952, over 3048743.81 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:58:29,304 INFO [train.py:904] (3/8) Epoch 28, batch 5700, loss[loss=0.203, simple_loss=0.2916, pruned_loss=0.05717, over 16650.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2912, pruned_loss=0.06039, over 3055662.00 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:59:05,449 INFO [optim.py:368] (3/8) 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 13:59:22,605 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 5750, loss[loss=0.1768, simple_loss=0.2739, pruned_loss=0.03989, over 16911.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2933, pruned_loss=0.06139, over 3039662.44 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:00:39,834 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:01:07,307 INFO [train.py:904] (3/8) Epoch 28, batch 5800, loss[loss=0.1659, simple_loss=0.269, pruned_loss=0.03144, over 16872.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2938, pruned_loss=0.06078, over 3028779.68 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:01:46,120 INFO [optim.py:368] (3/8) 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 14:01:57,429 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 28, batch 5850, loss[loss=0.2366, simple_loss=0.3037, pruned_loss=0.08478, over 11416.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2918, pruned_loss=0.05941, over 3018558.66 frames. ], batch size: 247, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:03:44,211 INFO [train.py:904] (3/8) Epoch 28, batch 5900, loss[loss=0.2433, simple_loss=0.3109, pruned_loss=0.08782, over 11454.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.291, pruned_loss=0.05902, over 3035244.81 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:04:26,173 INFO [optim.py:368] (3/8) 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 14:05:07,644 INFO [train.py:904] (3/8) Epoch 28, batch 5950, loss[loss=0.2076, simple_loss=0.2973, pruned_loss=0.05893, over 16267.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2913, pruned_loss=0.05732, over 3049590.54 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:20,830 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 6000, loss[loss=0.2126, simple_loss=0.2963, pruned_loss=0.06449, over 11657.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2904, pruned_loss=0.05696, over 3054341.58 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:23,831 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 14:06:34,277 INFO [train.py:938] (3/8) Epoch 28, validation: loss=0.148, simple_loss=0.2602, pruned_loss=0.0179, over 944034.00 frames. 2023-05-02 14:06:34,277 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 14:07:11,168 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.966e+02 2.746e+02 3.426e+02 4.098e+02 7.990e+02, threshold=6.852e+02, percent-clipped=5.0 2023-05-02 14:07:25,952 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 6050, loss[loss=0.2218, simple_loss=0.3125, pruned_loss=0.06552, over 16641.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2887, pruned_loss=0.05585, over 3079994.98 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:08:06,535 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:08:24,210 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5383, 3.5343, 3.5228, 2.6898, 3.3801, 2.0596, 3.1996, 2.8219], device='cuda:3'), covar=tensor([0.0201, 0.0165, 0.0217, 0.0266, 0.0136, 0.2607, 0.0179, 0.0316], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0187, 0.0190, 0.0218, 0.0202, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:09:02,476 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280149.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:08,878 INFO [train.py:904] (3/8) Epoch 28, batch 6100, loss[loss=0.185, simple_loss=0.2713, pruned_loss=0.04933, over 16522.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.288, pruned_loss=0.0549, over 3095000.97 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:09:10,113 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-05-02 14:09:51,400 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.432e+02 2.900e+02 3.731e+02 6.785e+02, threshold=5.800e+02, percent-clipped=0.0 2023-05-02 14:10:02,006 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280185.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:13,195 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8796, 5.1753, 5.3842, 5.1533, 5.2755, 5.7571, 5.2559, 4.9976], device='cuda:3'), covar=tensor([0.1066, 0.1831, 0.2552, 0.1896, 0.2137, 0.0908, 0.1602, 0.2563], device='cuda:3'), in_proj_covar=tensor([0.0426, 0.0637, 0.0699, 0.0515, 0.0691, 0.0723, 0.0543, 0.0691], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 14:10:22,448 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 6150, loss[loss=0.179, simple_loss=0.2633, pruned_loss=0.04733, over 17078.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2861, pruned_loss=0.054, over 3111723.50 frames. ], batch size: 55, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:11:16,582 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280233.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:35,648 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:45,495 INFO [train.py:904] (3/8) Epoch 28, batch 6200, loss[loss=0.1937, simple_loss=0.2819, pruned_loss=0.05275, over 16234.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2845, pruned_loss=0.05393, over 3112680.95 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:12:08,812 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 14:12:24,023 INFO [optim.py:368] (3/8) 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 14:12:39,516 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4759, 1.6970, 2.1384, 2.4562, 2.4818, 2.7336, 1.8406, 2.7203], device='cuda:3'), covar=tensor([0.0257, 0.0626, 0.0373, 0.0390, 0.0368, 0.0243, 0.0626, 0.0187], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0193, 0.0209, 0.0166, 0.0203, 0.0166], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:13:00,715 INFO [train.py:904] (3/8) Epoch 28, batch 6250, loss[loss=0.1783, simple_loss=0.2733, pruned_loss=0.04163, over 17045.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2839, pruned_loss=0.05384, over 3090953.80 frames. ], batch size: 53, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:13:06,066 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0316, 5.0376, 4.9139, 4.1038, 4.9299, 1.8392, 4.6527, 4.5132], device='cuda:3'), covar=tensor([0.0159, 0.0161, 0.0259, 0.0451, 0.0184, 0.2957, 0.0209, 0.0291], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0175, 0.0214, 0.0187, 0.0190, 0.0218, 0.0202, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:13:10,301 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 14:13:36,649 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:13:44,951 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3250, 2.5085, 2.5507, 4.0810, 2.3441, 2.8670, 2.5228, 2.6818], device='cuda:3'), covar=tensor([0.1481, 0.3534, 0.2918, 0.0561, 0.4017, 0.2418, 0.3805, 0.3064], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0476, 0.0386, 0.0337, 0.0446, 0.0545, 0.0447, 0.0555], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:14:05,565 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 14:14:14,430 INFO [train.py:904] (3/8) Epoch 28, batch 6300, loss[loss=0.17, simple_loss=0.258, pruned_loss=0.04103, over 17074.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2832, pruned_loss=0.05284, over 3104416.70 frames. ], batch size: 49, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:14:53,563 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.897e+02 2.662e+02 3.040e+02 3.867e+02 7.586e+02, threshold=6.080e+02, percent-clipped=1.0 2023-05-02 14:15:09,872 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:15:31,097 INFO [train.py:904] (3/8) Epoch 28, batch 6350, loss[loss=0.192, simple_loss=0.2759, pruned_loss=0.05403, over 16625.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2844, pruned_loss=0.05419, over 3085402.39 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:15:31,564 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8056, 4.7839, 5.1299, 5.1027, 5.1442, 4.8143, 4.7728, 4.6397], device='cuda:3'), covar=tensor([0.0324, 0.0627, 0.0377, 0.0402, 0.0474, 0.0401, 0.1008, 0.0513], device='cuda:3'), in_proj_covar=tensor([0.0434, 0.0490, 0.0474, 0.0438, 0.0520, 0.0499, 0.0575, 0.0402], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 14:15:36,843 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:31,383 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:43,550 INFO [train.py:904] (3/8) Epoch 28, batch 6400, loss[loss=0.2517, simple_loss=0.3196, pruned_loss=0.09195, over 11202.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2854, pruned_loss=0.05586, over 3078831.14 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:17:19,317 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.958e+02 3.396e+02 3.952e+02 7.468e+02, threshold=6.793e+02, percent-clipped=6.0 2023-05-02 14:17:56,100 INFO [train.py:904] (3/8) Epoch 28, batch 6450, loss[loss=0.2133, simple_loss=0.2853, pruned_loss=0.07067, over 11562.00 frames. ], tot_loss[loss=0.199, simple_loss=0.286, pruned_loss=0.056, over 3052290.46 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:18:18,428 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:18:41,527 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:02,762 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:13,553 INFO [train.py:904] (3/8) Epoch 28, batch 6500, loss[loss=0.1891, simple_loss=0.2788, pruned_loss=0.04972, over 16620.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2836, pruned_loss=0.05504, over 3062139.02 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:19:49,976 INFO [optim.py:368] (3/8) 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 14:19:52,147 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:15,882 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:28,523 INFO [train.py:904] (3/8) Epoch 28, batch 6550, loss[loss=0.2152, simple_loss=0.3082, pruned_loss=0.06113, over 16538.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2861, pruned_loss=0.05595, over 3074110.77 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:20:37,406 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:42,900 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2257, 4.3089, 4.1422, 3.8760, 3.8839, 4.2354, 3.8864, 4.0051], device='cuda:3'), covar=tensor([0.0643, 0.0792, 0.0315, 0.0303, 0.0760, 0.0537, 0.0976, 0.0652], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0468, 0.0362, 0.0365, 0.0360, 0.0418, 0.0247, 0.0432], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:20:45,259 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:21:34,440 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1679, 2.1274, 2.7142, 3.1320, 2.9659, 3.6746, 2.1408, 3.6391], device='cuda:3'), covar=tensor([0.0275, 0.0589, 0.0373, 0.0346, 0.0338, 0.0172, 0.0711, 0.0158], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0186, 0.0191, 0.0207, 0.0165, 0.0202, 0.0165], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:21:44,262 INFO [train.py:904] (3/8) Epoch 28, batch 6600, loss[loss=0.2398, simple_loss=0.3081, pruned_loss=0.08575, over 11812.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2881, pruned_loss=0.05615, over 3058808.10 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:22:18,049 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:22:28,859 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:59,886 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 6650, loss[loss=0.2632, simple_loss=0.3206, pruned_loss=0.1029, over 11327.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2884, pruned_loss=0.05677, over 3065098.72 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:23:07,195 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:20,731 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3894, 3.3708, 3.4094, 3.4812, 3.5185, 3.2722, 3.4948, 3.5617], device='cuda:3'), covar=tensor([0.1210, 0.0854, 0.0978, 0.0639, 0.0707, 0.2495, 0.1205, 0.0870], device='cuda:3'), in_proj_covar=tensor([0.0667, 0.0814, 0.0944, 0.0830, 0.0635, 0.0663, 0.0696, 0.0801], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:24:02,743 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 6700, loss[loss=0.1982, simple_loss=0.2882, pruned_loss=0.05408, over 16611.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2874, pruned_loss=0.05763, over 3049803.65 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:24:18,762 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:25,627 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-05-02 14:24:30,598 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:25:06,376 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8493, 2.7087, 2.8909, 2.1709, 2.6952, 2.1705, 2.7236, 2.9439], device='cuda:3'), covar=tensor([0.0277, 0.0829, 0.0517, 0.1813, 0.0820, 0.0891, 0.0582, 0.0737], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0158, 0.0149, 0.0134, 0.0148, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 14:25:13,162 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:25:29,255 INFO [train.py:904] (3/8) Epoch 28, batch 6750, loss[loss=0.185, simple_loss=0.276, pruned_loss=0.04706, over 16304.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2868, pruned_loss=0.05751, over 3049647.92 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:25:47,132 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2033, 3.2769, 2.0224, 3.6021, 2.5109, 3.6185, 2.1512, 2.6755], device='cuda:3'), covar=tensor([0.0346, 0.0432, 0.1788, 0.0241, 0.0928, 0.0594, 0.1724, 0.0857], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:26:41,172 INFO [train.py:904] (3/8) Epoch 28, batch 6800, loss[loss=0.1834, simple_loss=0.2776, pruned_loss=0.04458, over 16728.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2866, pruned_loss=0.05714, over 3069611.51 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:13,224 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:18,678 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 2.733e+02 3.177e+02 3.942e+02 7.279e+02, threshold=6.355e+02, percent-clipped=1.0 2023-05-02 14:27:36,559 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280889.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:40,460 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8673, 2.7560, 2.9190, 2.2216, 2.7215, 2.2143, 2.7944, 2.9756], device='cuda:3'), covar=tensor([0.0255, 0.0747, 0.0486, 0.1670, 0.0761, 0.0880, 0.0555, 0.0678], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0173, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 14:27:55,353 INFO [train.py:904] (3/8) Epoch 28, batch 6850, loss[loss=0.1817, simple_loss=0.2908, pruned_loss=0.0363, over 16775.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2878, pruned_loss=0.05688, over 3093418.92 frames. ], batch size: 89, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:56,252 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:28:04,546 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8073, 4.8176, 5.1742, 5.1375, 5.1546, 4.8528, 4.7804, 4.7203], device='cuda:3'), covar=tensor([0.0358, 0.0535, 0.0476, 0.0445, 0.0501, 0.0435, 0.1020, 0.0454], device='cuda:3'), in_proj_covar=tensor([0.0434, 0.0488, 0.0474, 0.0438, 0.0520, 0.0500, 0.0575, 0.0400], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 14:28:54,805 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4234, 5.7811, 5.4884, 5.5934, 5.2599, 5.2094, 5.1352, 5.8757], device='cuda:3'), covar=tensor([0.1309, 0.0903, 0.1114, 0.0853, 0.0829, 0.0715, 0.1198, 0.0873], device='cuda:3'), in_proj_covar=tensor([0.0717, 0.0871, 0.0713, 0.0671, 0.0549, 0.0547, 0.0725, 0.0677], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:28:55,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0849, 5.3946, 5.1522, 5.1866, 4.9184, 4.8461, 4.7651, 5.4697], device='cuda:3'), covar=tensor([0.1253, 0.0912, 0.1028, 0.0877, 0.0796, 0.0886, 0.1217, 0.0910], device='cuda:3'), in_proj_covar=tensor([0.0717, 0.0871, 0.0713, 0.0671, 0.0549, 0.0547, 0.0725, 0.0676], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:29:00,167 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3600, 3.7938, 3.8280, 2.4642, 3.5216, 3.8725, 3.5441, 2.1170], device='cuda:3'), covar=tensor([0.0586, 0.0082, 0.0071, 0.0470, 0.0122, 0.0134, 0.0116, 0.0549], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0136, 0.0102, 0.0116, 0.0098, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 14:29:06,760 INFO [train.py:904] (3/8) Epoch 28, batch 6900, loss[loss=0.1775, simple_loss=0.2775, pruned_loss=0.03869, over 16882.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2893, pruned_loss=0.05625, over 3097841.01 frames. ], batch size: 96, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:29:34,423 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:29:47,267 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3217, 4.3971, 4.2511, 3.9617, 3.9482, 4.3373, 4.0648, 4.1075], device='cuda:3'), covar=tensor([0.0709, 0.0718, 0.0323, 0.0338, 0.0866, 0.0567, 0.0724, 0.0676], device='cuda:3'), in_proj_covar=tensor([0.0307, 0.0464, 0.0358, 0.0362, 0.0357, 0.0414, 0.0246, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:29:53,915 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 6950, loss[loss=0.1748, simple_loss=0.2682, pruned_loss=0.04065, over 16849.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2908, pruned_loss=0.0577, over 3078083.44 frames. ], batch size: 102, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:30:33,737 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 14:31:03,952 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:21,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7155, 3.8907, 2.4906, 4.3392, 2.9911, 4.2621, 2.4761, 3.0345], device='cuda:3'), covar=tensor([0.0325, 0.0364, 0.1687, 0.0282, 0.0833, 0.0585, 0.1831, 0.0900], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0173, 0.0180, 0.0220, 0.0204, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:31:30,269 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281049.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:31:35,801 INFO [train.py:904] (3/8) Epoch 28, batch 7000, loss[loss=0.2236, simple_loss=0.2977, pruned_loss=0.0748, over 11872.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2907, pruned_loss=0.05714, over 3077369.43 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:44,111 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281058.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:32:12,984 INFO [optim.py:368] (3/8) 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 14:32:50,413 INFO [train.py:904] (3/8) Epoch 28, batch 7050, loss[loss=0.2223, simple_loss=0.301, pruned_loss=0.07185, over 11409.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2917, pruned_loss=0.05686, over 3086128.48 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:33:01,152 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281110.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:33:38,605 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2058, 3.2714, 2.0731, 3.6119, 2.4777, 3.6218, 2.2728, 2.6972], device='cuda:3'), covar=tensor([0.0357, 0.0434, 0.1684, 0.0213, 0.0899, 0.0602, 0.1499, 0.0807], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:33:43,613 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 14:33:48,683 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 14:34:04,177 INFO [train.py:904] (3/8) Epoch 28, batch 7100, loss[loss=0.1867, simple_loss=0.2807, pruned_loss=0.0464, over 17251.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2907, pruned_loss=0.05743, over 3060679.52 frames. ], batch size: 52, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:34:09,536 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 14:34:29,298 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281169.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:37,869 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 14:34:59,570 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 7150, loss[loss=0.2057, simple_loss=0.2917, pruned_loss=0.05983, over 16605.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2886, pruned_loss=0.05684, over 3073612.98 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:35:21,006 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281203.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:35:47,783 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:00,713 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281230.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:36:09,137 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:29,667 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281251.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:31,490 INFO [train.py:904] (3/8) Epoch 28, batch 7200, loss[loss=0.1927, simple_loss=0.2834, pruned_loss=0.051, over 16665.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2858, pruned_loss=0.05467, over 3087780.63 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:36:33,048 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-05-02 14:36:38,225 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3334, 4.4152, 4.2343, 3.9283, 3.9528, 4.3402, 4.0091, 4.0746], device='cuda:3'), covar=tensor([0.0610, 0.0582, 0.0294, 0.0303, 0.0775, 0.0551, 0.0803, 0.0631], device='cuda:3'), in_proj_covar=tensor([0.0308, 0.0466, 0.0360, 0.0363, 0.0359, 0.0415, 0.0247, 0.0430], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:36:43,169 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-05-02 14:36:56,582 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281270.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:37:07,756 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 7250, loss[loss=0.1847, simple_loss=0.2719, pruned_loss=0.04877, over 16437.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2842, pruned_loss=0.05389, over 3081721.00 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:38:09,206 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:38:42,295 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 14:38:59,568 INFO [train.py:904] (3/8) Epoch 28, batch 7300, loss[loss=0.2083, simple_loss=0.2971, pruned_loss=0.05975, over 16189.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2836, pruned_loss=0.05378, over 3083662.67 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:39:08,078 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281358.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:39:39,520 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.942e+02 3.043e+02 3.727e+02 4.632e+02 1.394e+03, threshold=7.454e+02, percent-clipped=7.0 2023-05-02 14:40:13,649 INFO [train.py:904] (3/8) Epoch 28, batch 7350, loss[loss=0.2321, simple_loss=0.3068, pruned_loss=0.07865, over 11331.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2848, pruned_loss=0.05484, over 3080533.09 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:40:16,307 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281405.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:40:17,486 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:40:50,070 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:27,313 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9544, 5.0148, 5.4143, 5.3755, 5.3989, 5.0617, 5.0036, 4.8535], device='cuda:3'), covar=tensor([0.0363, 0.0482, 0.0377, 0.0392, 0.0463, 0.0421, 0.0954, 0.0498], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0483, 0.0468, 0.0433, 0.0514, 0.0494, 0.0568, 0.0396], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 14:41:28,055 INFO [train.py:904] (3/8) Epoch 28, batch 7400, loss[loss=0.2136, simple_loss=0.2997, pruned_loss=0.06375, over 17028.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2853, pruned_loss=0.05538, over 3079827.78 frames. ], batch size: 53, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:41:33,764 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:41:57,974 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 14:42:05,091 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3809, 4.3729, 4.2546, 3.4005, 4.3098, 1.7004, 4.0968, 3.8183], device='cuda:3'), covar=tensor([0.0121, 0.0121, 0.0212, 0.0355, 0.0094, 0.3175, 0.0147, 0.0324], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0175, 0.0213, 0.0186, 0.0189, 0.0218, 0.0201, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:42:08,141 INFO [optim.py:368] (3/8) 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 14:42:13,929 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9525, 3.1299, 2.7408, 5.2630, 4.0779, 4.3079, 1.9511, 3.0064], device='cuda:3'), covar=tensor([0.1261, 0.0769, 0.1274, 0.0158, 0.0347, 0.0493, 0.1535, 0.0939], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0181, 0.0202, 0.0204, 0.0208, 0.0219, 0.0210, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:42:18,928 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:42:28,750 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4242, 2.9802, 2.7283, 2.3264, 2.3281, 2.3081, 2.9753, 2.9074], device='cuda:3'), covar=tensor([0.2396, 0.0653, 0.1560, 0.2551, 0.2280, 0.2224, 0.0544, 0.1272], device='cuda:3'), in_proj_covar=tensor([0.0335, 0.0273, 0.0312, 0.0326, 0.0304, 0.0276, 0.0304, 0.0350], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 14:42:41,851 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 14:42:42,677 INFO [train.py:904] (3/8) Epoch 28, batch 7450, loss[loss=0.2116, simple_loss=0.301, pruned_loss=0.06105, over 16402.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2865, pruned_loss=0.05683, over 3058914.50 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:43:05,948 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:43:11,511 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7698, 5.0641, 4.8497, 4.8593, 4.6351, 4.5783, 4.4962, 5.1652], device='cuda:3'), covar=tensor([0.1246, 0.0902, 0.0937, 0.0918, 0.0759, 0.1093, 0.1210, 0.0825], device='cuda:3'), in_proj_covar=tensor([0.0712, 0.0867, 0.0709, 0.0667, 0.0546, 0.0546, 0.0721, 0.0674], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:43:17,713 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281525.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:43:58,008 INFO [train.py:904] (3/8) Epoch 28, batch 7500, loss[loss=0.1827, simple_loss=0.273, pruned_loss=0.0462, over 16641.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2864, pruned_loss=0.05577, over 3073039.47 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:44:36,767 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.813e+02 3.506e+02 4.192e+02 1.275e+03, threshold=7.011e+02, percent-clipped=6.0 2023-05-02 14:45:11,672 INFO [train.py:904] (3/8) Epoch 28, batch 7550, loss[loss=0.1923, simple_loss=0.2806, pruned_loss=0.05206, over 16781.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2847, pruned_loss=0.05552, over 3082180.16 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:45:15,097 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5441, 3.6662, 2.2566, 4.1816, 2.7343, 4.1094, 2.5211, 2.9532], device='cuda:3'), covar=tensor([0.0307, 0.0384, 0.1693, 0.0205, 0.0855, 0.0551, 0.1467, 0.0803], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0171, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:45:52,762 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-05-02 14:46:25,961 INFO [train.py:904] (3/8) Epoch 28, batch 7600, loss[loss=0.192, simple_loss=0.292, pruned_loss=0.04603, over 16827.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2844, pruned_loss=0.05536, over 3097159.47 frames. ], batch size: 102, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:46:29,146 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 14:47:04,823 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 7650, loss[loss=0.2221, simple_loss=0.3082, pruned_loss=0.06799, over 16749.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2846, pruned_loss=0.05545, over 3101492.44 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:43,575 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281705.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:48:51,547 INFO [train.py:904] (3/8) Epoch 28, batch 7700, loss[loss=0.1855, simple_loss=0.2774, pruned_loss=0.04676, over 15334.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2847, pruned_loss=0.05609, over 3085465.03 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:48:52,552 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281753.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:49:01,621 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6028, 2.5547, 1.9099, 2.6599, 2.1573, 2.7614, 2.1976, 2.3969], device='cuda:3'), covar=tensor([0.0354, 0.0409, 0.1309, 0.0314, 0.0684, 0.0545, 0.1171, 0.0626], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0172, 0.0180, 0.0220, 0.0205, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:49:27,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8582, 4.9213, 4.7554, 4.4003, 4.4233, 4.8227, 4.6307, 4.5349], device='cuda:3'), covar=tensor([0.0675, 0.0698, 0.0334, 0.0359, 0.1000, 0.0621, 0.0509, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0463, 0.0357, 0.0360, 0.0356, 0.0413, 0.0247, 0.0428], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:49:31,077 INFO [optim.py:368] (3/8) 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 14:49:35,140 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:07,528 INFO [train.py:904] (3/8) Epoch 28, batch 7750, loss[loss=0.19, simple_loss=0.2772, pruned_loss=0.05137, over 16746.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.285, pruned_loss=0.05628, over 3084354.88 frames. ], batch size: 76, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:50:21,372 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:40,399 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281825.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:21,154 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281852.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:21,936 INFO [train.py:904] (3/8) Epoch 28, batch 7800, loss[loss=0.1842, simple_loss=0.2732, pruned_loss=0.04762, over 16988.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2857, pruned_loss=0.05684, over 3080739.32 frames. ], batch size: 41, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:51:52,697 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.715e+02 3.377e+02 4.149e+02 7.003e+02, threshold=6.754e+02, percent-clipped=1.0 2023-05-02 14:52:37,333 INFO [train.py:904] (3/8) Epoch 28, batch 7850, loss[loss=0.2341, simple_loss=0.3119, pruned_loss=0.07817, over 15470.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2869, pruned_loss=0.05657, over 3076288.61 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:52:37,849 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4674, 3.5169, 2.2111, 3.9725, 2.6874, 3.9261, 2.3477, 2.8380], device='cuda:3'), covar=tensor([0.0304, 0.0385, 0.1737, 0.0247, 0.0873, 0.0671, 0.1608, 0.0837], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0172, 0.0180, 0.0220, 0.0205, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:52:47,394 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281910.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:52:51,801 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281913.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:37,329 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 7900, loss[loss=0.2377, simple_loss=0.3097, pruned_loss=0.0828, over 11442.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2852, pruned_loss=0.05577, over 3095418.53 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:54:16,287 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281971.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:54:29,909 INFO [optim.py:368] (3/8) 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 14:55:00,264 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9478, 5.0038, 4.8520, 4.4732, 4.5174, 4.8937, 4.6904, 4.5763], device='cuda:3'), covar=tensor([0.0594, 0.0491, 0.0281, 0.0307, 0.0946, 0.0449, 0.0436, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0463, 0.0358, 0.0360, 0.0357, 0.0414, 0.0248, 0.0429], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:55:09,799 INFO [train.py:904] (3/8) Epoch 28, batch 7950, loss[loss=0.1993, simple_loss=0.285, pruned_loss=0.05677, over 16352.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.286, pruned_loss=0.05621, over 3108931.48 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:55:14,733 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:56:27,144 INFO [train.py:904] (3/8) Epoch 28, batch 8000, loss[loss=0.1939, simple_loss=0.2832, pruned_loss=0.05234, over 16365.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.287, pruned_loss=0.05761, over 3079929.30 frames. ], batch size: 35, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:07,769 INFO [optim.py:368] (3/8) 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 14:57:11,187 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:22,573 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2731, 3.3921, 2.0235, 3.7391, 2.5729, 3.7566, 2.2625, 2.7403], device='cuda:3'), covar=tensor([0.0376, 0.0431, 0.1847, 0.0285, 0.0900, 0.0664, 0.1624, 0.0875], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0173, 0.0180, 0.0221, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 14:57:33,365 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282096.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:42,452 INFO [train.py:904] (3/8) Epoch 28, batch 8050, loss[loss=0.1909, simple_loss=0.2859, pruned_loss=0.048, over 17004.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2869, pruned_loss=0.05709, over 3086847.01 frames. ], batch size: 41, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:56,160 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:19,932 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0445, 4.0327, 3.9458, 3.0698, 3.9853, 1.8512, 3.7968, 3.4077], device='cuda:3'), covar=tensor([0.0148, 0.0125, 0.0214, 0.0306, 0.0108, 0.3014, 0.0147, 0.0341], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0173, 0.0213, 0.0185, 0.0187, 0.0216, 0.0200, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 14:58:21,350 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 14:58:22,175 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:57,914 INFO [train.py:904] (3/8) Epoch 28, batch 8100, loss[loss=0.1921, simple_loss=0.2791, pruned_loss=0.05257, over 16236.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2868, pruned_loss=0.05685, over 3078325.07 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:59:03,732 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282157.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:08,426 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 28, batch 8150, loss[loss=0.1653, simple_loss=0.2627, pruned_loss=0.03393, over 16831.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2842, pruned_loss=0.05552, over 3082086.00 frames. ], batch size: 102, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:00:21,757 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282208.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:00:28,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7107, 3.9123, 4.2172, 2.1007, 4.5048, 4.5734, 3.3658, 3.1736], device='cuda:3'), covar=tensor([0.1095, 0.0224, 0.0235, 0.1346, 0.0084, 0.0149, 0.0430, 0.0591], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:00:48,266 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6065, 1.8464, 2.2929, 2.6030, 2.5834, 3.0379, 2.0142, 2.9735], device='cuda:3'), covar=tensor([0.0272, 0.0610, 0.0392, 0.0419, 0.0391, 0.0229, 0.0648, 0.0189], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0191, 0.0207, 0.0164, 0.0202, 0.0164], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:00:58,470 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 15:01:29,257 INFO [train.py:904] (3/8) Epoch 28, batch 8200, loss[loss=0.1808, simple_loss=0.2702, pruned_loss=0.04573, over 16573.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2816, pruned_loss=0.05473, over 3091719.94 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:01:50,081 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282266.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:02:13,182 INFO [optim.py:368] (3/8) 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 15:02:15,796 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 15:02:46,940 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:02:50,327 INFO [train.py:904] (3/8) Epoch 28, batch 8250, loss[loss=0.177, simple_loss=0.2666, pruned_loss=0.0437, over 17217.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2804, pruned_loss=0.05239, over 3080494.44 frames. ], batch size: 44, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:03:32,134 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 15:03:37,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8369, 3.7347, 3.9142, 4.0074, 4.0856, 3.6957, 4.0487, 4.1198], device='cuda:3'), covar=tensor([0.1708, 0.1252, 0.1325, 0.0750, 0.0611, 0.1872, 0.0807, 0.0850], device='cuda:3'), in_proj_covar=tensor([0.0662, 0.0813, 0.0938, 0.0824, 0.0629, 0.0657, 0.0692, 0.0799], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:04:07,365 INFO [train.py:904] (3/8) Epoch 28, batch 8300, loss[loss=0.1748, simple_loss=0.2583, pruned_loss=0.04569, over 11839.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2778, pruned_loss=0.04955, over 3054326.20 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:04:50,784 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 28, batch 8350, loss[loss=0.19, simple_loss=0.2895, pruned_loss=0.04529, over 15317.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2775, pruned_loss=0.04771, over 3063947.85 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:05:51,003 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-02 15:06:17,905 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6103, 2.7039, 2.4093, 2.5445, 3.0201, 2.7029, 3.0428, 3.2623], device='cuda:3'), covar=tensor([0.0165, 0.0444, 0.0573, 0.0465, 0.0339, 0.0463, 0.0335, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0222, 0.0236, 0.0228, 0.0228, 0.0238, 0.0237, 0.0234, 0.0237], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:06:43,068 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282452.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:06:43,833 INFO [train.py:904] (3/8) Epoch 28, batch 8400, loss[loss=0.162, simple_loss=0.2481, pruned_loss=0.03794, over 12177.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2748, pruned_loss=0.046, over 3041935.05 frames. ], batch size: 250, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:07:26,987 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.221e+02 2.637e+02 3.332e+02 6.864e+02, threshold=5.273e+02, percent-clipped=5.0 2023-05-02 15:07:39,987 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:08:02,091 INFO [train.py:904] (3/8) Epoch 28, batch 8450, loss[loss=0.1622, simple_loss=0.2519, pruned_loss=0.03623, over 12364.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2725, pruned_loss=0.04409, over 3024504.02 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:08:11,394 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:17,236 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:21,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0912, 2.3589, 2.3929, 3.0467, 1.9399, 3.2269, 1.8845, 2.8539], device='cuda:3'), covar=tensor([0.1136, 0.0646, 0.0999, 0.0181, 0.0080, 0.0323, 0.1443, 0.0602], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0198, 0.0199, 0.0205, 0.0215, 0.0207, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:09:21,852 INFO [train.py:904] (3/8) Epoch 28, batch 8500, loss[loss=0.1631, simple_loss=0.2442, pruned_loss=0.04096, over 11884.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2693, pruned_loss=0.04185, over 3046266.08 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:09:27,448 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:43,362 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282566.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:09:49,654 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8140, 4.0819, 4.1125, 2.9313, 3.5633, 4.1062, 3.7540, 2.4471], device='cuda:3'), covar=tensor([0.0448, 0.0062, 0.0053, 0.0360, 0.0138, 0.0117, 0.0083, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0132, 0.0100, 0.0113, 0.0096, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 15:09:55,036 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1201, 2.1835, 2.1672, 3.6485, 2.0856, 2.5174, 2.2725, 2.2943], device='cuda:3'), covar=tensor([0.1432, 0.4052, 0.3446, 0.0634, 0.4817, 0.2750, 0.4049, 0.3757], device='cuda:3'), in_proj_covar=tensor([0.0412, 0.0465, 0.0379, 0.0328, 0.0438, 0.0532, 0.0439, 0.0544], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:10:07,074 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 2.128e+02 2.662e+02 3.202e+02 5.740e+02, threshold=5.324e+02, percent-clipped=2.0 2023-05-02 15:10:42,293 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 8550, loss[loss=0.1734, simple_loss=0.2748, pruned_loss=0.03596, over 15285.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2674, pruned_loss=0.0409, over 3035145.74 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:11:05,487 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282614.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:11:21,223 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4872, 5.5620, 5.3966, 4.9663, 5.0269, 5.4556, 5.3057, 5.0955], device='cuda:3'), covar=tensor([0.0665, 0.0593, 0.0313, 0.0356, 0.1095, 0.0551, 0.0296, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0306, 0.0464, 0.0358, 0.0359, 0.0354, 0.0414, 0.0248, 0.0427], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:11:47,043 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9007, 2.1546, 2.3575, 3.1230, 2.1658, 2.3461, 2.3133, 2.2469], device='cuda:3'), covar=tensor([0.1477, 0.3934, 0.3039, 0.0797, 0.4887, 0.2874, 0.3730, 0.4120], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0467, 0.0380, 0.0329, 0.0439, 0.0533, 0.0440, 0.0546], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:12:14,466 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:12:21,560 INFO [train.py:904] (3/8) Epoch 28, batch 8600, loss[loss=0.1636, simple_loss=0.2604, pruned_loss=0.03341, over 16467.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.267, pruned_loss=0.03996, over 3021874.13 frames. ], batch size: 68, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:12:30,310 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9587, 2.7108, 3.0051, 2.1874, 2.7388, 2.1274, 2.8010, 2.9392], device='cuda:3'), covar=tensor([0.0258, 0.0960, 0.0456, 0.1890, 0.0775, 0.0976, 0.0620, 0.0865], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0167, 0.0168, 0.0155, 0.0145, 0.0130, 0.0143, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:12:32,751 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282658.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:12:38,332 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 15:12:45,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8099, 3.0275, 2.7757, 5.0861, 3.7270, 4.4378, 1.6528, 3.3578], device='cuda:3'), covar=tensor([0.1337, 0.0741, 0.1119, 0.0112, 0.0157, 0.0308, 0.1684, 0.0629], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0178, 0.0198, 0.0199, 0.0204, 0.0215, 0.0208, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:13:07,618 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2985, 3.4329, 2.2472, 3.7466, 2.6304, 3.7180, 2.3329, 2.7748], device='cuda:3'), covar=tensor([0.0406, 0.0412, 0.1646, 0.0264, 0.0828, 0.0681, 0.1563, 0.0817], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0178, 0.0194, 0.0169, 0.0177, 0.0217, 0.0202, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:13:12,566 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282677.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:17,329 INFO [optim.py:368] (3/8) 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 15:13:35,728 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0029, 4.2651, 4.1266, 4.1298, 3.8557, 3.8881, 3.8792, 4.2836], device='cuda:3'), covar=tensor([0.1149, 0.0987, 0.1020, 0.0919, 0.0882, 0.1936, 0.1087, 0.0974], device='cuda:3'), in_proj_covar=tensor([0.0705, 0.0853, 0.0699, 0.0660, 0.0538, 0.0542, 0.0711, 0.0664], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:13:40,140 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2554, 4.2313, 4.1251, 3.0057, 4.1502, 1.6507, 3.8661, 3.8224], device='cuda:3'), covar=tensor([0.0179, 0.0166, 0.0259, 0.0614, 0.0162, 0.3633, 0.0225, 0.0418], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0170, 0.0208, 0.0181, 0.0184, 0.0213, 0.0196, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:13:50,768 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 8650, loss[loss=0.1765, simple_loss=0.2763, pruned_loss=0.03837, over 16652.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2652, pruned_loss=0.0386, over 3025277.20 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:14:28,342 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4013, 2.7782, 3.2073, 2.0245, 2.8020, 2.1122, 3.0759, 3.0360], device='cuda:3'), covar=tensor([0.0254, 0.0984, 0.0560, 0.2230, 0.0836, 0.1067, 0.0653, 0.0969], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0154, 0.0145, 0.0130, 0.0143, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:14:36,260 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282719.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:14,352 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,266 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,929 INFO [train.py:904] (3/8) Epoch 28, batch 8700, loss[loss=0.1588, simple_loss=0.258, pruned_loss=0.02978, over 16873.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2626, pruned_loss=0.03766, over 3034163.96 frames. ], batch size: 102, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:15:53,235 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282759.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:16:15,001 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8503, 3.8416, 3.9922, 3.7749, 3.9356, 4.3539, 3.9864, 3.7091], device='cuda:3'), covar=tensor([0.2121, 0.2167, 0.2537, 0.2343, 0.2709, 0.1456, 0.1560, 0.2507], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0620, 0.0684, 0.0502, 0.0673, 0.0710, 0.0535, 0.0674], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 15:16:24,780 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0276, 4.8366, 5.0210, 5.2050, 5.4257, 4.7744, 5.4394, 5.4000], device='cuda:3'), covar=tensor([0.2232, 0.1402, 0.1960, 0.0908, 0.0667, 0.0963, 0.0753, 0.1098], device='cuda:3'), in_proj_covar=tensor([0.0654, 0.0801, 0.0923, 0.0815, 0.0622, 0.0649, 0.0682, 0.0790], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:16:29,665 INFO [optim.py:368] (3/8) 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 15:17:06,890 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:17:12,009 INFO [train.py:904] (3/8) Epoch 28, batch 8750, loss[loss=0.1574, simple_loss=0.2442, pruned_loss=0.03532, over 12320.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2623, pruned_loss=0.03715, over 3039340.65 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:17:18,900 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282805.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:18:31,842 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3984, 3.2187, 3.5063, 1.9096, 3.6831, 3.7052, 2.9632, 2.8120], device='cuda:3'), covar=tensor([0.0811, 0.0307, 0.0228, 0.1162, 0.0088, 0.0183, 0.0452, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0108, 0.0099, 0.0135, 0.0083, 0.0127, 0.0126, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 15:18:46,444 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:18:49,350 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0467, 3.1223, 1.9874, 3.3039, 2.3672, 3.3004, 2.1083, 2.5750], device='cuda:3'), covar=tensor([0.0344, 0.0374, 0.1554, 0.0302, 0.0824, 0.0557, 0.1605, 0.0846], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0176, 0.0192, 0.0167, 0.0175, 0.0214, 0.0200, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:19:02,829 INFO [train.py:904] (3/8) Epoch 28, batch 8800, loss[loss=0.1782, simple_loss=0.2783, pruned_loss=0.03908, over 16975.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.261, pruned_loss=0.03596, over 3057336.77 frames. ], batch size: 109, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:19:29,170 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:20:00,725 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.123e+02 2.465e+02 3.098e+02 6.856e+02, threshold=4.929e+02, percent-clipped=4.0 2023-05-02 15:20:47,892 INFO [train.py:904] (3/8) Epoch 28, batch 8850, loss[loss=0.1745, simple_loss=0.2769, pruned_loss=0.03608, over 16109.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2644, pruned_loss=0.03566, over 3056290.87 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:22:35,072 INFO [train.py:904] (3/8) Epoch 28, batch 8900, loss[loss=0.1601, simple_loss=0.2599, pruned_loss=0.03017, over 16826.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2648, pruned_loss=0.03513, over 3063975.26 frames. ], batch size: 90, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:22:50,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5001, 3.4485, 3.5424, 3.6133, 3.6533, 3.3569, 3.6456, 3.7005], device='cuda:3'), covar=tensor([0.1412, 0.0975, 0.1094, 0.0682, 0.0658, 0.2296, 0.0770, 0.0720], device='cuda:3'), in_proj_covar=tensor([0.0644, 0.0788, 0.0910, 0.0805, 0.0612, 0.0639, 0.0672, 0.0777], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:22:54,736 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9737, 4.3035, 3.2036, 2.4666, 2.6223, 2.6427, 4.5987, 3.5901], device='cuda:3'), covar=tensor([0.2771, 0.0494, 0.1743, 0.2997, 0.2928, 0.2216, 0.0352, 0.1303], device='cuda:3'), in_proj_covar=tensor([0.0332, 0.0269, 0.0308, 0.0323, 0.0299, 0.0274, 0.0300, 0.0345], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 15:23:38,030 INFO [optim.py:368] (3/8) 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 15:24:19,110 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0183, 5.5211, 5.6917, 5.4202, 5.4812, 6.0055, 5.5283, 5.2373], device='cuda:3'), covar=tensor([0.0991, 0.1745, 0.2129, 0.2033, 0.2219, 0.0804, 0.1508, 0.2495], device='cuda:3'), in_proj_covar=tensor([0.0414, 0.0616, 0.0680, 0.0498, 0.0669, 0.0707, 0.0531, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 15:24:39,869 INFO [train.py:904] (3/8) Epoch 28, batch 8950, loss[loss=0.1423, simple_loss=0.242, pruned_loss=0.02125, over 15318.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2642, pruned_loss=0.03536, over 3067042.98 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:25:04,559 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283014.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:25:23,999 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6193, 2.4493, 2.3145, 3.9461, 2.2739, 3.8652, 1.4041, 2.9429], device='cuda:3'), covar=tensor([0.1478, 0.0855, 0.1338, 0.0127, 0.0106, 0.0317, 0.1864, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0196, 0.0202, 0.0213, 0.0206, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:25:45,364 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:26:27,357 INFO [train.py:904] (3/8) Epoch 28, batch 9000, loss[loss=0.1505, simple_loss=0.2439, pruned_loss=0.02857, over 16435.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2609, pruned_loss=0.03442, over 3060009.88 frames. ], batch size: 147, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:26:27,358 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 15:26:38,047 INFO [train.py:938] (3/8) Epoch 28, validation: loss=0.1436, simple_loss=0.2472, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 15:26:38,048 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 15:26:41,523 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283054.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:27:36,918 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.008e+02 2.381e+02 2.804e+02 4.985e+02, threshold=4.761e+02, percent-clipped=1.0 2023-05-02 15:28:19,314 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 15:28:21,092 INFO [train.py:904] (3/8) Epoch 28, batch 9050, loss[loss=0.1626, simple_loss=0.2527, pruned_loss=0.03623, over 16817.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2618, pruned_loss=0.03491, over 3069321.13 frames. ], batch size: 90, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:28:48,323 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:46,354 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:48,309 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:04,160 INFO [train.py:904] (3/8) Epoch 28, batch 9100, loss[loss=0.1602, simple_loss=0.2445, pruned_loss=0.03798, over 12380.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2609, pruned_loss=0.03501, over 3071909.71 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:30:05,683 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8078, 3.7532, 3.8890, 3.6679, 3.9138, 4.2813, 3.9768, 3.6553], device='cuda:3'), covar=tensor([0.2206, 0.2730, 0.2843, 0.2788, 0.2877, 0.2152, 0.1649, 0.2634], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0612, 0.0676, 0.0497, 0.0666, 0.0704, 0.0528, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 15:30:20,072 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283161.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:34,619 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 15:30:58,149 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:08,544 INFO [optim.py:368] (3/8) 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 15:31:38,532 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283193.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:47,827 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9152, 4.9132, 4.6769, 4.1384, 4.7788, 1.7813, 4.5229, 4.4901], device='cuda:3'), covar=tensor([0.0106, 0.0101, 0.0232, 0.0388, 0.0112, 0.2894, 0.0146, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0170, 0.0209, 0.0180, 0.0185, 0.0214, 0.0197, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:32:01,011 INFO [train.py:904] (3/8) Epoch 28, batch 9150, loss[loss=0.1636, simple_loss=0.2518, pruned_loss=0.03765, over 11951.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2613, pruned_loss=0.03459, over 3055160.53 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:32:06,387 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283205.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:32:44,762 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7728, 1.9939, 2.3558, 2.7484, 2.6608, 3.0811, 2.0142, 3.0852], device='cuda:3'), covar=tensor([0.0261, 0.0600, 0.0387, 0.0359, 0.0366, 0.0242, 0.0632, 0.0161], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0195, 0.0183, 0.0187, 0.0204, 0.0162, 0.0199, 0.0161], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:33:35,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6181, 4.4791, 4.6449, 4.7897, 4.9549, 4.4646, 4.9727, 4.9535], device='cuda:3'), covar=tensor([0.1904, 0.1175, 0.1510, 0.0781, 0.0575, 0.1119, 0.0637, 0.0927], device='cuda:3'), in_proj_covar=tensor([0.0645, 0.0789, 0.0911, 0.0806, 0.0613, 0.0639, 0.0673, 0.0778], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:33:44,276 INFO [train.py:904] (3/8) Epoch 28, batch 9200, loss[loss=0.1426, simple_loss=0.2328, pruned_loss=0.02616, over 12232.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2572, pruned_loss=0.03353, over 3071731.49 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:34:20,983 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 15:34:22,077 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4802, 4.7428, 4.4361, 4.1347, 3.7815, 4.6519, 4.4497, 4.2439], device='cuda:3'), covar=tensor([0.0957, 0.0762, 0.0518, 0.0468, 0.1715, 0.0639, 0.0656, 0.0890], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0453, 0.0352, 0.0351, 0.0347, 0.0405, 0.0242, 0.0419], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:34:34,283 INFO [optim.py:368] (3/8) 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 15:34:39,310 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:20,276 INFO [train.py:904] (3/8) Epoch 28, batch 9250, loss[loss=0.1453, simple_loss=0.2403, pruned_loss=0.02514, over 16615.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2574, pruned_loss=0.03366, over 3082983.07 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:35:44,160 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283314.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:28,139 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:36,869 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 15:36:55,349 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:37:14,584 INFO [train.py:904] (3/8) Epoch 28, batch 9300, loss[loss=0.1467, simple_loss=0.241, pruned_loss=0.02625, over 16475.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2559, pruned_loss=0.03308, over 3089067.99 frames. ], batch size: 68, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:37:16,909 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:01,710 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 15:38:16,508 INFO [optim.py:368] (3/8) 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 15:38:17,487 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283381.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:52,027 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8704, 2.8495, 2.5746, 4.6208, 3.1367, 4.1396, 1.6191, 3.0801], device='cuda:3'), covar=tensor([0.1345, 0.0781, 0.1221, 0.0168, 0.0160, 0.0375, 0.1724, 0.0714], device='cuda:3'), in_proj_covar=tensor([0.0170, 0.0176, 0.0196, 0.0196, 0.0200, 0.0212, 0.0206, 0.0195], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 15:38:58,083 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283402.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 15:38:59,593 INFO [train.py:904] (3/8) Epoch 28, batch 9350, loss[loss=0.1856, simple_loss=0.2765, pruned_loss=0.04735, over 16640.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2555, pruned_loss=0.03294, over 3095548.50 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:39:50,882 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 15:40:41,025 INFO [train.py:904] (3/8) Epoch 28, batch 9400, loss[loss=0.1798, simple_loss=0.2896, pruned_loss=0.03497, over 16859.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2557, pruned_loss=0.03327, over 3066460.00 frames. ], batch size: 90, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:41:00,134 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283461.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:18,936 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:35,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8669, 5.1730, 5.3166, 5.0839, 5.1459, 5.6763, 5.1948, 4.9219], device='cuda:3'), covar=tensor([0.0948, 0.1728, 0.1896, 0.1978, 0.2284, 0.0876, 0.1555, 0.2508], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0612, 0.0675, 0.0497, 0.0665, 0.0703, 0.0527, 0.0665], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 15:41:39,543 INFO [optim.py:368] (3/8) 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 15:42:06,289 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0118, 4.1276, 3.9536, 3.6241, 3.6978, 4.0541, 3.7347, 3.8129], device='cuda:3'), covar=tensor([0.0634, 0.0584, 0.0344, 0.0348, 0.0730, 0.0537, 0.1033, 0.0638], device='cuda:3'), in_proj_covar=tensor([0.0300, 0.0451, 0.0351, 0.0351, 0.0345, 0.0405, 0.0241, 0.0417], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:42:20,564 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:42:24,701 INFO [train.py:904] (3/8) Epoch 28, batch 9450, loss[loss=0.1758, simple_loss=0.2636, pruned_loss=0.04397, over 12591.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2577, pruned_loss=0.03367, over 3053926.32 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:42:36,846 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283509.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:43:01,661 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2008, 2.2833, 2.2410, 3.9049, 2.1833, 2.6621, 2.3652, 2.4007], device='cuda:3'), covar=tensor([0.1354, 0.3808, 0.3323, 0.0549, 0.4397, 0.2517, 0.3873, 0.3587], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0461, 0.0378, 0.0325, 0.0436, 0.0528, 0.0436, 0.0540], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:43:48,941 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283544.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:44:06,791 INFO [train.py:904] (3/8) Epoch 28, batch 9500, loss[loss=0.152, simple_loss=0.2495, pruned_loss=0.02725, over 16360.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2566, pruned_loss=0.03306, over 3057837.32 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:03,873 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.031e+02 2.419e+02 2.973e+02 5.266e+02, threshold=4.838e+02, percent-clipped=2.0 2023-05-02 15:45:53,240 INFO [train.py:904] (3/8) Epoch 28, batch 9550, loss[loss=0.1657, simple_loss=0.2719, pruned_loss=0.02975, over 15304.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2559, pruned_loss=0.03286, over 3059610.34 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:59,193 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283605.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:47:08,472 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9382, 2.7612, 2.6452, 1.9813, 2.5564, 2.8262, 2.6236, 1.9699], device='cuda:3'), covar=tensor([0.0412, 0.0083, 0.0083, 0.0354, 0.0154, 0.0101, 0.0113, 0.0437], device='cuda:3'), in_proj_covar=tensor([0.0133, 0.0086, 0.0087, 0.0130, 0.0099, 0.0110, 0.0094, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 15:47:09,794 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 9600, loss[loss=0.1375, simple_loss=0.2335, pruned_loss=0.02075, over 17151.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2573, pruned_loss=0.03333, over 3057690.06 frames. ], batch size: 49, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:48:11,555 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7923, 4.8583, 4.6806, 4.2690, 4.3863, 4.7799, 4.5909, 4.4713], device='cuda:3'), covar=tensor([0.0645, 0.0635, 0.0320, 0.0345, 0.0949, 0.0576, 0.0381, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0448, 0.0350, 0.0349, 0.0345, 0.0403, 0.0240, 0.0415], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:48:29,440 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.192e+02 2.621e+02 3.097e+02 5.622e+02, threshold=5.243e+02, percent-clipped=5.0 2023-05-02 15:49:23,006 INFO [train.py:904] (3/8) Epoch 28, batch 9650, loss[loss=0.1665, simple_loss=0.2637, pruned_loss=0.03465, over 16963.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2595, pruned_loss=0.03391, over 3054198.18 frames. ], batch size: 109, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:10,204 INFO [train.py:904] (3/8) Epoch 28, batch 9700, loss[loss=0.1705, simple_loss=0.2724, pruned_loss=0.03435, over 15364.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2584, pruned_loss=0.0334, over 3063151.24 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:43,415 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9153, 2.1971, 2.3822, 3.2148, 2.1950, 2.3941, 2.3723, 2.2913], device='cuda:3'), covar=tensor([0.1342, 0.3727, 0.2862, 0.0725, 0.4502, 0.2713, 0.3519, 0.3822], device='cuda:3'), in_proj_covar=tensor([0.0409, 0.0462, 0.0378, 0.0325, 0.0437, 0.0528, 0.0436, 0.0540], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:51:47,282 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283771.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:08,986 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.180e+02 2.423e+02 2.886e+02 5.203e+02, threshold=4.846e+02, percent-clipped=0.0 2023-05-02 15:52:48,099 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 9750, loss[loss=0.1498, simple_loss=0.2524, pruned_loss=0.02359, over 15260.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2569, pruned_loss=0.03344, over 3051080.10 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:53:17,450 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0572, 4.1595, 3.9634, 3.6827, 3.7730, 4.0866, 3.7395, 3.8894], device='cuda:3'), covar=tensor([0.0606, 0.0745, 0.0296, 0.0260, 0.0629, 0.0504, 0.0932, 0.0566], device='cuda:3'), in_proj_covar=tensor([0.0299, 0.0449, 0.0350, 0.0349, 0.0345, 0.0404, 0.0240, 0.0415], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:53:24,760 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283819.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:13,291 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283842.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:24,070 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 9800, loss[loss=0.1707, simple_loss=0.2683, pruned_loss=0.03656, over 16864.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2575, pruned_loss=0.03271, over 3069275.94 frames. ], batch size: 116, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:55:23,051 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.061e+02 2.416e+02 3.000e+02 5.773e+02, threshold=4.831e+02, percent-clipped=3.0 2023-05-02 15:55:38,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1954, 2.3088, 2.4220, 3.9790, 2.2399, 2.6414, 2.4101, 2.4674], device='cuda:3'), covar=tensor([0.1376, 0.3765, 0.3217, 0.0565, 0.4300, 0.2708, 0.3759, 0.3671], device='cuda:3'), in_proj_covar=tensor([0.0410, 0.0463, 0.0379, 0.0325, 0.0438, 0.0529, 0.0436, 0.0541], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 15:56:11,011 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283900.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:56:15,639 INFO [train.py:904] (3/8) Epoch 28, batch 9850, loss[loss=0.1574, simple_loss=0.2551, pruned_loss=0.02981, over 16853.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2589, pruned_loss=0.03259, over 3075834.00 frames. ], batch size: 124, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:56:17,417 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:57:36,902 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:58:06,898 INFO [train.py:904] (3/8) Epoch 28, batch 9900, loss[loss=0.1665, simple_loss=0.2695, pruned_loss=0.03177, over 15292.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2596, pruned_loss=0.03281, over 3062473.48 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:58:38,491 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:59:13,294 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:00:07,756 INFO [train.py:904] (3/8) Epoch 28, batch 9950, loss[loss=0.1699, simple_loss=0.2722, pruned_loss=0.03381, over 16327.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2615, pruned_loss=0.03323, over 3065106.98 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:01:08,221 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 28, batch 10000, loss[loss=0.1572, simple_loss=0.2606, pruned_loss=0.02696, over 16266.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2603, pruned_loss=0.03288, over 3078280.22 frames. ], batch size: 166, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:02:27,828 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 16:02:47,662 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9385, 3.7503, 4.0934, 2.1933, 4.2615, 4.3280, 3.3981, 3.3268], device='cuda:3'), covar=tensor([0.0638, 0.0252, 0.0217, 0.1152, 0.0067, 0.0140, 0.0328, 0.0409], device='cuda:3'), in_proj_covar=tensor([0.0143, 0.0106, 0.0097, 0.0133, 0.0082, 0.0124, 0.0124, 0.0125], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 16:03:03,885 INFO [optim.py:368] (3/8) 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 16:03:50,387 INFO [train.py:904] (3/8) Epoch 28, batch 10050, loss[loss=0.1716, simple_loss=0.2649, pruned_loss=0.03915, over 16946.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2607, pruned_loss=0.03287, over 3091298.60 frames. ], batch size: 109, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:05:24,422 INFO [train.py:904] (3/8) Epoch 28, batch 10100, loss[loss=0.1495, simple_loss=0.2448, pruned_loss=0.02707, over 16827.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2604, pruned_loss=0.03294, over 3095511.75 frames. ], batch size: 83, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:06:20,722 INFO [optim.py:368] (3/8) 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 16:06:34,369 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 16:06:37,173 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7924, 5.0230, 5.1662, 4.9656, 5.0118, 5.5694, 5.0717, 4.8128], device='cuda:3'), covar=tensor([0.1073, 0.1887, 0.2133, 0.2084, 0.2374, 0.0888, 0.1560, 0.2366], device='cuda:3'), in_proj_covar=tensor([0.0403, 0.0602, 0.0668, 0.0492, 0.0658, 0.0694, 0.0520, 0.0656], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 16:06:39,382 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:06:42,003 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284200.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:06:44,473 INFO [train.py:904] (3/8) Epoch 28, batch 10150, loss[loss=0.1558, simple_loss=0.2413, pruned_loss=0.03513, over 12194.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2595, pruned_loss=0.03319, over 3065021.40 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:07:10,315 INFO [train.py:904] (3/8) Epoch 29, batch 0, loss[loss=0.2134, simple_loss=0.2983, pruned_loss=0.06427, over 17060.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2983, pruned_loss=0.06427, over 17060.00 frames. ], batch size: 53, lr: 2.34e-03, grad_scale: 8.0 2023-05-02 16:07:10,315 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 16:07:17,746 INFO [train.py:938] (3/8) Epoch 29, validation: loss=0.1427, simple_loss=0.246, pruned_loss=0.0197, over 944034.00 frames. 2023-05-02 16:07:17,746 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 16:07:53,594 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 16:08:07,723 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4123, 3.5333, 3.7848, 2.6097, 3.4887, 3.8614, 3.5505, 2.0292], device='cuda:3'), covar=tensor([0.0624, 0.0393, 0.0082, 0.0478, 0.0149, 0.0128, 0.0140, 0.0646], device='cuda:3'), in_proj_covar=tensor([0.0135, 0.0087, 0.0088, 0.0132, 0.0100, 0.0111, 0.0095, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 16:08:10,065 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5036, 3.5577, 4.0328, 2.2850, 3.2605, 2.5993, 3.8289, 3.9031], device='cuda:3'), covar=tensor([0.0292, 0.1119, 0.0574, 0.2255, 0.0922, 0.1145, 0.0688, 0.1199], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0163, 0.0165, 0.0153, 0.0143, 0.0129, 0.0141, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 16:08:18,231 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 50, loss[loss=0.1804, simple_loss=0.2752, pruned_loss=0.04284, over 16730.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2647, pruned_loss=0.04436, over 753432.16 frames. ], batch size: 57, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:09:03,191 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8677, 2.9466, 3.2233, 2.1413, 2.8384, 2.2370, 3.3666, 3.3968], device='cuda:3'), covar=tensor([0.0268, 0.1026, 0.0668, 0.2029, 0.0945, 0.1103, 0.0594, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0156, 0.0164, 0.0166, 0.0153, 0.0144, 0.0129, 0.0142, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 16:09:08,277 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 100, loss[loss=0.1464, simple_loss=0.2395, pruned_loss=0.02662, over 16793.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2593, pruned_loss=0.04155, over 1324572.05 frames. ], batch size: 39, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:10:02,064 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:10:46,098 INFO [train.py:904] (3/8) Epoch 29, batch 150, loss[loss=0.169, simple_loss=0.2527, pruned_loss=0.04266, over 16515.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2581, pruned_loss=0.0413, over 1770674.85 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:10:48,634 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 16:11:25,623 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 200, loss[loss=0.1962, simple_loss=0.2646, pruned_loss=0.06391, over 16750.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2591, pruned_loss=0.04166, over 2118997.06 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:11:55,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7617, 4.9417, 5.0509, 4.8199, 4.8668, 5.5174, 5.0034, 4.7381], device='cuda:3'), covar=tensor([0.1485, 0.2257, 0.3018, 0.2643, 0.2997, 0.1193, 0.1966, 0.2742], device='cuda:3'), in_proj_covar=tensor([0.0415, 0.0621, 0.0689, 0.0507, 0.0676, 0.0712, 0.0534, 0.0675], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 16:11:55,942 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 16:12:51,492 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284443.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:13:01,996 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6660, 4.7784, 4.9281, 4.6982, 4.7517, 5.3952, 4.8386, 4.5471], device='cuda:3'), covar=tensor([0.1542, 0.2176, 0.2749, 0.2519, 0.2824, 0.1214, 0.2049, 0.2778], device='cuda:3'), in_proj_covar=tensor([0.0417, 0.0625, 0.0693, 0.0510, 0.0680, 0.0716, 0.0538, 0.0679], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 16:13:04,662 INFO [train.py:904] (3/8) Epoch 29, batch 250, loss[loss=0.1797, simple_loss=0.2538, pruned_loss=0.05281, over 16862.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2565, pruned_loss=0.04152, over 2388310.40 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:13:11,808 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6374, 3.7938, 2.4635, 4.3512, 2.9230, 4.2747, 2.5205, 3.1491], device='cuda:3'), covar=tensor([0.0395, 0.0463, 0.1727, 0.0358, 0.0901, 0.0638, 0.1714, 0.0881], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0171, 0.0180, 0.0218, 0.0205, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 16:13:24,383 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1465, 2.0988, 2.7576, 3.0988, 2.8906, 3.5292, 2.1111, 3.5758], device='cuda:3'), covar=tensor([0.0236, 0.0663, 0.0348, 0.0327, 0.0380, 0.0266, 0.0730, 0.0171], device='cuda:3'), in_proj_covar=tensor([0.0194, 0.0197, 0.0184, 0.0188, 0.0206, 0.0163, 0.0200, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:13:36,049 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-05-02 16:13:47,576 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.177e+02 2.500e+02 3.084e+02 6.085e+02, threshold=5.001e+02, percent-clipped=1.0 2023-05-02 16:14:08,303 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284498.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:17,000 INFO [train.py:904] (3/8) Epoch 29, batch 300, loss[loss=0.1803, simple_loss=0.2543, pruned_loss=0.05312, over 16813.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.254, pruned_loss=0.04052, over 2592204.54 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:14:17,393 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284504.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:15:14,279 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284546.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:15:23,225 INFO [train.py:904] (3/8) Epoch 29, batch 350, loss[loss=0.1641, simple_loss=0.2408, pruned_loss=0.04367, over 16839.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2519, pruned_loss=0.03974, over 2748985.15 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:15:40,491 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 16:16:02,935 INFO [optim.py:368] (3/8) 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 16:16:31,999 INFO [train.py:904] (3/8) Epoch 29, batch 400, loss[loss=0.1752, simple_loss=0.2581, pruned_loss=0.04612, over 16874.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2506, pruned_loss=0.03933, over 2866313.35 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:16:34,203 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4145, 4.0529, 4.5135, 2.4889, 4.7215, 4.8051, 3.5740, 3.8095], device='cuda:3'), covar=tensor([0.0657, 0.0286, 0.0258, 0.1130, 0.0076, 0.0179, 0.0420, 0.0394], device='cuda:3'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0137, 0.0085, 0.0129, 0.0128, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 16:16:57,104 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:17:41,206 INFO [train.py:904] (3/8) Epoch 29, batch 450, loss[loss=0.1846, simple_loss=0.2777, pruned_loss=0.04569, over 17081.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2499, pruned_loss=0.03842, over 2972757.30 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:17:59,177 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-05-02 16:18:02,994 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284670.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:18:11,903 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9390, 5.2831, 5.3734, 5.2356, 5.2405, 5.8165, 5.2921, 5.0788], device='cuda:3'), covar=tensor([0.1086, 0.1999, 0.3023, 0.2223, 0.2826, 0.1118, 0.1745, 0.2273], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0632, 0.0703, 0.0517, 0.0691, 0.0724, 0.0545, 0.0687], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 16:18:18,523 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 500, loss[loss=0.1805, simple_loss=0.2519, pruned_loss=0.05453, over 16760.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2484, pruned_loss=0.03804, over 3039295.55 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:19:56,204 INFO [train.py:904] (3/8) Epoch 29, batch 550, loss[loss=0.1795, simple_loss=0.2629, pruned_loss=0.04808, over 16353.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2479, pruned_loss=0.03752, over 3101585.76 frames. ], batch size: 145, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:19:59,644 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3419, 5.3137, 5.0840, 4.5483, 5.1029, 2.0364, 4.8935, 5.0316], device='cuda:3'), covar=tensor([0.0107, 0.0099, 0.0244, 0.0422, 0.0104, 0.2802, 0.0149, 0.0219], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0173, 0.0211, 0.0181, 0.0187, 0.0217, 0.0198, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:20:35,714 INFO [optim.py:368] (3/8) 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 16:20:38,817 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-02 16:20:57,907 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:04,186 INFO [train.py:904] (3/8) Epoch 29, batch 600, loss[loss=0.1516, simple_loss=0.2469, pruned_loss=0.02818, over 17190.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2479, pruned_loss=0.03741, over 3155799.31 frames. ], batch size: 46, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:21:07,176 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4018, 4.0068, 4.5032, 2.4397, 4.6818, 4.7790, 3.5113, 3.8061], device='cuda:3'), covar=tensor([0.0640, 0.0264, 0.0222, 0.1157, 0.0081, 0.0168, 0.0472, 0.0401], device='cuda:3'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0139, 0.0086, 0.0130, 0.0129, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 16:21:17,565 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 16:21:21,991 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:22:12,347 INFO [train.py:904] (3/8) Epoch 29, batch 650, loss[loss=0.1517, simple_loss=0.2491, pruned_loss=0.02714, over 17206.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2472, pruned_loss=0.03701, over 3190712.99 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:22:37,633 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0511, 5.5315, 5.6267, 5.3427, 5.4559, 6.0464, 5.5505, 5.2819], device='cuda:3'), covar=tensor([0.1063, 0.2038, 0.2575, 0.2382, 0.2693, 0.0999, 0.1528, 0.2407], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0637, 0.0708, 0.0519, 0.0695, 0.0729, 0.0548, 0.0691], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 16:22:46,722 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:22:53,855 INFO [optim.py:368] (3/8) 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 16:23:22,616 INFO [train.py:904] (3/8) Epoch 29, batch 700, loss[loss=0.1686, simple_loss=0.2629, pruned_loss=0.03711, over 17132.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2467, pruned_loss=0.03654, over 3210428.81 frames. ], batch size: 48, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:24:12,249 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:24:30,116 INFO [train.py:904] (3/8) Epoch 29, batch 750, loss[loss=0.1734, simple_loss=0.2569, pruned_loss=0.04495, over 16806.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2471, pruned_loss=0.03665, over 3234831.33 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:25:13,140 INFO [optim.py:368] (3/8) 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 16:25:17,742 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0260, 3.9164, 4.0957, 4.1997, 4.2579, 3.8526, 4.1074, 4.2882], device='cuda:3'), covar=tensor([0.1640, 0.1197, 0.1287, 0.0737, 0.0721, 0.1621, 0.2749, 0.0985], device='cuda:3'), in_proj_covar=tensor([0.0683, 0.0833, 0.0962, 0.0847, 0.0644, 0.0669, 0.0713, 0.0822], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:25:30,054 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:37,987 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:42,931 INFO [train.py:904] (3/8) Epoch 29, batch 800, loss[loss=0.1625, simple_loss=0.2381, pruned_loss=0.04342, over 16863.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2463, pruned_loss=0.03629, over 3249560.95 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:26:34,281 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285042.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:26:49,079 INFO [train.py:904] (3/8) Epoch 29, batch 850, loss[loss=0.1377, simple_loss=0.2194, pruned_loss=0.02796, over 16801.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2462, pruned_loss=0.03582, over 3266565.79 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:26:53,043 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 16:27:49,570 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:54,201 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285103.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 16:27:55,639 INFO [train.py:904] (3/8) Epoch 29, batch 900, loss[loss=0.1825, simple_loss=0.2617, pruned_loss=0.05166, over 16869.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.2449, pruned_loss=0.03505, over 3287749.48 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:28:34,580 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.85 vs. limit=5.0 2023-05-02 16:28:55,566 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285147.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:05,599 INFO [train.py:904] (3/8) Epoch 29, batch 950, loss[loss=0.1703, simple_loss=0.2597, pruned_loss=0.04047, over 17015.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2449, pruned_loss=0.03493, over 3303902.83 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:29:30,563 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:38,308 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8663, 2.8881, 2.6248, 2.8338, 3.1748, 2.9659, 3.4685, 3.3457], device='cuda:3'), covar=tensor([0.0171, 0.0464, 0.0543, 0.0479, 0.0347, 0.0442, 0.0250, 0.0315], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0247, 0.0236, 0.0237, 0.0247, 0.0246, 0.0243, 0.0246], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:29:47,164 INFO [optim.py:368] (3/8) 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 16:30:14,195 INFO [train.py:904] (3/8) Epoch 29, batch 1000, loss[loss=0.1666, simple_loss=0.2497, pruned_loss=0.04176, over 11624.00 frames. ], tot_loss[loss=0.1567, simple_loss=0.2437, pruned_loss=0.03485, over 3300201.34 frames. ], batch size: 246, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:31:02,857 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5180, 3.5634, 4.1965, 2.2656, 3.3528, 2.6594, 3.9407, 3.8070], device='cuda:3'), covar=tensor([0.0252, 0.1071, 0.0440, 0.2142, 0.0832, 0.1000, 0.0565, 0.1134], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 16:31:21,134 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1774, 3.2773, 3.6044, 2.2448, 3.0501, 2.4700, 3.6246, 3.6316], device='cuda:3'), covar=tensor([0.0296, 0.1080, 0.0632, 0.2157, 0.0920, 0.1042, 0.0648, 0.1116], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 16:31:24,212 INFO [train.py:904] (3/8) Epoch 29, batch 1050, loss[loss=0.16, simple_loss=0.2521, pruned_loss=0.03394, over 17128.00 frames. ], tot_loss[loss=0.1563, simple_loss=0.2432, pruned_loss=0.03467, over 3310147.03 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:32:05,303 INFO [optim.py:368] (3/8) 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 16:32:19,764 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:32:31,238 INFO [train.py:904] (3/8) Epoch 29, batch 1100, loss[loss=0.1502, simple_loss=0.2443, pruned_loss=0.02805, over 17241.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2433, pruned_loss=0.03484, over 3317204.38 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:33:37,331 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:33:37,798 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.08 vs. limit=2.0 2023-05-02 16:33:40,187 INFO [train.py:904] (3/8) Epoch 29, batch 1150, loss[loss=0.1342, simple_loss=0.222, pruned_loss=0.02324, over 16983.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.243, pruned_loss=0.03445, over 3329915.15 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:34:00,655 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.7310, 6.0965, 5.8253, 5.9326, 5.5330, 5.6107, 5.4852, 6.2470], device='cuda:3'), covar=tensor([0.1436, 0.0944, 0.1051, 0.0912, 0.0879, 0.0625, 0.1302, 0.0814], device='cuda:3'), in_proj_covar=tensor([0.0725, 0.0877, 0.0717, 0.0681, 0.0555, 0.0553, 0.0738, 0.0687], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:34:06,990 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8778, 2.2559, 2.5233, 3.1456, 2.2813, 2.3873, 2.4412, 2.3704], device='cuda:3'), covar=tensor([0.1575, 0.3593, 0.2838, 0.0923, 0.4248, 0.2779, 0.3460, 0.3744], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0477, 0.0391, 0.0338, 0.0449, 0.0546, 0.0450, 0.0560], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:34:22,232 INFO [optim.py:368] (3/8) 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 16:34:40,261 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285398.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:34:47,458 INFO [train.py:904] (3/8) Epoch 29, batch 1200, loss[loss=0.1396, simple_loss=0.2186, pruned_loss=0.03031, over 16725.00 frames. ], tot_loss[loss=0.1558, simple_loss=0.2424, pruned_loss=0.0346, over 3332930.22 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:35:22,688 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 16:35:56,660 INFO [train.py:904] (3/8) Epoch 29, batch 1250, loss[loss=0.1621, simple_loss=0.2523, pruned_loss=0.03589, over 17011.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.243, pruned_loss=0.03596, over 3326518.64 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:36:21,860 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:36:38,698 INFO [optim.py:368] (3/8) 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 16:37:05,030 INFO [train.py:904] (3/8) Epoch 29, batch 1300, loss[loss=0.1529, simple_loss=0.2489, pruned_loss=0.02848, over 17097.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2431, pruned_loss=0.0358, over 3321638.02 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:37:17,089 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1686, 2.4782, 2.6159, 1.9848, 2.7451, 2.7745, 2.5128, 2.4314], device='cuda:3'), covar=tensor([0.0727, 0.0312, 0.0313, 0.0991, 0.0178, 0.0328, 0.0504, 0.0504], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 16:37:27,840 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285520.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:38:00,095 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.6442, 5.9905, 5.7541, 5.7926, 5.3266, 5.5437, 5.3094, 6.1302], device='cuda:3'), covar=tensor([0.1466, 0.1059, 0.1090, 0.1021, 0.0952, 0.0628, 0.1373, 0.0997], device='cuda:3'), in_proj_covar=tensor([0.0727, 0.0878, 0.0719, 0.0681, 0.0556, 0.0554, 0.0738, 0.0689], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:38:13,879 INFO [train.py:904] (3/8) Epoch 29, batch 1350, loss[loss=0.17, simple_loss=0.2579, pruned_loss=0.04108, over 16494.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2432, pruned_loss=0.03563, over 3317669.50 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:38:58,361 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.007e+02 2.357e+02 2.705e+02 4.553e+02, threshold=4.714e+02, percent-clipped=0.0 2023-05-02 16:39:11,787 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285595.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:39:17,465 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4282, 2.4212, 2.3873, 4.2185, 2.3491, 2.7818, 2.4554, 2.5736], device='cuda:3'), covar=tensor([0.1364, 0.3779, 0.3404, 0.0552, 0.4259, 0.2745, 0.3846, 0.3731], device='cuda:3'), in_proj_covar=tensor([0.0425, 0.0479, 0.0392, 0.0339, 0.0450, 0.0549, 0.0452, 0.0562], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:39:24,823 INFO [train.py:904] (3/8) Epoch 29, batch 1400, loss[loss=0.1532, simple_loss=0.2307, pruned_loss=0.03788, over 16465.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2438, pruned_loss=0.03541, over 3320165.66 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:39:25,751 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9682, 5.3294, 5.5977, 5.5643, 5.6065, 5.2765, 4.9322, 5.0752], device='cuda:3'), covar=tensor([0.0725, 0.1015, 0.0710, 0.0863, 0.0866, 0.0797, 0.1553, 0.0613], device='cuda:3'), in_proj_covar=tensor([0.0445, 0.0503, 0.0485, 0.0446, 0.0532, 0.0508, 0.0584, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 16:39:31,844 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 16:39:49,588 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:39:52,581 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1098, 5.1758, 5.5863, 5.5768, 5.5902, 5.2131, 5.2079, 5.0124], device='cuda:3'), covar=tensor([0.0359, 0.0513, 0.0387, 0.0382, 0.0468, 0.0416, 0.0958, 0.0440], device='cuda:3'), in_proj_covar=tensor([0.0445, 0.0502, 0.0485, 0.0446, 0.0532, 0.0508, 0.0583, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 16:40:19,785 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:31,731 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 1450, loss[loss=0.1577, simple_loss=0.2485, pruned_loss=0.03346, over 17190.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2434, pruned_loss=0.0355, over 3307151.39 frames. ], batch size: 46, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:40:50,662 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 16:41:13,935 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:14,297 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 16:41:16,389 INFO [optim.py:368] (3/8) 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 16:41:34,314 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285698.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:41:37,509 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 1500, loss[loss=0.1629, simple_loss=0.2452, pruned_loss=0.04031, over 12530.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2434, pruned_loss=0.03543, over 3309870.07 frames. ], batch size: 247, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:42:40,636 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285746.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:42:51,734 INFO [train.py:904] (3/8) Epoch 29, batch 1550, loss[loss=0.1924, simple_loss=0.2619, pruned_loss=0.06149, over 16913.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2451, pruned_loss=0.03654, over 3306312.22 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:43:34,663 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.209e+02 2.639e+02 3.034e+02 6.721e+02, threshold=5.278e+02, percent-clipped=2.0 2023-05-02 16:43:44,865 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-02 16:44:00,617 INFO [train.py:904] (3/8) Epoch 29, batch 1600, loss[loss=0.1538, simple_loss=0.2429, pruned_loss=0.0323, over 16832.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2468, pruned_loss=0.03683, over 3305774.17 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:44:30,856 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 16:45:09,517 INFO [train.py:904] (3/8) Epoch 29, batch 1650, loss[loss=0.1472, simple_loss=0.2447, pruned_loss=0.02487, over 17131.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2483, pruned_loss=0.0368, over 3317342.61 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:45:50,989 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 1700, loss[loss=0.1718, simple_loss=0.263, pruned_loss=0.04031, over 17177.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2496, pruned_loss=0.03748, over 3315771.00 frames. ], batch size: 46, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:46:58,566 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-05-02 16:47:24,308 INFO [train.py:904] (3/8) Epoch 29, batch 1750, loss[loss=0.1715, simple_loss=0.2592, pruned_loss=0.0419, over 15420.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2505, pruned_loss=0.03782, over 3308497.19 frames. ], batch size: 190, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:47:24,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0927, 3.2702, 3.0725, 1.9600, 2.6684, 2.1523, 3.4618, 3.6223], device='cuda:3'), covar=tensor([0.0222, 0.0892, 0.0761, 0.2392, 0.1149, 0.1234, 0.0593, 0.0900], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0158, 0.0149, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 16:47:58,302 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.447e+02 2.051e+02 2.355e+02 2.951e+02 5.834e+02, threshold=4.710e+02, percent-clipped=3.0 2023-05-02 16:48:36,742 INFO [train.py:904] (3/8) Epoch 29, batch 1800, loss[loss=0.1717, simple_loss=0.256, pruned_loss=0.04373, over 16172.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2513, pruned_loss=0.03782, over 3317895.73 frames. ], batch size: 165, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:49:43,957 INFO [train.py:904] (3/8) Epoch 29, batch 1850, loss[loss=0.1562, simple_loss=0.2528, pruned_loss=0.02979, over 17074.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2518, pruned_loss=0.03794, over 3313005.19 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:50:28,094 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 1900, loss[loss=0.1618, simple_loss=0.2583, pruned_loss=0.03264, over 17053.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2513, pruned_loss=0.03749, over 3309078.16 frames. ], batch size: 50, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:51:18,931 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2309, 5.7542, 5.9044, 5.5939, 5.6313, 6.2224, 5.7556, 5.4912], device='cuda:3'), covar=tensor([0.0939, 0.2132, 0.2716, 0.2036, 0.2901, 0.1076, 0.1520, 0.2122], device='cuda:3'), in_proj_covar=tensor([0.0436, 0.0647, 0.0721, 0.0529, 0.0709, 0.0740, 0.0557, 0.0705], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 16:51:20,198 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286123.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:51:29,261 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:52:04,245 INFO [train.py:904] (3/8) Epoch 29, batch 1950, loss[loss=0.1706, simple_loss=0.262, pruned_loss=0.0396, over 17028.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2521, pruned_loss=0.03738, over 3307183.10 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:52:28,933 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4516, 3.5527, 3.8605, 2.7015, 3.4661, 3.8813, 3.5857, 2.3007], device='cuda:3'), covar=tensor([0.0545, 0.0267, 0.0074, 0.0426, 0.0145, 0.0140, 0.0121, 0.0527], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 16:52:42,635 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9792, 4.9555, 4.7663, 4.2653, 4.8859, 1.9888, 4.6435, 4.6000], device='cuda:3'), covar=tensor([0.0129, 0.0115, 0.0224, 0.0398, 0.0116, 0.2777, 0.0167, 0.0224], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0177, 0.0216, 0.0186, 0.0193, 0.0220, 0.0204, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:52:46,851 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286184.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:52:48,709 INFO [optim.py:368] (3/8) 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 16:52:55,344 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 2000, loss[loss=0.164, simple_loss=0.2518, pruned_loss=0.03811, over 16271.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.252, pruned_loss=0.03737, over 3312761.65 frames. ], batch size: 165, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:21,795 INFO [train.py:904] (3/8) Epoch 29, batch 2050, loss[loss=0.1767, simple_loss=0.2524, pruned_loss=0.0505, over 16728.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2525, pruned_loss=0.03776, over 3305893.69 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:54,851 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.125e+02 2.436e+02 2.931e+02 6.185e+02, threshold=4.871e+02, percent-clipped=3.0 2023-05-02 16:55:18,543 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 2100, loss[loss=0.1747, simple_loss=0.2646, pruned_loss=0.04244, over 16484.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2526, pruned_loss=0.03803, over 3310937.59 frames. ], batch size: 68, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:00,346 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286326.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:26,322 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:40,457 INFO [train.py:904] (3/8) Epoch 29, batch 2150, loss[loss=0.1352, simple_loss=0.2226, pruned_loss=0.02385, over 17011.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2531, pruned_loss=0.0386, over 3303817.77 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:44,715 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286356.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:52,440 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 16:56:57,445 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9459, 3.0440, 2.6549, 2.8713, 3.2413, 3.0999, 3.5216, 3.4940], device='cuda:3'), covar=tensor([0.0182, 0.0429, 0.0512, 0.0423, 0.0312, 0.0398, 0.0306, 0.0282], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0254, 0.0241, 0.0242, 0.0254, 0.0252, 0.0250, 0.0252], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:57:06,918 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2622, 4.3896, 4.6711, 4.6470, 4.7097, 4.4250, 4.3347, 4.3096], device='cuda:3'), covar=tensor([0.0593, 0.0901, 0.0546, 0.0617, 0.0785, 0.0683, 0.1337, 0.0819], device='cuda:3'), in_proj_covar=tensor([0.0452, 0.0510, 0.0492, 0.0454, 0.0541, 0.0516, 0.0592, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 16:57:11,648 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0071, 4.7527, 5.0339, 5.2205, 5.4059, 4.7900, 5.3736, 5.3791], device='cuda:3'), covar=tensor([0.1943, 0.1614, 0.1940, 0.0813, 0.0550, 0.0988, 0.0619, 0.0701], device='cuda:3'), in_proj_covar=tensor([0.0704, 0.0864, 0.0998, 0.0876, 0.0666, 0.0691, 0.0733, 0.0845], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 16:57:24,982 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 2200, loss[loss=0.2165, simple_loss=0.2984, pruned_loss=0.06729, over 11782.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2545, pruned_loss=0.0397, over 3287992.69 frames. ], batch size: 247, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:57:52,990 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:58:59,516 INFO [train.py:904] (3/8) Epoch 29, batch 2250, loss[loss=0.1555, simple_loss=0.2365, pruned_loss=0.0372, over 16325.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2558, pruned_loss=0.03985, over 3280034.25 frames. ], batch size: 36, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:59:16,122 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9497, 2.1541, 2.6491, 2.8574, 2.7464, 3.4028, 2.5230, 3.3727], device='cuda:3'), covar=tensor([0.0287, 0.0649, 0.0354, 0.0434, 0.0441, 0.0251, 0.0517, 0.0192], device='cuda:3'), in_proj_covar=tensor([0.0203, 0.0203, 0.0190, 0.0197, 0.0214, 0.0170, 0.0207, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 16:59:33,867 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286479.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 16:59:43,673 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:00:08,670 INFO [train.py:904] (3/8) Epoch 29, batch 2300, loss[loss=0.1719, simple_loss=0.261, pruned_loss=0.04137, over 17109.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2552, pruned_loss=0.03941, over 3296393.09 frames. ], batch size: 53, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:00:11,532 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 17:00:26,956 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3411, 2.8264, 3.0998, 1.9845, 3.2103, 3.2014, 2.7223, 2.5649], device='cuda:3'), covar=tensor([0.0823, 0.0338, 0.0273, 0.1055, 0.0149, 0.0282, 0.0526, 0.0484], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0113, 0.0104, 0.0141, 0.0088, 0.0134, 0.0132, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 17:01:17,609 INFO [train.py:904] (3/8) Epoch 29, batch 2350, loss[loss=0.1608, simple_loss=0.2509, pruned_loss=0.03535, over 17223.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2553, pruned_loss=0.03955, over 3298203.50 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:02:03,068 INFO [optim.py:368] (3/8) 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 17:02:04,826 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286588.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:02:16,972 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1489, 3.2669, 3.5624, 2.2936, 3.0366, 2.4168, 3.6607, 3.6356], device='cuda:3'), covar=tensor([0.0260, 0.0955, 0.0659, 0.1973, 0.0899, 0.1064, 0.0491, 0.0887], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0172, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 17:02:27,440 INFO [train.py:904] (3/8) Epoch 29, batch 2400, loss[loss=0.1742, simple_loss=0.248, pruned_loss=0.05022, over 16426.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2548, pruned_loss=0.03949, over 3307802.13 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 17:03:30,979 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:34,191 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286651.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:37,416 INFO [train.py:904] (3/8) Epoch 29, batch 2450, loss[loss=0.1597, simple_loss=0.2555, pruned_loss=0.0319, over 17111.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2554, pruned_loss=0.0394, over 3317458.07 frames. ], batch size: 49, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:03:59,884 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:04:25,640 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 17:04:41,082 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 2500, loss[loss=0.1497, simple_loss=0.246, pruned_loss=0.02669, over 17234.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2549, pruned_loss=0.0392, over 3309575.30 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:04:47,878 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8289, 2.5047, 2.4428, 3.8039, 2.9600, 3.8729, 1.5836, 2.8518], device='cuda:3'), covar=tensor([0.1586, 0.0865, 0.1337, 0.0221, 0.0146, 0.0401, 0.1867, 0.0947], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0207, 0.0206, 0.0221, 0.0211, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 17:05:17,533 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 17:05:24,974 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286732.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:05:42,086 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1361, 4.0052, 4.2077, 4.3180, 4.3744, 3.9982, 4.1698, 4.3891], device='cuda:3'), covar=tensor([0.1655, 0.1273, 0.1221, 0.0678, 0.0697, 0.1474, 0.2575, 0.0862], device='cuda:3'), in_proj_covar=tensor([0.0704, 0.0864, 0.0995, 0.0875, 0.0667, 0.0693, 0.0734, 0.0848], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:05:55,670 INFO [train.py:904] (3/8) Epoch 29, batch 2550, loss[loss=0.1685, simple_loss=0.249, pruned_loss=0.04397, over 16858.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2558, pruned_loss=0.03955, over 3312960.87 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:06:19,042 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:33,006 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286779.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:06:40,953 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:44,662 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.042e+02 2.418e+02 2.906e+02 6.403e+02, threshold=4.836e+02, percent-clipped=2.0 2023-05-02 17:07:07,733 INFO [train.py:904] (3/8) Epoch 29, batch 2600, loss[loss=0.1694, simple_loss=0.2566, pruned_loss=0.04107, over 16439.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2556, pruned_loss=0.03906, over 3318057.76 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:07:39,059 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:07:45,105 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:08:15,898 INFO [train.py:904] (3/8) Epoch 29, batch 2650, loss[loss=0.1594, simple_loss=0.2449, pruned_loss=0.03693, over 17218.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.256, pruned_loss=0.03884, over 3314174.68 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:08:49,617 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.56 vs. limit=5.0 2023-05-02 17:09:00,291 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 2700, loss[loss=0.1442, simple_loss=0.2336, pruned_loss=0.02739, over 16801.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2564, pruned_loss=0.03854, over 3321566.25 frames. ], batch size: 39, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:09:40,896 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0792, 5.1511, 5.5256, 5.5131, 5.5704, 5.2107, 5.1492, 4.9479], device='cuda:3'), covar=tensor([0.0357, 0.0560, 0.0373, 0.0422, 0.0442, 0.0401, 0.0938, 0.0470], device='cuda:3'), in_proj_covar=tensor([0.0451, 0.0510, 0.0493, 0.0451, 0.0539, 0.0516, 0.0594, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 17:10:17,471 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286944.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:26,948 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 2750, loss[loss=0.1809, simple_loss=0.2639, pruned_loss=0.04893, over 16748.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2569, pruned_loss=0.03817, over 3324296.35 frames. ], batch size: 89, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:17,993 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:35,307 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:40,741 INFO [train.py:904] (3/8) Epoch 29, batch 2800, loss[loss=0.1672, simple_loss=0.2628, pruned_loss=0.03576, over 17089.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2566, pruned_loss=0.03794, over 3317821.76 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:56,762 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8094, 4.3085, 4.2492, 3.1022, 3.5312, 4.2496, 3.8181, 2.5625], device='cuda:3'), covar=tensor([0.0469, 0.0089, 0.0069, 0.0377, 0.0183, 0.0117, 0.0121, 0.0481], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0091, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 17:12:12,974 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:19,146 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0528, 4.0113, 3.9501, 3.0708, 3.9370, 1.7285, 3.7203, 3.3656], device='cuda:3'), covar=tensor([0.0159, 0.0152, 0.0213, 0.0381, 0.0123, 0.3516, 0.0154, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0179, 0.0218, 0.0188, 0.0195, 0.0222, 0.0205, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:12:42,702 INFO [zipformer.py:625] (3/8) 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] (3/8) attn_weights_entropy = tensor([4.9976, 5.3256, 5.0884, 5.0781, 4.8829, 4.8106, 4.7341, 5.4156], device='cuda:3'), covar=tensor([0.1258, 0.0872, 0.0990, 0.0887, 0.0804, 0.0998, 0.1307, 0.0868], device='cuda:3'), in_proj_covar=tensor([0.0733, 0.0888, 0.0727, 0.0688, 0.0561, 0.0559, 0.0745, 0.0698], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:12:50,055 INFO [train.py:904] (3/8) Epoch 29, batch 2850, loss[loss=0.1688, simple_loss=0.2412, pruned_loss=0.04819, over 16893.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2559, pruned_loss=0.03794, over 3321814.37 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:13:39,986 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.226e+02 2.685e+02 3.371e+02 7.491e+02, threshold=5.370e+02, percent-clipped=5.0 2023-05-02 17:14:00,585 INFO [train.py:904] (3/8) Epoch 29, batch 2900, loss[loss=0.1757, simple_loss=0.249, pruned_loss=0.05126, over 16820.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2551, pruned_loss=0.03864, over 3315189.91 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:14:23,276 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9525, 4.9168, 4.7505, 4.2126, 4.8395, 1.9195, 4.5919, 4.5050], device='cuda:3'), covar=tensor([0.0138, 0.0135, 0.0239, 0.0404, 0.0139, 0.2949, 0.0169, 0.0265], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0179, 0.0217, 0.0189, 0.0195, 0.0222, 0.0205, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:14:31,456 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:15:10,985 INFO [train.py:904] (3/8) Epoch 29, batch 2950, loss[loss=0.1787, simple_loss=0.2613, pruned_loss=0.04801, over 16404.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2538, pruned_loss=0.03914, over 3313602.70 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:15:59,394 INFO [optim.py:368] (3/8) 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 17:16:12,650 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 17:16:20,171 INFO [train.py:904] (3/8) Epoch 29, batch 3000, loss[loss=0.1432, simple_loss=0.2341, pruned_loss=0.0261, over 17174.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2529, pruned_loss=0.03847, over 3326428.45 frames. ], batch size: 46, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:16:20,172 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 17:16:28,747 INFO [train.py:938] (3/8) Epoch 29, validation: loss=0.1336, simple_loss=0.2385, pruned_loss=0.01438, over 944034.00 frames. 2023-05-02 17:16:28,748 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 17:17:25,944 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:17:40,006 INFO [train.py:904] (3/8) Epoch 29, batch 3050, loss[loss=0.158, simple_loss=0.26, pruned_loss=0.02803, over 17147.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2531, pruned_loss=0.03913, over 3321789.96 frames. ], batch size: 47, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:18:11,135 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-05-02 17:18:29,494 INFO [optim.py:368] (3/8) 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 17:18:34,330 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287292.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:18:49,717 INFO [train.py:904] (3/8) Epoch 29, batch 3100, loss[loss=0.17, simple_loss=0.2567, pruned_loss=0.04167, over 16677.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2525, pruned_loss=0.03892, over 3322898.77 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:19:22,362 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 3150, loss[loss=0.1943, simple_loss=0.2604, pruned_loss=0.06413, over 16902.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2514, pruned_loss=0.03873, over 3325003.59 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:20:30,289 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:38,229 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7859, 3.5576, 3.9209, 2.1313, 4.0696, 4.0709, 3.1969, 3.0452], device='cuda:3'), covar=tensor([0.0765, 0.0289, 0.0222, 0.1226, 0.0116, 0.0213, 0.0464, 0.0484], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 17:20:49,323 INFO [optim.py:368] (3/8) 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 17:21:10,100 INFO [train.py:904] (3/8) Epoch 29, batch 3200, loss[loss=0.1632, simple_loss=0.2483, pruned_loss=0.0391, over 16458.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2511, pruned_loss=0.03802, over 3328383.96 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:21:20,061 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:21:40,358 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:18,322 INFO [train.py:904] (3/8) Epoch 29, batch 3250, loss[loss=0.1889, simple_loss=0.2627, pruned_loss=0.05758, over 16406.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2515, pruned_loss=0.03828, over 3326482.88 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:22:43,559 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:45,674 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287474.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:23:05,920 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.139e+02 2.558e+02 3.030e+02 4.692e+02, threshold=5.116e+02, percent-clipped=0.0 2023-05-02 17:23:12,612 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2701, 5.9094, 6.0923, 5.7035, 5.8938, 6.3913, 5.9393, 5.6645], device='cuda:3'), covar=tensor([0.0955, 0.1734, 0.2073, 0.2259, 0.2479, 0.0950, 0.1533, 0.2217], device='cuda:3'), in_proj_covar=tensor([0.0438, 0.0652, 0.0725, 0.0532, 0.0712, 0.0743, 0.0560, 0.0705], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 17:23:18,602 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8868, 3.7670, 4.2315, 2.1886, 4.3974, 4.4414, 3.2251, 3.3314], device='cuda:3'), covar=tensor([0.0789, 0.0262, 0.0220, 0.1189, 0.0100, 0.0234, 0.0475, 0.0458], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 17:23:26,856 INFO [train.py:904] (3/8) Epoch 29, batch 3300, loss[loss=0.1562, simple_loss=0.2541, pruned_loss=0.02918, over 17137.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2519, pruned_loss=0.03825, over 3320007.64 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:11,467 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:29,466 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-02 17:24:34,314 INFO [train.py:904] (3/8) Epoch 29, batch 3350, loss[loss=0.1553, simple_loss=0.2501, pruned_loss=0.0303, over 16707.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2528, pruned_loss=0.0383, over 3327962.75 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:46,540 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287562.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:55,110 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 17:25:00,595 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9463, 2.8615, 2.7818, 5.0727, 4.0203, 4.3530, 1.7501, 3.1845], device='cuda:3'), covar=tensor([0.1339, 0.0832, 0.1212, 0.0197, 0.0237, 0.0423, 0.1639, 0.0799], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0209, 0.0208, 0.0221, 0.0211, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 17:25:14,588 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287583.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:21,672 INFO [optim.py:368] (3/8) 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 17:25:33,245 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 3400, loss[loss=0.1483, simple_loss=0.2412, pruned_loss=0.02765, over 16561.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2523, pruned_loss=0.03797, over 3328620.39 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:25:45,221 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-05-02 17:26:10,417 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:16,676 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1963, 5.1460, 5.0282, 4.5105, 4.6864, 5.0972, 4.9293, 4.6510], device='cuda:3'), covar=tensor([0.0643, 0.0599, 0.0379, 0.0438, 0.1198, 0.0541, 0.0382, 0.0869], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0496, 0.0385, 0.0387, 0.0380, 0.0444, 0.0263, 0.0460], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 17:26:40,161 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287644.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:53,540 INFO [train.py:904] (3/8) Epoch 29, batch 3450, loss[loss=0.1605, simple_loss=0.2598, pruned_loss=0.03062, over 17275.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2516, pruned_loss=0.03709, over 3334644.34 frames. ], batch size: 52, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:27:43,291 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 3500, loss[loss=0.1615, simple_loss=0.2508, pruned_loss=0.03616, over 16828.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2507, pruned_loss=0.03671, over 3335251.67 frames. ], batch size: 90, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:28:17,577 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9779, 4.6808, 4.6219, 3.4664, 3.7969, 4.6061, 4.1238, 2.8080], device='cuda:3'), covar=tensor([0.0532, 0.0064, 0.0049, 0.0372, 0.0144, 0.0092, 0.0102, 0.0507], device='cuda:3'), in_proj_covar=tensor([0.0142, 0.0093, 0.0094, 0.0140, 0.0106, 0.0120, 0.0102, 0.0135], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 17:28:26,767 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:14,562 INFO [train.py:904] (3/8) Epoch 29, batch 3550, loss[loss=0.1513, simple_loss=0.2402, pruned_loss=0.03124, over 17213.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.25, pruned_loss=0.03648, over 3337644.91 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:29:24,345 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5309, 5.9264, 5.6915, 5.7516, 5.3887, 5.4763, 5.3054, 6.0468], device='cuda:3'), covar=tensor([0.1730, 0.1039, 0.1068, 0.0952, 0.0969, 0.0701, 0.1475, 0.0966], device='cuda:3'), in_proj_covar=tensor([0.0742, 0.0894, 0.0733, 0.0694, 0.0568, 0.0565, 0.0751, 0.0702], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:29:33,113 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:42,526 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:54,025 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287781.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:29:58,253 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1724, 5.1285, 5.0153, 4.5145, 4.7039, 5.0593, 4.9684, 4.6928], device='cuda:3'), covar=tensor([0.0642, 0.0633, 0.0347, 0.0418, 0.1045, 0.0569, 0.0384, 0.0804], device='cuda:3'), in_proj_covar=tensor([0.0330, 0.0499, 0.0387, 0.0389, 0.0383, 0.0447, 0.0265, 0.0462], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 17:30:07,186 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.118e+02 2.537e+02 3.010e+02 5.172e+02, threshold=5.074e+02, percent-clipped=1.0 2023-05-02 17:30:26,881 INFO [train.py:904] (3/8) Epoch 29, batch 3600, loss[loss=0.1652, simple_loss=0.2473, pruned_loss=0.04148, over 15510.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2487, pruned_loss=0.03611, over 3333032.15 frames. ], batch size: 191, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:31:10,903 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287834.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:31:40,142 INFO [train.py:904] (3/8) Epoch 29, batch 3650, loss[loss=0.1614, simple_loss=0.2314, pruned_loss=0.04571, over 16901.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2478, pruned_loss=0.03661, over 3308474.90 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:32:08,440 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 17:32:35,499 INFO [optim.py:368] (3/8) 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 17:32:38,200 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:32:38,288 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8791, 3.8255, 3.9543, 4.0382, 4.0867, 3.7023, 3.9657, 4.1391], device='cuda:3'), covar=tensor([0.1480, 0.1012, 0.1139, 0.0620, 0.0679, 0.1916, 0.2030, 0.0721], device='cuda:3'), in_proj_covar=tensor([0.0712, 0.0869, 0.1002, 0.0882, 0.0673, 0.0697, 0.0736, 0.0852], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:32:38,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0694, 4.0266, 3.9939, 3.3502, 3.9628, 1.9262, 3.7975, 3.4353], device='cuda:3'), covar=tensor([0.0179, 0.0153, 0.0222, 0.0319, 0.0114, 0.2961, 0.0136, 0.0329], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0182, 0.0221, 0.0192, 0.0198, 0.0224, 0.0209, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:32:57,081 INFO [train.py:904] (3/8) Epoch 29, batch 3700, loss[loss=0.18, simple_loss=0.2577, pruned_loss=0.05121, over 16797.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2467, pruned_loss=0.03814, over 3282045.75 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:33:18,825 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:33:31,826 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-05-02 17:33:48,765 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:34:09,992 INFO [train.py:904] (3/8) Epoch 29, batch 3750, loss[loss=0.164, simple_loss=0.2473, pruned_loss=0.04037, over 16480.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2465, pruned_loss=0.03967, over 3273550.43 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:35:05,542 INFO [optim.py:368] (3/8) 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 17:35:29,235 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288003.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:35:30,029 INFO [train.py:904] (3/8) Epoch 29, batch 3800, loss[loss=0.1788, simple_loss=0.2632, pruned_loss=0.04724, over 16539.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2476, pruned_loss=0.04076, over 3283101.11 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:40,375 INFO [train.py:904] (3/8) Epoch 29, batch 3850, loss[loss=0.1637, simple_loss=0.2409, pruned_loss=0.04331, over 16426.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2484, pruned_loss=0.04171, over 3271223.98 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:55,433 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288064.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:36:59,579 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288067.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:37:12,437 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288076.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:37:30,812 INFO [optim.py:368] (3/8) 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 17:37:50,150 INFO [train.py:904] (3/8) Epoch 29, batch 3900, loss[loss=0.1741, simple_loss=0.2558, pruned_loss=0.04625, over 16566.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2483, pruned_loss=0.04256, over 3279206.23 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:37:53,586 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288106.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:38:05,838 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:27,297 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:39:00,689 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 3950, loss[loss=0.1736, simple_loss=0.2426, pruned_loss=0.05225, over 16877.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2479, pruned_loss=0.04303, over 3287593.38 frames. ], batch size: 109, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:39:22,298 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288167.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:39:55,368 INFO [optim.py:368] (3/8) 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 17:39:58,700 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4000, loss[loss=0.1711, simple_loss=0.2632, pruned_loss=0.03945, over 16780.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.248, pruned_loss=0.04363, over 3289553.30 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:40:30,273 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288213.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:40:37,367 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:40:41,715 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2490, 5.3206, 5.6243, 5.5949, 5.6845, 5.2968, 5.2558, 4.9209], device='cuda:3'), covar=tensor([0.0315, 0.0435, 0.0361, 0.0410, 0.0432, 0.0407, 0.0901, 0.0532], device='cuda:3'), in_proj_covar=tensor([0.0452, 0.0512, 0.0494, 0.0454, 0.0542, 0.0520, 0.0598, 0.0419], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 17:41:07,448 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288239.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:08,413 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:18,874 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5498, 2.7557, 2.3456, 2.5623, 3.0348, 2.6854, 2.9636, 3.2510], device='cuda:3'), covar=tensor([0.0113, 0.0413, 0.0574, 0.0467, 0.0284, 0.0430, 0.0276, 0.0271], device='cuda:3'), in_proj_covar=tensor([0.0240, 0.0251, 0.0239, 0.0240, 0.0252, 0.0250, 0.0248, 0.0250], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:41:29,240 INFO [train.py:904] (3/8) Epoch 29, batch 4050, loss[loss=0.1783, simple_loss=0.2623, pruned_loss=0.04717, over 16636.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2489, pruned_loss=0.04285, over 3290370.61 frames. ], batch size: 76, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:41:32,677 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0699, 5.3441, 5.1360, 5.1891, 4.8686, 4.8377, 4.7498, 5.4740], device='cuda:3'), covar=tensor([0.1285, 0.0837, 0.1037, 0.0840, 0.0782, 0.0932, 0.1147, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0742, 0.0893, 0.0731, 0.0695, 0.0569, 0.0566, 0.0752, 0.0702], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:41:47,212 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:21,065 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.858e+02 2.084e+02 2.415e+02 5.689e+02, threshold=4.167e+02, percent-clipped=1.0 2023-05-02 17:42:42,061 INFO [train.py:904] (3/8) Epoch 29, batch 4100, loss[loss=0.189, simple_loss=0.268, pruned_loss=0.05501, over 11855.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2503, pruned_loss=0.0423, over 3287482.55 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:43:17,873 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:45,780 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:49,058 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8952, 2.2176, 2.3704, 3.4168, 2.1576, 2.4756, 2.3095, 2.3463], device='cuda:3'), covar=tensor([0.1596, 0.3518, 0.2962, 0.0718, 0.4266, 0.2433, 0.3573, 0.3505], device='cuda:3'), in_proj_covar=tensor([0.0427, 0.0482, 0.0391, 0.0342, 0.0448, 0.0554, 0.0454, 0.0565], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:43:59,219 INFO [train.py:904] (3/8) Epoch 29, batch 4150, loss[loss=0.1971, simple_loss=0.2812, pruned_loss=0.05653, over 17039.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2575, pruned_loss=0.04444, over 3254857.31 frames. ], batch size: 50, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:44:07,903 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:33,375 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:51,110 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:52,908 INFO [optim.py:368] (3/8) 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 17:45:02,567 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4200, loss[loss=0.2222, simple_loss=0.3024, pruned_loss=0.07101, over 11186.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2647, pruned_loss=0.04611, over 3216717.54 frames. ], batch size: 247, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:45:18,714 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:19,800 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6299, 4.6202, 4.3820, 3.6446, 4.5365, 1.6695, 4.2856, 3.8745], device='cuda:3'), covar=tensor([0.0108, 0.0104, 0.0213, 0.0356, 0.0092, 0.3290, 0.0120, 0.0344], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0181, 0.0221, 0.0192, 0.0198, 0.0224, 0.0209, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:45:45,765 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:53,531 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4250, loss[loss=0.1634, simple_loss=0.2681, pruned_loss=0.02932, over 16684.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2685, pruned_loss=0.04605, over 3207705.46 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:46:36,672 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:43,211 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288462.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:46:45,745 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2368, 3.4458, 3.7065, 2.3912, 3.2295, 2.4776, 3.5979, 3.7525], device='cuda:3'), covar=tensor([0.0267, 0.0831, 0.0571, 0.1934, 0.0801, 0.0968, 0.0618, 0.0882], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 17:47:05,781 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 29, batch 4300, loss[loss=0.204, simple_loss=0.2987, pruned_loss=0.05467, over 15380.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.27, pruned_loss=0.04546, over 3189284.81 frames. ], batch size: 191, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:47:51,123 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4350, loss[loss=0.194, simple_loss=0.2862, pruned_loss=0.05086, over 16714.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2725, pruned_loss=0.04617, over 3174322.22 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:49:53,411 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.193e+02 2.447e+02 2.913e+02 5.150e+02, threshold=4.894e+02, percent-clipped=1.0 2023-05-02 17:50:13,841 INFO [train.py:904] (3/8) Epoch 29, batch 4400, loss[loss=0.1871, simple_loss=0.2743, pruned_loss=0.04992, over 16609.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2749, pruned_loss=0.04748, over 3165035.41 frames. ], batch size: 35, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:50:22,326 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4450, loss[loss=0.1878, simple_loss=0.2807, pruned_loss=0.04743, over 17223.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2781, pruned_loss=0.04876, over 3183792.71 frames. ], batch size: 52, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:51:34,140 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:49,659 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288670.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:52:08,447 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:18,785 INFO [optim.py:368] (3/8) 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 17:52:34,491 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4500, loss[loss=0.2135, simple_loss=0.2939, pruned_loss=0.06652, over 15425.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2789, pruned_loss=0.04967, over 3189591.76 frames. ], batch size: 190, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:52:42,936 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:01,791 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0284, 3.3278, 3.4939, 2.1651, 3.0957, 2.1372, 3.4001, 3.5848], device='cuda:3'), covar=tensor([0.0223, 0.0832, 0.0553, 0.2193, 0.0826, 0.1122, 0.0607, 0.0888], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0173, 0.0172, 0.0158, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 17:53:49,130 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:51,481 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-02 17:53:51,862 INFO [train.py:904] (3/8) Epoch 29, batch 4550, loss[loss=0.2002, simple_loss=0.2917, pruned_loss=0.05436, over 16903.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2796, pruned_loss=0.05036, over 3209263.11 frames. ], batch size: 96, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:54:03,002 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288762.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:54:13,457 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 17:54:40,455 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 17:54:44,107 INFO [optim.py:368] (3/8) 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 17:54:53,700 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:54:58,821 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1850, 4.2781, 4.4737, 4.1274, 4.2748, 4.8206, 4.3274, 3.9695], device='cuda:3'), covar=tensor([0.1792, 0.2032, 0.2171, 0.2250, 0.2685, 0.1122, 0.1610, 0.2511], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0640, 0.0706, 0.0519, 0.0696, 0.0728, 0.0548, 0.0690], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 17:55:04,694 INFO [train.py:904] (3/8) Epoch 29, batch 4600, loss[loss=0.1671, simple_loss=0.2637, pruned_loss=0.03528, over 16486.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2807, pruned_loss=0.05078, over 3210404.86 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:55:11,566 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288808.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:55:14,422 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288810.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:55:14,678 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1288, 2.4413, 2.3391, 3.7201, 2.2098, 2.7092, 2.4923, 2.5334], device='cuda:3'), covar=tensor([0.1487, 0.3188, 0.3036, 0.0672, 0.4198, 0.2323, 0.3197, 0.3253], device='cuda:3'), in_proj_covar=tensor([0.0424, 0.0478, 0.0388, 0.0339, 0.0447, 0.0550, 0.0450, 0.0560], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 17:56:18,390 INFO [train.py:904] (3/8) Epoch 29, batch 4650, loss[loss=0.1931, simple_loss=0.2844, pruned_loss=0.05091, over 16724.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2801, pruned_loss=0.05128, over 3204790.65 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:56:21,002 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288856.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:56:24,852 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 17:57:29,743 INFO [train.py:904] (3/8) Epoch 29, batch 4700, loss[loss=0.1769, simple_loss=0.2632, pruned_loss=0.04531, over 16824.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2771, pruned_loss=0.04986, over 3204010.48 frames. ], batch size: 39, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:57:37,897 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 17:58:21,495 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 17:58:21,672 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 17:58:41,606 INFO [train.py:904] (3/8) Epoch 29, batch 4750, loss[loss=0.1825, simple_loss=0.2622, pruned_loss=0.05141, over 16738.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2733, pruned_loss=0.04785, over 3217618.69 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:58:57,735 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288965.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:59:23,643 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:35,275 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.780e+02 1.987e+02 2.347e+02 4.131e+02, threshold=3.973e+02, percent-clipped=1.0 2023-05-02 17:59:50,950 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:54,500 INFO [train.py:904] (3/8) Epoch 29, batch 4800, loss[loss=0.1909, simple_loss=0.2813, pruned_loss=0.05023, over 16661.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2697, pruned_loss=0.04593, over 3219143.62 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:00:37,053 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:00,454 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289046.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:01:05,208 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:08,943 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:12,125 INFO [train.py:904] (3/8) Epoch 29, batch 4850, loss[loss=0.182, simple_loss=0.2788, pruned_loss=0.04264, over 16853.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2702, pruned_loss=0.04506, over 3200940.53 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:08,677 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:02:28,897 INFO [train.py:904] (3/8) Epoch 29, batch 4900, loss[loss=0.1601, simple_loss=0.2532, pruned_loss=0.03356, over 16604.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2689, pruned_loss=0.04361, over 3185715.14 frames. ], batch size: 76, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:33,591 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289107.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:03:06,152 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-02 18:03:16,444 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289136.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:42,156 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 4950, loss[loss=0.1786, simple_loss=0.2767, pruned_loss=0.04024, over 16439.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2687, pruned_loss=0.04316, over 3197276.41 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:38,081 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.040e+02 2.385e+02 2.853e+02 5.286e+02, threshold=4.771e+02, percent-clipped=2.0 2023-05-02 18:04:40,828 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0328, 3.9861, 3.8937, 2.8679, 3.9183, 1.5762, 3.6677, 3.3399], device='cuda:3'), covar=tensor([0.0162, 0.0172, 0.0225, 0.0540, 0.0132, 0.3591, 0.0190, 0.0414], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0178, 0.0217, 0.0188, 0.0194, 0.0220, 0.0205, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:04:46,861 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289197.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:04:53,396 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 18:04:56,967 INFO [train.py:904] (3/8) Epoch 29, batch 5000, loss[loss=0.177, simple_loss=0.2733, pruned_loss=0.04033, over 16437.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2701, pruned_loss=0.04293, over 3201302.72 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:57,488 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9523, 1.9966, 2.4867, 2.8806, 2.8836, 3.5493, 1.9782, 3.4705], device='cuda:3'), covar=tensor([0.0291, 0.0594, 0.0430, 0.0393, 0.0366, 0.0149, 0.0735, 0.0135], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0195, 0.0212, 0.0169, 0.0206, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 18:06:09,972 INFO [train.py:904] (3/8) Epoch 29, batch 5050, loss[loss=0.1668, simple_loss=0.2666, pruned_loss=0.03349, over 16726.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2711, pruned_loss=0.04289, over 3210530.53 frames. ], batch size: 89, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:06:18,572 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6239, 2.4028, 2.2316, 3.8967, 2.2724, 3.6793, 1.5769, 2.6668], device='cuda:3'), covar=tensor([0.1545, 0.1031, 0.1598, 0.0186, 0.0204, 0.0485, 0.1911, 0.1043], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0181, 0.0200, 0.0204, 0.0206, 0.0217, 0.0210, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:06:27,028 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289265.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:06:30,875 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 18:07:03,093 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 1.941e+02 2.202e+02 2.540e+02 3.761e+02, threshold=4.404e+02, percent-clipped=0.0 2023-05-02 18:07:22,109 INFO [train.py:904] (3/8) Epoch 29, batch 5100, loss[loss=0.1926, simple_loss=0.2761, pruned_loss=0.05452, over 12392.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2696, pruned_loss=0.04242, over 3211768.71 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:07:34,854 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:08:31,808 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8141, 3.7168, 3.8667, 3.9787, 4.0773, 3.6966, 4.0381, 4.1243], device='cuda:3'), covar=tensor([0.1563, 0.1127, 0.1386, 0.0743, 0.0566, 0.1941, 0.0801, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0685, 0.0835, 0.0963, 0.0850, 0.0647, 0.0668, 0.0706, 0.0821], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:08:36,361 INFO [train.py:904] (3/8) Epoch 29, batch 5150, loss[loss=0.1795, simple_loss=0.2764, pruned_loss=0.04127, over 16860.00 frames. ], tot_loss[loss=0.176, simple_loss=0.269, pruned_loss=0.04153, over 3205281.15 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:09:29,038 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289402.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:09:47,654 INFO [train.py:904] (3/8) Epoch 29, batch 5200, loss[loss=0.1681, simple_loss=0.2419, pruned_loss=0.04722, over 16994.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2675, pruned_loss=0.04105, over 3218060.21 frames. ], batch size: 55, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:10:23,623 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5364, 2.6468, 2.1793, 2.4823, 3.0386, 2.6864, 3.0409, 3.2322], device='cuda:3'), covar=tensor([0.0142, 0.0548, 0.0705, 0.0527, 0.0346, 0.0455, 0.0295, 0.0304], device='cuda:3'), in_proj_covar=tensor([0.0235, 0.0247, 0.0236, 0.0237, 0.0248, 0.0247, 0.0243, 0.0248], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:10:59,573 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:11:00,510 INFO [train.py:904] (3/8) Epoch 29, batch 5250, loss[loss=0.157, simple_loss=0.2578, pruned_loss=0.02807, over 16858.00 frames. ], tot_loss[loss=0.173, simple_loss=0.265, pruned_loss=0.04054, over 3215827.98 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:11:55,944 INFO [optim.py:368] (3/8) 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 18:11:58,115 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:10,716 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:14,309 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289503.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:15,673 INFO [train.py:904] (3/8) Epoch 29, batch 5300, loss[loss=0.1784, simple_loss=0.2595, pruned_loss=0.04861, over 12130.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2618, pruned_loss=0.04, over 3191572.86 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:12:48,423 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7146, 3.4578, 4.0039, 1.9131, 4.2187, 4.1374, 3.0674, 3.0941], device='cuda:3'), covar=tensor([0.0787, 0.0318, 0.0198, 0.1337, 0.0065, 0.0148, 0.0425, 0.0498], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0131, 0.0130, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:13:26,898 INFO [train.py:904] (3/8) Epoch 29, batch 5350, loss[loss=0.1775, simple_loss=0.2667, pruned_loss=0.04417, over 17094.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2607, pruned_loss=0.03958, over 3199894.83 frames. ], batch size: 47, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:42,025 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:13:53,334 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3541, 5.6432, 5.3785, 5.4576, 5.1437, 5.1013, 5.0385, 5.7417], device='cuda:3'), covar=tensor([0.1198, 0.0745, 0.0944, 0.0820, 0.0784, 0.0755, 0.1177, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0727, 0.0874, 0.0718, 0.0676, 0.0557, 0.0554, 0.0731, 0.0687], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:14:21,403 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 5400, loss[loss=0.1782, simple_loss=0.271, pruned_loss=0.04277, over 12148.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2632, pruned_loss=0.04015, over 3197615.61 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:15:23,740 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 18:15:57,431 INFO [train.py:904] (3/8) Epoch 29, batch 5450, loss[loss=0.1914, simple_loss=0.2861, pruned_loss=0.04836, over 16713.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2661, pruned_loss=0.04148, over 3191911.14 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:16:38,809 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1516, 3.5411, 3.6393, 2.4116, 3.2506, 3.6508, 3.3512, 2.0667], device='cuda:3'), covar=tensor([0.0593, 0.0090, 0.0073, 0.0445, 0.0154, 0.0133, 0.0124, 0.0514], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0136, 0.0103, 0.0116, 0.0099, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 18:16:53,637 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.272e+02 2.915e+02 3.734e+02 7.781e+02, threshold=5.831e+02, percent-clipped=14.0 2023-05-02 18:16:57,256 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3625, 3.2239, 3.6894, 1.7846, 3.8331, 3.7837, 2.7874, 2.8563], device='cuda:3'), covar=tensor([0.0905, 0.0338, 0.0215, 0.1359, 0.0083, 0.0208, 0.0514, 0.0490], device='cuda:3'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0131, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:17:12,039 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289702.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:17:14,645 INFO [train.py:904] (3/8) Epoch 29, batch 5500, loss[loss=0.2621, simple_loss=0.3324, pruned_loss=0.09588, over 12009.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2731, pruned_loss=0.04559, over 3162766.56 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:17:26,803 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4099, 4.6729, 4.4787, 4.4990, 4.2106, 4.1751, 4.1845, 4.7280], device='cuda:3'), covar=tensor([0.1265, 0.0872, 0.1034, 0.0925, 0.0868, 0.1641, 0.1189, 0.0889], device='cuda:3'), in_proj_covar=tensor([0.0721, 0.0866, 0.0713, 0.0671, 0.0553, 0.0550, 0.0726, 0.0682], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:17:58,963 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1826, 3.2179, 1.9880, 3.4527, 2.4429, 3.4872, 2.1712, 2.6901], device='cuda:3'), covar=tensor([0.0358, 0.0435, 0.1738, 0.0238, 0.0909, 0.0644, 0.1566, 0.0749], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0183, 0.0198, 0.0174, 0.0181, 0.0222, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:18:26,456 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289750.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:18:31,707 INFO [train.py:904] (3/8) Epoch 29, batch 5550, loss[loss=0.2749, simple_loss=0.3398, pruned_loss=0.105, over 11392.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2801, pruned_loss=0.05052, over 3127192.29 frames. ], batch size: 249, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:19:30,745 INFO [optim.py:368] (3/8) 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 18:19:33,683 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 5600, loss[loss=0.2727, simple_loss=0.3339, pruned_loss=0.1057, over 11166.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2846, pruned_loss=0.05437, over 3089783.00 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:20:47,994 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5307, 4.4318, 4.5719, 4.7141, 4.8908, 4.4177, 4.8728, 4.9000], device='cuda:3'), covar=tensor([0.2136, 0.1349, 0.1685, 0.0809, 0.0608, 0.1110, 0.0680, 0.0810], device='cuda:3'), in_proj_covar=tensor([0.0688, 0.0838, 0.0966, 0.0853, 0.0650, 0.0672, 0.0708, 0.0825], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:20:54,481 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289840.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:21:15,639 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1077, 5.1374, 4.9881, 4.6280, 4.6377, 5.0348, 4.8905, 4.7772], device='cuda:3'), covar=tensor([0.0584, 0.0411, 0.0282, 0.0299, 0.0996, 0.0453, 0.0350, 0.0603], device='cuda:3'), in_proj_covar=tensor([0.0311, 0.0472, 0.0365, 0.0366, 0.0362, 0.0423, 0.0251, 0.0437], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:21:16,879 INFO [train.py:904] (3/8) Epoch 29, batch 5650, loss[loss=0.1799, simple_loss=0.2678, pruned_loss=0.04601, over 16732.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2887, pruned_loss=0.05726, over 3081020.64 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:21:25,653 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:22:15,729 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 5700, loss[loss=0.2136, simple_loss=0.2994, pruned_loss=0.06387, over 15371.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.29, pruned_loss=0.05881, over 3080603.15 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:23:01,368 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289920.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:23:55,706 INFO [train.py:904] (3/8) Epoch 29, batch 5750, loss[loss=0.2091, simple_loss=0.2977, pruned_loss=0.06022, over 16868.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2925, pruned_loss=0.06017, over 3055697.42 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:24:40,821 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289981.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:24:55,858 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4984, 4.5166, 4.8585, 4.8291, 4.8537, 4.5728, 4.4853, 4.5110], device='cuda:3'), covar=tensor([0.0442, 0.0847, 0.0497, 0.0465, 0.0501, 0.0552, 0.1124, 0.0537], device='cuda:3'), in_proj_covar=tensor([0.0441, 0.0500, 0.0483, 0.0442, 0.0529, 0.0506, 0.0587, 0.0409], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 18:24:56,847 INFO [optim.py:368] (3/8) 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 18:25:08,740 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289998.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:25:20,970 INFO [train.py:904] (3/8) Epoch 29, batch 5800, loss[loss=0.1821, simple_loss=0.2785, pruned_loss=0.04288, over 16878.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2928, pruned_loss=0.05951, over 3050679.06 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:25:34,397 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 18:25:47,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0567, 3.3315, 3.5176, 2.1546, 3.0586, 2.2403, 3.4730, 3.6436], device='cuda:3'), covar=tensor([0.0248, 0.0891, 0.0595, 0.2099, 0.0866, 0.1039, 0.0656, 0.0958], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 18:26:34,275 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:26:39,147 INFO [train.py:904] (3/8) Epoch 29, batch 5850, loss[loss=0.1806, simple_loss=0.2764, pruned_loss=0.04235, over 16361.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2903, pruned_loss=0.05788, over 3056348.41 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:48,272 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290059.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:27:39,837 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.628e+02 3.224e+02 3.685e+02 6.132e+02, threshold=6.447e+02, percent-clipped=0.0 2023-05-02 18:28:01,356 INFO [train.py:904] (3/8) Epoch 29, batch 5900, loss[loss=0.1733, simple_loss=0.2708, pruned_loss=0.03794, over 16754.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2894, pruned_loss=0.05672, over 3083829.51 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:28:15,102 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 5950, loss[loss=0.1858, simple_loss=0.2768, pruned_loss=0.04743, over 16422.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2901, pruned_loss=0.05578, over 3088532.24 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:29:30,176 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:30:18,749 INFO [optim.py:368] (3/8) 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 18:30:38,241 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2584, 2.9261, 2.6767, 2.3421, 2.2479, 2.2972, 2.9771, 2.8285], device='cuda:3'), covar=tensor([0.2587, 0.0672, 0.1746, 0.2559, 0.2497, 0.2192, 0.0541, 0.1485], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0330, 0.0308, 0.0279, 0.0308, 0.0354], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 18:30:40,227 INFO [train.py:904] (3/8) Epoch 29, batch 6000, loss[loss=0.1748, simple_loss=0.2736, pruned_loss=0.03805, over 16875.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2891, pruned_loss=0.05569, over 3079479.71 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:30:40,227 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 18:30:50,251 INFO [train.py:938] (3/8) 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] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 18:30:55,863 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:31:07,936 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8426, 1.4668, 1.7690, 1.6595, 1.8785, 1.9877, 1.6623, 1.9413], device='cuda:3'), covar=tensor([0.0298, 0.0461, 0.0250, 0.0371, 0.0308, 0.0249, 0.0491, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0198, 0.0197, 0.0186, 0.0193, 0.0210, 0.0167, 0.0203, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:31:54,820 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2688, 5.0555, 5.2476, 5.4587, 5.6632, 5.0338, 5.6211, 5.6650], device='cuda:3'), covar=tensor([0.2271, 0.1575, 0.2092, 0.0981, 0.0788, 0.1078, 0.0877, 0.0841], device='cuda:3'), in_proj_covar=tensor([0.0684, 0.0835, 0.0961, 0.0847, 0.0646, 0.0667, 0.0707, 0.0823], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:32:11,209 INFO [train.py:904] (3/8) Epoch 29, batch 6050, loss[loss=0.1917, simple_loss=0.2906, pruned_loss=0.04642, over 16897.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2878, pruned_loss=0.05532, over 3087780.00 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:32:15,054 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9276, 3.7719, 4.2074, 2.0909, 4.4280, 4.3735, 3.3021, 3.4129], device='cuda:3'), covar=tensor([0.0715, 0.0257, 0.0176, 0.1276, 0.0067, 0.0173, 0.0367, 0.0411], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0132, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:32:42,689 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290276.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:33:04,446 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.468e+02 2.905e+02 3.476e+02 8.072e+02, threshold=5.810e+02, percent-clipped=1.0 2023-05-02 18:33:29,592 INFO [train.py:904] (3/8) Epoch 29, batch 6100, loss[loss=0.1757, simple_loss=0.2724, pruned_loss=0.03952, over 16846.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2873, pruned_loss=0.05449, over 3091088.88 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:33:53,038 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:34:21,545 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 18:34:46,605 INFO [train.py:904] (3/8) Epoch 29, batch 6150, loss[loss=0.1812, simple_loss=0.2699, pruned_loss=0.04622, over 16950.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2863, pruned_loss=0.05465, over 3077248.78 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:34:47,113 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:05,035 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5345, 3.7242, 2.8356, 2.3349, 2.5302, 2.5166, 4.1012, 3.3622], device='cuda:3'), covar=tensor([0.3201, 0.0763, 0.1907, 0.2765, 0.2698, 0.2135, 0.0447, 0.1321], device='cuda:3'), in_proj_covar=tensor([0.0338, 0.0277, 0.0315, 0.0331, 0.0308, 0.0280, 0.0308, 0.0355], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 18:35:27,393 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290379.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:44,980 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 6200, loss[loss=0.1692, simple_loss=0.2616, pruned_loss=0.03844, over 16792.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2838, pruned_loss=0.0535, over 3097255.92 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:36:08,271 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:37:21,096 INFO [train.py:904] (3/8) Epoch 29, batch 6250, loss[loss=0.1802, simple_loss=0.2793, pruned_loss=0.04058, over 16800.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2831, pruned_loss=0.05313, over 3101430.65 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:37:41,294 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2388, 2.4366, 2.4873, 4.0626, 2.2455, 2.7580, 2.4825, 2.5598], device='cuda:3'), covar=tensor([0.1461, 0.3363, 0.2865, 0.0521, 0.3999, 0.2432, 0.3747, 0.2982], device='cuda:3'), in_proj_covar=tensor([0.0423, 0.0475, 0.0387, 0.0337, 0.0446, 0.0547, 0.0449, 0.0558], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:38:15,680 INFO [optim.py:368] (3/8) 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 18:38:34,373 INFO [train.py:904] (3/8) Epoch 29, batch 6300, loss[loss=0.2018, simple_loss=0.279, pruned_loss=0.0623, over 11407.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2825, pruned_loss=0.05279, over 3089642.75 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:38:38,268 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7430, 2.8965, 2.3515, 2.5943, 3.2413, 2.8916, 3.3174, 3.4343], device='cuda:3'), covar=tensor([0.0132, 0.0466, 0.0627, 0.0509, 0.0290, 0.0378, 0.0287, 0.0287], device='cuda:3'), in_proj_covar=tensor([0.0232, 0.0244, 0.0234, 0.0235, 0.0245, 0.0244, 0.0240, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:39:03,927 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5571, 3.3386, 3.7957, 1.9267, 3.9511, 3.9405, 2.9879, 2.9032], device='cuda:3'), covar=tensor([0.0850, 0.0318, 0.0214, 0.1252, 0.0092, 0.0218, 0.0472, 0.0540], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0133, 0.0130, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:39:52,799 INFO [train.py:904] (3/8) Epoch 29, batch 6350, loss[loss=0.1828, simple_loss=0.272, pruned_loss=0.04679, over 16380.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2827, pruned_loss=0.05353, over 3082439.50 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:40:28,254 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290576.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:40:50,035 INFO [optim.py:368] (3/8) 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 18:41:09,296 INFO [train.py:904] (3/8) Epoch 29, batch 6400, loss[loss=0.2116, simple_loss=0.2925, pruned_loss=0.06534, over 16958.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.283, pruned_loss=0.05439, over 3085724.96 frames. ], batch size: 109, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:41:17,196 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3411, 3.4820, 3.7976, 2.0963, 3.2110, 2.3706, 3.7069, 3.8397], device='cuda:3'), covar=tensor([0.0225, 0.0846, 0.0561, 0.2247, 0.0824, 0.1026, 0.0623, 0.0953], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 18:41:39,256 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290624.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 18:42:23,290 INFO [train.py:904] (3/8) Epoch 29, batch 6450, loss[loss=0.2148, simple_loss=0.2821, pruned_loss=0.07374, over 11649.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2831, pruned_loss=0.05421, over 3078081.04 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:42:23,721 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:42:38,778 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3822, 4.3673, 4.2767, 3.4203, 4.3220, 1.7463, 4.1240, 3.7920], device='cuda:3'), covar=tensor([0.0138, 0.0105, 0.0192, 0.0315, 0.0099, 0.3134, 0.0139, 0.0303], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0179, 0.0217, 0.0189, 0.0194, 0.0221, 0.0205, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:42:53,362 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 29, batch 6500, loss[loss=0.1967, simple_loss=0.2845, pruned_loss=0.05446, over 16832.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2815, pruned_loss=0.05367, over 3082923.30 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:43:44,250 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290706.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:44:05,239 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1563, 5.4897, 5.2134, 5.2712, 4.9737, 4.9301, 4.8622, 5.6080], device='cuda:3'), covar=tensor([0.1317, 0.0829, 0.0996, 0.0995, 0.0808, 0.0909, 0.1255, 0.0784], device='cuda:3'), in_proj_covar=tensor([0.0722, 0.0865, 0.0714, 0.0674, 0.0552, 0.0553, 0.0727, 0.0681], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 18:44:31,092 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 18:44:59,509 INFO [train.py:904] (3/8) Epoch 29, batch 6550, loss[loss=0.1909, simple_loss=0.2965, pruned_loss=0.04262, over 16247.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.285, pruned_loss=0.05464, over 3078560.65 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:44:59,830 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:45:58,155 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.498e+02 2.774e+02 3.264e+02 3.867e+02 7.263e+02, threshold=6.527e+02, percent-clipped=1.0 2023-05-02 18:46:19,466 INFO [train.py:904] (3/8) Epoch 29, batch 6600, loss[loss=0.2059, simple_loss=0.2977, pruned_loss=0.05705, over 16710.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.287, pruned_loss=0.05535, over 3078056.62 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:47:38,810 INFO [train.py:904] (3/8) Epoch 29, batch 6650, loss[loss=0.1854, simple_loss=0.2774, pruned_loss=0.04672, over 16540.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2868, pruned_loss=0.05553, over 3087940.12 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:48:35,967 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 6700, loss[loss=0.2325, simple_loss=0.3041, pruned_loss=0.08049, over 16356.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2862, pruned_loss=0.05599, over 3079424.06 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:12,978 INFO [train.py:904] (3/8) Epoch 29, batch 6750, loss[loss=0.2007, simple_loss=0.279, pruned_loss=0.06125, over 16253.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.285, pruned_loss=0.05583, over 3094617.93 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:43,278 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 18:51:28,424 INFO [train.py:904] (3/8) Epoch 29, batch 6800, loss[loss=0.233, simple_loss=0.3024, pruned_loss=0.08181, over 11277.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2847, pruned_loss=0.05588, over 3092548.33 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:51:57,164 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291022.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:52:03,261 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-02 18:52:46,127 INFO [train.py:904] (3/8) Epoch 29, batch 6850, loss[loss=0.2025, simple_loss=0.2923, pruned_loss=0.05636, over 15410.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2857, pruned_loss=0.05629, over 3101263.91 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:53:09,415 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291069.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:53:28,423 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8273, 3.9574, 2.6374, 4.4708, 3.0765, 4.3140, 2.5438, 3.0439], device='cuda:3'), covar=tensor([0.0290, 0.0345, 0.1507, 0.0260, 0.0760, 0.0642, 0.1673, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0181, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:53:42,653 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 6900, loss[loss=0.2069, simple_loss=0.2984, pruned_loss=0.05766, over 15334.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2873, pruned_loss=0.05537, over 3103633.63 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:54:44,526 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 6950, loss[loss=0.2447, simple_loss=0.3093, pruned_loss=0.09006, over 10732.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2891, pruned_loss=0.05708, over 3088375.79 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:18,587 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 7000, loss[loss=0.2049, simple_loss=0.2967, pruned_loss=0.05656, over 15390.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2892, pruned_loss=0.0563, over 3096695.05 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:41,522 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 18:57:04,045 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2600, 4.0140, 4.0307, 2.4780, 3.6305, 4.0758, 3.7140, 2.1279], device='cuda:3'), covar=tensor([0.0658, 0.0068, 0.0062, 0.0501, 0.0114, 0.0116, 0.0107, 0.0564], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 18:57:50,282 INFO [train.py:904] (3/8) Epoch 29, batch 7050, loss[loss=0.2071, simple_loss=0.2908, pruned_loss=0.06166, over 16600.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2907, pruned_loss=0.05654, over 3097510.11 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:57:55,830 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6217, 2.5442, 1.9653, 2.6471, 2.1472, 2.7586, 2.1619, 2.3264], device='cuda:3'), covar=tensor([0.0293, 0.0349, 0.1178, 0.0242, 0.0600, 0.0424, 0.1062, 0.0593], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0181, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 18:58:49,441 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.753e+02 3.384e+02 4.060e+02 1.222e+03, threshold=6.768e+02, percent-clipped=4.0 2023-05-02 18:59:07,325 INFO [train.py:904] (3/8) Epoch 29, batch 7100, loss[loss=0.1865, simple_loss=0.2763, pruned_loss=0.04837, over 16337.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2888, pruned_loss=0.05609, over 3102292.18 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:59:19,284 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5200, 3.6735, 2.6001, 2.2379, 2.4917, 2.2954, 3.8216, 3.2245], device='cuda:3'), covar=tensor([0.3421, 0.0744, 0.2355, 0.2994, 0.2823, 0.2552, 0.0610, 0.1601], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0276, 0.0315, 0.0329, 0.0307, 0.0279, 0.0306, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 19:00:25,886 INFO [train.py:904] (3/8) Epoch 29, batch 7150, loss[loss=0.1962, simple_loss=0.2897, pruned_loss=0.05132, over 16444.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2867, pruned_loss=0.05565, over 3122959.86 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:23,563 INFO [optim.py:368] (3/8) 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 19:01:41,833 INFO [train.py:904] (3/8) Epoch 29, batch 7200, loss[loss=0.1811, simple_loss=0.2802, pruned_loss=0.04097, over 16762.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2848, pruned_loss=0.05414, over 3105899.44 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:45,992 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7631, 4.8246, 4.6501, 4.3163, 4.3024, 4.7305, 4.5891, 4.4634], device='cuda:3'), covar=tensor([0.0635, 0.0531, 0.0332, 0.0337, 0.1032, 0.0485, 0.0395, 0.0648], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0470, 0.0363, 0.0363, 0.0360, 0.0418, 0.0249, 0.0435], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:02:14,451 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=291425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:03:00,075 INFO [train.py:904] (3/8) Epoch 29, batch 7250, loss[loss=0.2166, simple_loss=0.2909, pruned_loss=0.07116, over 11387.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2824, pruned_loss=0.05286, over 3111125.69 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:03:58,888 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 7300, loss[loss=0.2027, simple_loss=0.2951, pruned_loss=0.05509, over 16388.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.282, pruned_loss=0.05272, over 3113162.06 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:05:32,024 INFO [train.py:904] (3/8) Epoch 29, batch 7350, loss[loss=0.2311, simple_loss=0.2995, pruned_loss=0.0813, over 11213.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2835, pruned_loss=0.05404, over 3086004.46 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:11,606 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 19:06:32,139 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 7400, loss[loss=0.208, simple_loss=0.288, pruned_loss=0.064, over 11658.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2839, pruned_loss=0.05399, over 3092690.72 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:08:07,280 INFO [train.py:904] (3/8) Epoch 29, batch 7450, loss[loss=0.1844, simple_loss=0.2804, pruned_loss=0.0442, over 16870.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2856, pruned_loss=0.05553, over 3061091.17 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:09:05,186 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0546, 3.2893, 3.3925, 2.0967, 3.1451, 3.3968, 3.1852, 1.9891], device='cuda:3'), covar=tensor([0.0606, 0.0104, 0.0083, 0.0523, 0.0129, 0.0136, 0.0116, 0.0518], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0117, 0.0099, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 19:09:10,902 INFO [optim.py:368] (3/8) 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 19:09:28,108 INFO [train.py:904] (3/8) Epoch 29, batch 7500, loss[loss=0.182, simple_loss=0.2666, pruned_loss=0.04867, over 16761.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2863, pruned_loss=0.05539, over 3041598.52 frames. ], batch size: 57, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:10:01,614 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=291725.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:10:45,759 INFO [train.py:904] (3/8) Epoch 29, batch 7550, loss[loss=0.1633, simple_loss=0.2503, pruned_loss=0.03813, over 16614.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2851, pruned_loss=0.05581, over 3044285.70 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:11:15,343 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:19,686 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:26,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5097, 3.5678, 3.3623, 2.9807, 3.2205, 3.4686, 3.3269, 3.2998], device='cuda:3'), covar=tensor([0.0588, 0.0636, 0.0299, 0.0276, 0.0444, 0.0477, 0.1260, 0.0498], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0468, 0.0361, 0.0362, 0.0357, 0.0417, 0.0248, 0.0434], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:11:44,714 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 7600, loss[loss=0.1906, simple_loss=0.2814, pruned_loss=0.04988, over 16515.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2847, pruned_loss=0.05601, over 3044064.44 frames. ], batch size: 75, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:12:19,354 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2201, 1.6410, 1.9818, 2.1295, 2.2791, 2.3922, 1.7387, 2.3478], device='cuda:3'), covar=tensor([0.0266, 0.0565, 0.0325, 0.0390, 0.0361, 0.0247, 0.0642, 0.0180], device='cuda:3'), in_proj_covar=tensor([0.0196, 0.0197, 0.0186, 0.0191, 0.0209, 0.0166, 0.0204, 0.0167], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:12:52,647 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291837.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:13:18,970 INFO [train.py:904] (3/8) Epoch 29, batch 7650, loss[loss=0.2046, simple_loss=0.2924, pruned_loss=0.05841, over 16202.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2855, pruned_loss=0.05681, over 3033284.19 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:14:09,146 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 19:14:20,910 INFO [optim.py:368] (3/8) 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 19:14:36,035 INFO [train.py:904] (3/8) Epoch 29, batch 7700, loss[loss=0.1847, simple_loss=0.2873, pruned_loss=0.04102, over 16822.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2847, pruned_loss=0.05608, over 3060502.38 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:14:43,781 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3507, 3.9844, 3.9830, 2.5690, 3.6086, 4.0397, 3.6065, 2.1738], device='cuda:3'), covar=tensor([0.0604, 0.0075, 0.0071, 0.0477, 0.0134, 0.0138, 0.0122, 0.0567], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0117, 0.0099, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 19:15:53,491 INFO [train.py:904] (3/8) Epoch 29, batch 7750, loss[loss=0.1937, simple_loss=0.2804, pruned_loss=0.05347, over 16680.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2851, pruned_loss=0.05584, over 3072591.92 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:16:55,756 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 2.177e+02 2.797e+02 3.314e+02 3.963e+02 7.468e+02, threshold=6.627e+02, percent-clipped=1.0 2023-05-02 19:17:14,568 INFO [train.py:904] (3/8) Epoch 29, batch 7800, loss[loss=0.1944, simple_loss=0.2772, pruned_loss=0.05579, over 16544.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2858, pruned_loss=0.05626, over 3078943.62 frames. ], batch size: 68, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:18:30,016 INFO [train.py:904] (3/8) Epoch 29, batch 7850, loss[loss=0.2135, simple_loss=0.2995, pruned_loss=0.06378, over 15341.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2863, pruned_loss=0.05622, over 3071286.26 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:18:30,452 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:19:30,095 INFO [optim.py:368] (3/8) 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 19:19:43,456 INFO [train.py:904] (3/8) Epoch 29, batch 7900, loss[loss=0.2412, simple_loss=0.311, pruned_loss=0.08569, over 11463.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2855, pruned_loss=0.0561, over 3070650.28 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:19:59,946 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=292115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:07,748 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9332, 2.4708, 1.9706, 2.1865, 2.7308, 2.4083, 2.5550, 2.8737], device='cuda:3'), covar=tensor([0.0234, 0.0429, 0.0630, 0.0530, 0.0326, 0.0413, 0.0263, 0.0318], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0243, 0.0233, 0.0233, 0.0244, 0.0241, 0.0239, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:20:08,004 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 19:20:19,977 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8680, 4.9243, 4.7280, 4.3892, 4.4325, 4.8172, 4.6209, 4.5390], device='cuda:3'), covar=tensor([0.0602, 0.0547, 0.0326, 0.0315, 0.0892, 0.0556, 0.0419, 0.0671], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0470, 0.0362, 0.0363, 0.0359, 0.0418, 0.0250, 0.0435], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:20:26,271 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:26,406 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2472, 5.2807, 5.0989, 4.7422, 4.7708, 5.1855, 5.0558, 4.8543], device='cuda:3'), covar=tensor([0.0675, 0.0595, 0.0326, 0.0333, 0.1017, 0.0549, 0.0310, 0.0694], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0470, 0.0362, 0.0363, 0.0359, 0.0418, 0.0250, 0.0435], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:21:01,246 INFO [train.py:904] (3/8) Epoch 29, batch 7950, loss[loss=0.2272, simple_loss=0.3035, pruned_loss=0.07551, over 16277.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2863, pruned_loss=0.05683, over 3064401.45 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:21:07,704 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 19:21:10,860 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 19:22:03,656 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.577e+02 3.154e+02 3.832e+02 7.252e+02, threshold=6.308e+02, percent-clipped=1.0 2023-05-02 19:22:18,473 INFO [train.py:904] (3/8) Epoch 29, batch 8000, loss[loss=0.1795, simple_loss=0.2797, pruned_loss=0.03965, over 16900.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2871, pruned_loss=0.05724, over 3061025.86 frames. ], batch size: 90, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:22:21,347 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3021, 5.3741, 5.7295, 5.7144, 5.7577, 5.4055, 5.3402, 5.1124], device='cuda:3'), covar=tensor([0.0288, 0.0494, 0.0350, 0.0362, 0.0474, 0.0332, 0.0951, 0.0436], device='cuda:3'), in_proj_covar=tensor([0.0440, 0.0499, 0.0479, 0.0443, 0.0527, 0.0506, 0.0584, 0.0406], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 19:23:31,106 INFO [train.py:904] (3/8) Epoch 29, batch 8050, loss[loss=0.2467, simple_loss=0.3095, pruned_loss=0.09197, over 11842.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2872, pruned_loss=0.05717, over 3056220.74 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:24:32,479 INFO [optim.py:368] (3/8) 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 19:24:46,378 INFO [train.py:904] (3/8) Epoch 29, batch 8100, loss[loss=0.203, simple_loss=0.2927, pruned_loss=0.05667, over 16416.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2869, pruned_loss=0.05656, over 3049743.80 frames. ], batch size: 68, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:25:02,627 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 19:25:03,332 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8426, 5.1810, 5.3993, 5.1498, 5.2092, 5.7369, 5.2160, 4.9402], device='cuda:3'), covar=tensor([0.1122, 0.1818, 0.2372, 0.1869, 0.2202, 0.0925, 0.1631, 0.2381], device='cuda:3'), in_proj_covar=tensor([0.0430, 0.0639, 0.0708, 0.0518, 0.0691, 0.0730, 0.0549, 0.0693], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 19:25:03,492 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2622, 2.4600, 1.7674, 2.0115, 2.7908, 2.3947, 2.8669, 3.0178], device='cuda:3'), covar=tensor([0.0206, 0.0597, 0.0864, 0.0729, 0.0356, 0.0577, 0.0346, 0.0352], device='cuda:3'), in_proj_covar=tensor([0.0230, 0.0243, 0.0233, 0.0233, 0.0244, 0.0241, 0.0239, 0.0242], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:25:09,316 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2189, 3.3161, 3.5739, 2.0901, 3.1492, 2.3487, 3.5769, 3.6990], device='cuda:3'), covar=tensor([0.0262, 0.0861, 0.0632, 0.2197, 0.0837, 0.1020, 0.0584, 0.0938], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0171, 0.0171, 0.0157, 0.0148, 0.0134, 0.0145, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 19:26:01,419 INFO [train.py:904] (3/8) Epoch 29, batch 8150, loss[loss=0.1707, simple_loss=0.2597, pruned_loss=0.04083, over 16866.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2837, pruned_loss=0.05498, over 3077879.87 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:01,238 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 8200, loss[loss=0.2086, simple_loss=0.2778, pruned_loss=0.0697, over 11301.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2818, pruned_loss=0.05465, over 3078937.38 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:25,887 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292410.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:27:41,480 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4209, 3.5780, 2.2293, 3.8161, 2.7091, 3.8426, 2.4009, 2.9042], device='cuda:3'), covar=tensor([0.0325, 0.0374, 0.1620, 0.0395, 0.0886, 0.0631, 0.1521, 0.0731], device='cuda:3'), in_proj_covar=tensor([0.0176, 0.0179, 0.0195, 0.0173, 0.0179, 0.0220, 0.0203, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:28:00,113 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292432.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:34,796 INFO [train.py:904] (3/8) Epoch 29, batch 8250, loss[loss=0.1843, simple_loss=0.2696, pruned_loss=0.04956, over 12070.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2806, pruned_loss=0.05243, over 3049357.50 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:29:17,989 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:29:41,435 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.407e+02 2.184e+02 2.675e+02 3.477e+02 6.200e+02, threshold=5.351e+02, percent-clipped=0.0 2023-05-02 19:29:55,807 INFO [train.py:904] (3/8) Epoch 29, batch 8300, loss[loss=0.1753, simple_loss=0.2746, pruned_loss=0.03804, over 15223.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2777, pruned_loss=0.04937, over 3051169.29 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:31:15,559 INFO [train.py:904] (3/8) Epoch 29, batch 8350, loss[loss=0.1801, simple_loss=0.2635, pruned_loss=0.04837, over 12139.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2766, pruned_loss=0.04756, over 3028557.98 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:32:20,504 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.032e+02 2.659e+02 3.191e+02 5.260e+02, threshold=5.318e+02, percent-clipped=0.0 2023-05-02 19:32:36,089 INFO [train.py:904] (3/8) Epoch 29, batch 8400, loss[loss=0.1729, simple_loss=0.2557, pruned_loss=0.04501, over 12350.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2742, pruned_loss=0.0455, over 3034257.71 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:33:56,324 INFO [train.py:904] (3/8) Epoch 29, batch 8450, loss[loss=0.1669, simple_loss=0.2714, pruned_loss=0.03123, over 16881.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2727, pruned_loss=0.04413, over 3026947.68 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:34:05,446 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4632, 3.3780, 3.5001, 3.5771, 3.6246, 3.3471, 3.5834, 3.6784], device='cuda:3'), covar=tensor([0.1433, 0.1016, 0.1065, 0.0707, 0.0659, 0.2394, 0.0982, 0.0923], device='cuda:3'), in_proj_covar=tensor([0.0670, 0.0817, 0.0941, 0.0832, 0.0635, 0.0654, 0.0695, 0.0812], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:35:03,480 INFO [optim.py:368] (3/8) 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 19:35:19,252 INFO [train.py:904] (3/8) Epoch 29, batch 8500, loss[loss=0.1545, simple_loss=0.2442, pruned_loss=0.03241, over 16236.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2691, pruned_loss=0.04223, over 3023001.62 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:35:28,844 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292710.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:35:42,658 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2094, 4.2708, 4.5521, 4.5242, 4.5282, 4.3041, 4.2907, 4.2785], device='cuda:3'), covar=tensor([0.0495, 0.0946, 0.0588, 0.0581, 0.0591, 0.0640, 0.1136, 0.0624], device='cuda:3'), in_proj_covar=tensor([0.0436, 0.0496, 0.0476, 0.0440, 0.0521, 0.0502, 0.0579, 0.0403], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 19:35:50,919 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3031, 3.5962, 3.5528, 2.5280, 3.2212, 3.6365, 3.4257, 2.0666], device='cuda:3'), covar=tensor([0.0532, 0.0078, 0.0077, 0.0385, 0.0141, 0.0112, 0.0097, 0.0564], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0089, 0.0090, 0.0133, 0.0101, 0.0114, 0.0097, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 19:35:50,984 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7472, 2.6343, 2.4327, 4.0009, 2.5224, 4.0016, 1.5501, 2.9633], device='cuda:3'), covar=tensor([0.1420, 0.0817, 0.1227, 0.0157, 0.0132, 0.0344, 0.1783, 0.0740], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0182, 0.0201, 0.0204, 0.0207, 0.0218, 0.0211, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:36:08,102 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8794, 5.1467, 5.2773, 5.0266, 5.0478, 5.6185, 5.1024, 4.8019], device='cuda:3'), covar=tensor([0.0984, 0.1793, 0.2055, 0.1987, 0.2220, 0.0944, 0.1646, 0.2499], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0627, 0.0695, 0.0508, 0.0679, 0.0717, 0.0541, 0.0681], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 19:36:42,995 INFO [train.py:904] (3/8) Epoch 29, batch 8550, loss[loss=0.1883, simple_loss=0.296, pruned_loss=0.04027, over 16345.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2673, pruned_loss=0.04144, over 3008682.02 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:36:52,468 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292758.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:38:04,312 INFO [optim.py:368] (3/8) 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 19:38:21,505 INFO [train.py:904] (3/8) Epoch 29, batch 8600, loss[loss=0.1546, simple_loss=0.2569, pruned_loss=0.0262, over 16914.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2665, pruned_loss=0.04014, over 2994503.62 frames. ], batch size: 90, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:38:59,583 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 19:39:03,815 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8528, 2.7080, 2.6420, 4.1712, 2.2736, 4.0293, 1.6776, 3.0930], device='cuda:3'), covar=tensor([0.1381, 0.0845, 0.1112, 0.0134, 0.0118, 0.0364, 0.1694, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0182, 0.0201, 0.0203, 0.0206, 0.0218, 0.0211, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:39:21,456 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1256, 3.3599, 3.8304, 2.2471, 3.1148, 2.3766, 3.5546, 3.4571], device='cuda:3'), covar=tensor([0.0229, 0.0881, 0.0431, 0.2025, 0.0792, 0.1010, 0.0635, 0.1051], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0168, 0.0168, 0.0155, 0.0146, 0.0132, 0.0144, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:39:41,631 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4281, 3.3526, 3.4734, 3.5324, 3.5855, 3.3257, 3.5437, 3.6254], device='cuda:3'), covar=tensor([0.1287, 0.0912, 0.0970, 0.0627, 0.0578, 0.2201, 0.0920, 0.0800], device='cuda:3'), in_proj_covar=tensor([0.0667, 0.0815, 0.0937, 0.0831, 0.0633, 0.0653, 0.0692, 0.0810], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:39:53,062 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0341, 3.1935, 3.7407, 2.0353, 3.0497, 2.2180, 3.4363, 3.3077], device='cuda:3'), covar=tensor([0.0261, 0.1006, 0.0505, 0.2283, 0.0840, 0.1103, 0.0672, 0.1143], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0168, 0.0168, 0.0155, 0.0146, 0.0132, 0.0144, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:40:01,866 INFO [train.py:904] (3/8) Epoch 29, batch 8650, loss[loss=0.1654, simple_loss=0.258, pruned_loss=0.03645, over 11896.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2649, pruned_loss=0.03825, over 3017784.91 frames. ], batch size: 250, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:40:33,391 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2475, 4.0859, 4.3402, 4.4621, 4.5735, 4.1192, 4.5412, 4.5831], device='cuda:3'), covar=tensor([0.1783, 0.1194, 0.1440, 0.0704, 0.0591, 0.1265, 0.0698, 0.0748], device='cuda:3'), in_proj_covar=tensor([0.0666, 0.0813, 0.0936, 0.0829, 0.0632, 0.0652, 0.0691, 0.0809], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:41:31,353 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 8700, loss[loss=0.1939, simple_loss=0.2844, pruned_loss=0.05171, over 16716.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2626, pruned_loss=0.03702, over 3029477.76 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:43:26,843 INFO [train.py:904] (3/8) Epoch 29, batch 8750, loss[loss=0.1466, simple_loss=0.2388, pruned_loss=0.02716, over 12247.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2622, pruned_loss=0.03657, over 3031271.56 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:44:01,945 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:44:24,404 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 19:45:00,945 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 2.165e+02 2.592e+02 3.174e+02 6.684e+02, threshold=5.184e+02, percent-clipped=4.0 2023-05-02 19:45:20,129 INFO [train.py:904] (3/8) Epoch 29, batch 8800, loss[loss=0.1776, simple_loss=0.2733, pruned_loss=0.04101, over 16887.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2605, pruned_loss=0.03523, over 3051850.46 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:45:33,660 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8642, 2.7772, 2.5988, 1.9704, 2.5267, 2.7951, 2.6873, 1.9004], device='cuda:3'), covar=tensor([0.0465, 0.0102, 0.0091, 0.0382, 0.0169, 0.0129, 0.0118, 0.0478], device='cuda:3'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0132, 0.0101, 0.0113, 0.0097, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-02 19:46:12,246 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:46:28,909 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0369, 4.3010, 4.1750, 4.1959, 3.8491, 3.8868, 3.9237, 4.3045], device='cuda:3'), covar=tensor([0.1026, 0.0863, 0.0872, 0.0695, 0.0705, 0.1713, 0.0964, 0.0949], device='cuda:3'), in_proj_covar=tensor([0.0710, 0.0854, 0.0702, 0.0662, 0.0544, 0.0544, 0.0714, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:46:48,691 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9761, 3.7937, 3.8338, 4.0957, 4.1823, 3.8317, 4.1891, 4.2140], device='cuda:3'), covar=tensor([0.1706, 0.1294, 0.1966, 0.0990, 0.0808, 0.2200, 0.0997, 0.0990], device='cuda:3'), in_proj_covar=tensor([0.0664, 0.0811, 0.0933, 0.0828, 0.0630, 0.0651, 0.0690, 0.0806], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 19:47:05,839 INFO [train.py:904] (3/8) Epoch 29, batch 8850, loss[loss=0.1678, simple_loss=0.2779, pruned_loss=0.02885, over 16589.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2639, pruned_loss=0.03466, over 3051302.81 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:47:30,865 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 19:48:05,753 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:48:24,684 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 19:48:35,604 INFO [optim.py:368] (3/8) 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 19:48:55,721 INFO [train.py:904] (3/8) Epoch 29, batch 8900, loss[loss=0.1508, simple_loss=0.2541, pruned_loss=0.02377, over 16869.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2635, pruned_loss=0.03376, over 3057396.50 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:49:23,409 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293117.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:49:59,990 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7087, 3.2741, 3.5726, 1.8848, 3.7302, 3.7608, 3.0833, 2.9553], device='cuda:3'), covar=tensor([0.0608, 0.0259, 0.0187, 0.1171, 0.0076, 0.0159, 0.0369, 0.0423], device='cuda:3'), in_proj_covar=tensor([0.0145, 0.0108, 0.0099, 0.0135, 0.0085, 0.0127, 0.0126, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 19:50:34,625 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:35,146 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-02 19:51:01,957 INFO [train.py:904] (3/8) Epoch 29, batch 8950, loss[loss=0.1561, simple_loss=0.2497, pruned_loss=0.03125, over 16366.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2638, pruned_loss=0.03421, over 3081758.87 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:51:52,484 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 29, batch 9000, loss[loss=0.1521, simple_loss=0.2479, pruned_loss=0.02816, over 15177.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2611, pruned_loss=0.03336, over 3086540.07 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:52:53,145 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 19:53:02,752 INFO [train.py:938] (3/8) Epoch 29, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01987, over 944034.00 frames. 2023-05-02 19:53:02,752 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 19:53:49,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8991, 3.1522, 3.5177, 2.0675, 2.9987, 2.2591, 3.3897, 3.3087], device='cuda:3'), covar=tensor([0.0250, 0.0909, 0.0489, 0.2167, 0.0779, 0.0994, 0.0601, 0.1065], device='cuda:3'), in_proj_covar=tensor([0.0158, 0.0167, 0.0168, 0.0155, 0.0145, 0.0131, 0.0143, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:54:47,962 INFO [train.py:904] (3/8) Epoch 29, batch 9050, loss[loss=0.151, simple_loss=0.241, pruned_loss=0.03051, over 16554.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2619, pruned_loss=0.03378, over 3095300.85 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:56:15,025 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.113e+02 2.508e+02 2.955e+02 4.534e+02, threshold=5.015e+02, percent-clipped=0.0 2023-05-02 19:56:16,671 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9276, 2.4479, 2.3831, 3.5653, 1.8280, 3.6196, 1.6546, 2.9583], device='cuda:3'), covar=tensor([0.1333, 0.0795, 0.1273, 0.0179, 0.0101, 0.0358, 0.1690, 0.0725], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0202, 0.0205, 0.0217, 0.0211, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 19:56:34,919 INFO [train.py:904] (3/8) Epoch 29, batch 9100, loss[loss=0.1617, simple_loss=0.2629, pruned_loss=0.03022, over 15418.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2617, pruned_loss=0.03449, over 3080879.07 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:57:17,188 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:34,110 INFO [train.py:904] (3/8) Epoch 29, batch 9150, loss[loss=0.1527, simple_loss=0.2425, pruned_loss=0.03149, over 12058.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2625, pruned_loss=0.03467, over 3082134.01 frames. ], batch size: 250, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:58:51,914 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293361.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:56,735 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:59:43,566 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 20:00:04,472 INFO [optim.py:368] (3/8) 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 20:00:20,590 INFO [train.py:904] (3/8) Epoch 29, batch 9200, loss[loss=0.157, simple_loss=0.2491, pruned_loss=0.03245, over 12429.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2578, pruned_loss=0.03368, over 3064370.72 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:00:55,055 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293422.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:00,762 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:13,329 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1664, 3.2055, 1.9324, 3.5084, 2.4541, 3.4825, 2.1027, 2.6056], device='cuda:3'), covar=tensor([0.0347, 0.0403, 0.1710, 0.0274, 0.0888, 0.0580, 0.1634, 0.0832], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0175, 0.0191, 0.0167, 0.0176, 0.0213, 0.0200, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:01:23,543 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:58,342 INFO [train.py:904] (3/8) Epoch 29, batch 9250, loss[loss=0.1809, simple_loss=0.2767, pruned_loss=0.04253, over 16703.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2576, pruned_loss=0.03385, over 3062034.57 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:02:33,894 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9779, 4.1162, 3.8870, 3.6399, 3.5748, 4.0340, 3.6560, 3.7364], device='cuda:3'), covar=tensor([0.0679, 0.0730, 0.0419, 0.0380, 0.0814, 0.0532, 0.1230, 0.0669], device='cuda:3'), in_proj_covar=tensor([0.0302, 0.0458, 0.0354, 0.0354, 0.0349, 0.0407, 0.0244, 0.0422], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:02:39,018 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:03:28,128 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-02 20:03:28,486 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 9300, loss[loss=0.1417, simple_loss=0.2386, pruned_loss=0.02244, over 16363.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2563, pruned_loss=0.03332, over 3060818.20 frames. ], batch size: 166, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:04:37,732 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7645, 2.5197, 2.3595, 3.9298, 2.1778, 3.8608, 1.5330, 2.9380], device='cuda:3'), covar=tensor([0.1457, 0.0941, 0.1351, 0.0207, 0.0137, 0.0397, 0.1938, 0.0765], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0200, 0.0202, 0.0214, 0.0209, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:05:33,313 INFO [train.py:904] (3/8) Epoch 29, batch 9350, loss[loss=0.1749, simple_loss=0.266, pruned_loss=0.04194, over 16819.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2567, pruned_loss=0.03327, over 3083880.27 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:06:40,903 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7559, 4.7294, 4.5062, 3.8043, 4.6166, 1.6519, 4.3953, 4.2597], device='cuda:3'), covar=tensor([0.0100, 0.0089, 0.0211, 0.0325, 0.0107, 0.3028, 0.0131, 0.0254], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0171, 0.0208, 0.0180, 0.0186, 0.0215, 0.0197, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:06:43,163 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4532, 2.0652, 1.7909, 1.6867, 2.2924, 1.8973, 1.8579, 2.3427], device='cuda:3'), covar=tensor([0.0215, 0.0491, 0.0621, 0.0581, 0.0344, 0.0462, 0.0208, 0.0335], device='cuda:3'), in_proj_covar=tensor([0.0222, 0.0238, 0.0227, 0.0227, 0.0237, 0.0235, 0.0231, 0.0234], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:06:53,138 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 20:06:56,970 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 9400, loss[loss=0.1621, simple_loss=0.2579, pruned_loss=0.03314, over 15242.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2567, pruned_loss=0.03323, over 3067808.55 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:07:53,837 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:08:55,052 INFO [train.py:904] (3/8) Epoch 29, batch 9450, loss[loss=0.154, simple_loss=0.2561, pruned_loss=0.0259, over 16887.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2585, pruned_loss=0.0334, over 3076826.08 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:09:29,417 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:10:18,287 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 29, batch 9500, loss[loss=0.1579, simple_loss=0.2537, pruned_loss=0.03105, over 16367.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2579, pruned_loss=0.03341, over 3082784.36 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:10:51,374 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-05-02 20:11:02,445 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:08,937 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:14,931 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8098, 2.3577, 2.3188, 3.6181, 2.0249, 3.6319, 1.5291, 2.9575], device='cuda:3'), covar=tensor([0.1414, 0.0910, 0.1310, 0.0171, 0.0116, 0.0369, 0.1810, 0.0735], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0199, 0.0202, 0.0214, 0.0209, 0.0197], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:11:23,487 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 20:11:41,932 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:12:17,574 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6151, 4.5736, 4.4228, 3.7781, 4.4981, 1.7250, 4.2570, 4.1325], device='cuda:3'), covar=tensor([0.0094, 0.0100, 0.0213, 0.0279, 0.0107, 0.2840, 0.0137, 0.0271], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0179, 0.0185, 0.0214, 0.0196, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:12:19,483 INFO [train.py:904] (3/8) Epoch 29, batch 9550, loss[loss=0.1842, simple_loss=0.2823, pruned_loss=0.04306, over 16304.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2576, pruned_loss=0.03375, over 3071736.57 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:12:59,123 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:25,492 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:43,637 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 2.273e+02 2.766e+02 3.522e+02 6.419e+02, threshold=5.532e+02, percent-clipped=5.0 2023-05-02 20:13:59,133 INFO [train.py:904] (3/8) Epoch 29, batch 9600, loss[loss=0.178, simple_loss=0.2757, pruned_loss=0.04019, over 15297.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2583, pruned_loss=0.03432, over 3052464.13 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:14:32,486 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:14:50,172 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7312, 5.0585, 4.8756, 4.8482, 4.5573, 4.5204, 4.4539, 5.1432], device='cuda:3'), covar=tensor([0.1361, 0.1044, 0.0962, 0.0891, 0.0875, 0.1237, 0.1265, 0.0893], device='cuda:3'), in_proj_covar=tensor([0.0702, 0.0846, 0.0694, 0.0657, 0.0538, 0.0540, 0.0708, 0.0663], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:15:45,740 INFO [train.py:904] (3/8) Epoch 29, batch 9650, loss[loss=0.1739, simple_loss=0.2696, pruned_loss=0.03904, over 16169.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2604, pruned_loss=0.03463, over 3045616.51 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:17:16,474 INFO [optim.py:368] (3/8) 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 20:17:35,790 INFO [train.py:904] (3/8) Epoch 29, batch 9700, loss[loss=0.149, simple_loss=0.2442, pruned_loss=0.02686, over 17022.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2589, pruned_loss=0.0342, over 3058168.45 frames. ], batch size: 55, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:19:17,221 INFO [train.py:904] (3/8) Epoch 29, batch 9750, loss[loss=0.1683, simple_loss=0.2638, pruned_loss=0.03646, over 16897.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2582, pruned_loss=0.03432, over 3061125.87 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:20:08,090 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-02 20:20:24,597 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 29, batch 9800, loss[loss=0.1735, simple_loss=0.2715, pruned_loss=0.03776, over 16885.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2585, pruned_loss=0.0338, over 3057392.33 frames. ], batch size: 116, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:20:59,711 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0802, 4.0380, 3.9406, 3.1779, 3.9930, 1.8302, 3.8203, 3.4516], device='cuda:3'), covar=tensor([0.0106, 0.0109, 0.0196, 0.0262, 0.0105, 0.3166, 0.0129, 0.0348], device='cuda:3'), in_proj_covar=tensor([0.0177, 0.0170, 0.0208, 0.0179, 0.0185, 0.0214, 0.0196, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:21:22,946 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:27,426 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294020.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:40,426 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 20:21:45,313 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1249, 2.5880, 2.6124, 1.9190, 2.7847, 2.8626, 2.5100, 2.5152], device='cuda:3'), covar=tensor([0.0688, 0.0286, 0.0259, 0.1030, 0.0129, 0.0240, 0.0449, 0.0444], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0108, 0.0098, 0.0135, 0.0084, 0.0127, 0.0126, 0.0126], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 20:22:29,330 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:22:39,769 INFO [train.py:904] (3/8) Epoch 29, batch 9850, loss[loss=0.1593, simple_loss=0.2592, pruned_loss=0.02968, over 16672.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2593, pruned_loss=0.03333, over 3058720.86 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 16.0 2023-05-02 20:23:02,392 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) 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] (3/8) Epoch 29, batch 9900, loss[loss=0.1589, simple_loss=0.2465, pruned_loss=0.03561, over 12672.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2598, pruned_loss=0.0332, over 3071934.03 frames. ], batch size: 249, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:26:11,495 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4236, 4.4853, 4.3143, 3.9703, 4.0238, 4.3916, 4.1504, 4.0793], device='cuda:3'), covar=tensor([0.0573, 0.0725, 0.0380, 0.0348, 0.0940, 0.0564, 0.0718, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0301, 0.0454, 0.0352, 0.0352, 0.0347, 0.0404, 0.0243, 0.0420], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:26:27,773 INFO [train.py:904] (3/8) Epoch 29, batch 9950, loss[loss=0.1693, simple_loss=0.2608, pruned_loss=0.03893, over 12555.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2623, pruned_loss=0.03371, over 3084287.12 frames. ], batch size: 250, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:27:06,737 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:28:08,137 INFO [optim.py:368] (3/8) 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 20:28:14,607 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0964, 3.1354, 1.9631, 3.3047, 2.3691, 3.3533, 2.1809, 2.6013], device='cuda:3'), covar=tensor([0.0330, 0.0378, 0.1607, 0.0293, 0.0821, 0.0504, 0.1531, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0171, 0.0174, 0.0189, 0.0165, 0.0174, 0.0211, 0.0199, 0.0177], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:28:27,670 INFO [train.py:904] (3/8) Epoch 29, batch 10000, loss[loss=0.1684, simple_loss=0.2757, pruned_loss=0.03057, over 15524.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2605, pruned_loss=0.03292, over 3090363.34 frames. ], batch size: 192, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:29:21,488 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 29, batch 10050, loss[loss=0.1685, simple_loss=0.2619, pruned_loss=0.03757, over 12329.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2606, pruned_loss=0.03297, over 3072967.27 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:31:27,887 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.082e+02 2.498e+02 3.055e+02 5.920e+02, threshold=4.997e+02, percent-clipped=2.0 2023-05-02 20:31:41,195 INFO [train.py:904] (3/8) Epoch 29, batch 10100, loss[loss=0.1403, simple_loss=0.2332, pruned_loss=0.0237, over 16894.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2608, pruned_loss=0.03299, over 3075631.14 frames. ], batch size: 96, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:32:42,406 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:32:46,796 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294343.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:32:51,058 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4414, 4.7197, 4.5477, 4.5231, 4.2429, 4.1789, 4.1928, 4.7333], device='cuda:3'), covar=tensor([0.1132, 0.0850, 0.0931, 0.0806, 0.0841, 0.1705, 0.1183, 0.0899], device='cuda:3'), in_proj_covar=tensor([0.0696, 0.0839, 0.0690, 0.0653, 0.0537, 0.0536, 0.0703, 0.0659], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:33:20,746 INFO [train.py:904] (3/8) Epoch 30, batch 0, loss[loss=0.1797, simple_loss=0.2816, pruned_loss=0.03893, over 17019.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2816, pruned_loss=0.03893, over 17019.00 frames. ], batch size: 50, lr: 2.26e-03, grad_scale: 8.0 2023-05-02 20:33:20,747 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 20:33:28,214 INFO [train.py:938] (3/8) Epoch 30, validation: loss=0.1426, simple_loss=0.2458, pruned_loss=0.01974, over 944034.00 frames. 2023-05-02 20:33:28,214 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 20:34:10,663 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0092, 2.4183, 2.2754, 2.9397, 1.9968, 3.1847, 1.8077, 2.7950], device='cuda:3'), covar=tensor([0.1191, 0.0665, 0.1190, 0.0164, 0.0099, 0.0383, 0.1558, 0.0717], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0178, 0.0198, 0.0199, 0.0200, 0.0214, 0.0209, 0.0196], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:34:28,805 INFO [optim.py:368] (3/8) 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 20:34:29,653 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294400.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:34:36,208 INFO [train.py:904] (3/8) Epoch 30, batch 50, loss[loss=0.1585, simple_loss=0.253, pruned_loss=0.03197, over 17165.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2636, pruned_loss=0.04609, over 751328.04 frames. ], batch size: 46, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:35:02,186 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294423.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:35:22,482 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9205, 2.0413, 2.5519, 2.8773, 2.7689, 3.3610, 2.4008, 3.3904], device='cuda:3'), covar=tensor([0.0329, 0.0618, 0.0427, 0.0430, 0.0436, 0.0240, 0.0604, 0.0204], device='cuda:3'), in_proj_covar=tensor([0.0193, 0.0194, 0.0183, 0.0187, 0.0206, 0.0162, 0.0200, 0.0163], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:35:43,886 INFO [train.py:904] (3/8) Epoch 30, batch 100, loss[loss=0.1463, simple_loss=0.2306, pruned_loss=0.03102, over 17027.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2606, pruned_loss=0.04403, over 1316252.52 frames. ], batch size: 41, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:35:48,568 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6020, 4.6200, 4.5596, 4.1170, 4.5648, 1.8585, 4.3889, 4.2140], device='cuda:3'), covar=tensor([0.0163, 0.0122, 0.0218, 0.0324, 0.0133, 0.2795, 0.0164, 0.0284], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0170, 0.0208, 0.0178, 0.0186, 0.0215, 0.0196, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:36:24,470 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294484.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:32,229 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:46,208 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 150, loss[loss=0.1957, simple_loss=0.294, pruned_loss=0.04871, over 16458.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2603, pruned_loss=0.0434, over 1751874.32 frames. ], batch size: 68, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:37:20,776 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294526.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:32,774 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-05-02 20:37:50,428 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7166, 3.7394, 2.3590, 4.0483, 3.0063, 3.9970, 2.5813, 3.1807], device='cuda:3'), covar=tensor([0.0330, 0.0465, 0.1606, 0.0403, 0.0787, 0.0743, 0.1397, 0.0719], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0179, 0.0194, 0.0170, 0.0178, 0.0217, 0.0204, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:37:53,403 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:59,956 INFO [train.py:904] (3/8) Epoch 30, batch 200, loss[loss=0.1583, simple_loss=0.2607, pruned_loss=0.02795, over 17101.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2595, pruned_loss=0.04236, over 2093395.98 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:38:23,399 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:38:41,538 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4335, 3.4351, 2.2405, 3.6427, 2.7862, 3.5905, 2.3211, 2.8737], device='cuda:3'), covar=tensor([0.0318, 0.0469, 0.1569, 0.0415, 0.0839, 0.0862, 0.1452, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0179, 0.0194, 0.0171, 0.0179, 0.0217, 0.0204, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:39:01,687 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 250, loss[loss=0.1595, simple_loss=0.2485, pruned_loss=0.03527, over 16697.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2572, pruned_loss=0.04195, over 2364213.43 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:39:45,437 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:01,616 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:02,900 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4058, 4.4943, 4.6436, 4.4486, 4.4924, 5.0372, 4.5556, 4.2455], device='cuda:3'), covar=tensor([0.1843, 0.2294, 0.2766, 0.2421, 0.2916, 0.1223, 0.1879, 0.2750], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0634, 0.0707, 0.0513, 0.0686, 0.0726, 0.0544, 0.0681], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 20:40:11,736 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9566, 2.6526, 2.4857, 4.2394, 3.5119, 4.1259, 1.6365, 3.0210], device='cuda:3'), covar=tensor([0.1437, 0.0755, 0.1276, 0.0201, 0.0169, 0.0418, 0.1687, 0.0877], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0204, 0.0204, 0.0217, 0.0212, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:40:15,312 INFO [train.py:904] (3/8) Epoch 30, batch 300, loss[loss=0.1687, simple_loss=0.2525, pruned_loss=0.04248, over 16532.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2558, pruned_loss=0.04115, over 2574595.76 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:41:05,691 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5757, 4.4221, 4.4622, 4.1378, 4.2433, 4.4719, 4.3425, 4.2705], device='cuda:3'), covar=tensor([0.0687, 0.1092, 0.0379, 0.0366, 0.0879, 0.0542, 0.0528, 0.0739], device='cuda:3'), in_proj_covar=tensor([0.0309, 0.0466, 0.0361, 0.0362, 0.0356, 0.0415, 0.0248, 0.0431], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 20:41:06,778 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:12,839 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:19,992 INFO [optim.py:368] (3/8) 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 20:41:25,716 INFO [train.py:904] (3/8) Epoch 30, batch 350, loss[loss=0.1931, simple_loss=0.2757, pruned_loss=0.05525, over 16697.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2534, pruned_loss=0.04002, over 2739653.98 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:42:33,013 INFO [train.py:904] (3/8) Epoch 30, batch 400, loss[loss=0.1407, simple_loss=0.2223, pruned_loss=0.02949, over 12308.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2517, pruned_loss=0.03954, over 2868017.70 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:42:55,400 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3078, 3.2773, 3.7212, 2.2587, 3.1753, 2.5157, 3.7137, 3.6140], device='cuda:3'), covar=tensor([0.0257, 0.1026, 0.0605, 0.2184, 0.0847, 0.1057, 0.0561, 0.1116], device='cuda:3'), in_proj_covar=tensor([0.0160, 0.0168, 0.0169, 0.0157, 0.0147, 0.0132, 0.0144, 0.0181], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:43:06,954 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 2.158e+02 2.521e+02 2.985e+02 1.602e+03, threshold=5.041e+02, percent-clipped=2.0 2023-05-02 20:43:41,854 INFO [train.py:904] (3/8) Epoch 30, batch 450, loss[loss=0.1551, simple_loss=0.2516, pruned_loss=0.02931, over 17132.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2509, pruned_loss=0.03914, over 2976204.77 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:43:50,099 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8480, 2.7538, 2.5957, 4.8638, 3.6464, 4.1941, 1.8160, 3.0887], device='cuda:3'), covar=tensor([0.1430, 0.0952, 0.1379, 0.0249, 0.0226, 0.0510, 0.1736, 0.0894], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0181, 0.0200, 0.0203, 0.0203, 0.0217, 0.0211, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:44:12,072 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294826.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:27,396 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 20:44:31,311 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8949, 4.3405, 4.3699, 3.2375, 3.6537, 4.3135, 3.9250, 2.7643], device='cuda:3'), covar=tensor([0.0512, 0.0077, 0.0054, 0.0370, 0.0164, 0.0105, 0.0098, 0.0450], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0092, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 20:44:36,117 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:49,593 INFO [train.py:904] (3/8) Epoch 30, batch 500, loss[loss=0.1533, simple_loss=0.2325, pruned_loss=0.03703, over 16480.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2502, pruned_loss=0.03853, over 3052903.52 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:45:17,739 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:45:19,793 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5394, 3.5278, 1.8792, 3.6852, 2.8062, 3.6242, 2.1167, 2.8625], device='cuda:3'), covar=tensor([0.0287, 0.0414, 0.1902, 0.0419, 0.0770, 0.0702, 0.1491, 0.0716], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0182, 0.0197, 0.0174, 0.0181, 0.0221, 0.0207, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:45:52,584 INFO [optim.py:368] (3/8) 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 20:45:57,301 INFO [train.py:904] (3/8) Epoch 30, batch 550, loss[loss=0.1569, simple_loss=0.2523, pruned_loss=0.03076, over 17091.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2492, pruned_loss=0.03754, over 3117984.74 frames. ], batch size: 47, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:46:13,074 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1198, 2.3656, 2.7638, 3.0877, 2.9314, 3.5788, 2.5594, 3.5825], device='cuda:3'), covar=tensor([0.0340, 0.0614, 0.0408, 0.0399, 0.0428, 0.0223, 0.0607, 0.0212], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0193, 0.0212, 0.0168, 0.0206, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 20:46:28,990 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:47:05,464 INFO [train.py:904] (3/8) Epoch 30, batch 600, loss[loss=0.1573, simple_loss=0.2349, pruned_loss=0.03986, over 16720.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2488, pruned_loss=0.03774, over 3151473.86 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:47:52,981 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:01,874 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:07,356 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.157e+02 2.572e+02 3.014e+02 4.443e+02, threshold=5.143e+02, percent-clipped=0.0 2023-05-02 20:48:13,121 INFO [train.py:904] (3/8) Epoch 30, batch 650, loss[loss=0.1523, simple_loss=0.2341, pruned_loss=0.03525, over 16797.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2476, pruned_loss=0.03721, over 3196098.84 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:04,906 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295043.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:15,127 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295050.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 20:49:20,778 INFO [train.py:904] (3/8) Epoch 30, batch 700, loss[loss=0.1595, simple_loss=0.2468, pruned_loss=0.03609, over 16469.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2473, pruned_loss=0.03672, over 3225110.50 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:39,349 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:54,480 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:50:03,977 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 20:50:23,319 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 750, loss[loss=0.1567, simple_loss=0.2423, pruned_loss=0.03558, over 16730.00 frames. ], tot_loss[loss=0.1593, simple_loss=0.2463, pruned_loss=0.03617, over 3252809.76 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:50:56,724 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7694, 3.9841, 4.0785, 2.8859, 3.6422, 4.1089, 3.7783, 2.4180], device='cuda:3'), covar=tensor([0.0562, 0.0338, 0.0088, 0.0460, 0.0160, 0.0144, 0.0129, 0.0558], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 20:51:00,101 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:03,348 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:24,557 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:38,209 INFO [train.py:904] (3/8) Epoch 30, batch 800, loss[loss=0.1589, simple_loss=0.26, pruned_loss=0.02891, over 17092.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2461, pruned_loss=0.03564, over 3275840.59 frames. ], batch size: 47, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:51:39,876 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9653, 2.0800, 2.5200, 2.8674, 2.8516, 3.0534, 2.2593, 3.1656], device='cuda:3'), covar=tensor([0.0244, 0.0592, 0.0410, 0.0346, 0.0353, 0.0302, 0.0635, 0.0207], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0193, 0.0212, 0.0168, 0.0205, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 20:51:49,427 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-05-02 20:51:51,488 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5160, 3.5221, 2.2615, 3.7276, 2.9308, 3.6781, 2.4438, 2.9568], device='cuda:3'), covar=tensor([0.0306, 0.0462, 0.1647, 0.0429, 0.0733, 0.0830, 0.1375, 0.0755], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0175, 0.0181, 0.0222, 0.0207, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:52:27,494 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0969, 3.0919, 3.1679, 2.1764, 3.0051, 3.2694, 3.0300, 1.9577], device='cuda:3'), covar=tensor([0.0586, 0.0147, 0.0094, 0.0481, 0.0156, 0.0149, 0.0136, 0.0566], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0100, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 20:52:29,857 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:52:42,511 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.068e+02 2.393e+02 2.794e+02 7.464e+02, threshold=4.786e+02, percent-clipped=2.0 2023-05-02 20:52:44,675 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1256, 3.1332, 3.4801, 2.1717, 3.0210, 2.2092, 3.6977, 3.5416], device='cuda:3'), covar=tensor([0.0246, 0.1065, 0.0646, 0.2282, 0.0925, 0.1186, 0.0498, 0.1047], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0158, 0.0148, 0.0133, 0.0145, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 20:52:48,745 INFO [train.py:904] (3/8) Epoch 30, batch 850, loss[loss=0.1695, simple_loss=0.2675, pruned_loss=0.03573, over 17091.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2449, pruned_loss=0.03546, over 3288749.77 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:53:18,734 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:21,291 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295227.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:56,740 INFO [train.py:904] (3/8) Epoch 30, batch 900, loss[loss=0.1535, simple_loss=0.2443, pruned_loss=0.03136, over 16696.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2449, pruned_loss=0.03513, over 3300332.88 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:54:27,485 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:30,574 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4537, 5.4556, 5.3109, 4.7619, 4.9578, 5.3874, 5.3080, 4.9495], device='cuda:3'), covar=tensor([0.0608, 0.0478, 0.0341, 0.0360, 0.1126, 0.0418, 0.0269, 0.0828], device='cuda:3'), in_proj_covar=tensor([0.0320, 0.0483, 0.0373, 0.0376, 0.0370, 0.0432, 0.0258, 0.0448], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 20:54:33,453 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295279.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:41,799 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6810, 3.9022, 4.0599, 2.9305, 3.6678, 4.1304, 3.7380, 2.4757], device='cuda:3'), covar=tensor([0.0515, 0.0286, 0.0074, 0.0387, 0.0134, 0.0135, 0.0127, 0.0489], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 20:54:43,609 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295286.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:02,821 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.063e+02 2.465e+02 2.923e+02 4.814e+02, threshold=4.929e+02, percent-clipped=1.0 2023-05-02 20:55:08,088 INFO [train.py:904] (3/8) Epoch 30, batch 950, loss[loss=0.1434, simple_loss=0.2239, pruned_loss=0.03143, over 16754.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.245, pruned_loss=0.03581, over 3298669.35 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:55:48,087 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4082, 3.4299, 3.9314, 2.2020, 3.2420, 2.4573, 3.8903, 3.7155], device='cuda:3'), covar=tensor([0.0293, 0.1082, 0.0519, 0.2173, 0.0854, 0.1053, 0.0583, 0.1118], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0158, 0.0149, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 20:55:49,172 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:59,243 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:06,370 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295345.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 20:56:17,470 INFO [train.py:904] (3/8) Epoch 30, batch 1000, loss[loss=0.1532, simple_loss=0.2345, pruned_loss=0.03591, over 16443.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2435, pruned_loss=0.03548, over 3310237.42 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:56:17,852 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7194, 3.8540, 2.1773, 4.2922, 2.9427, 4.2129, 2.1316, 3.0634], device='cuda:3'), covar=tensor([0.0335, 0.0378, 0.1893, 0.0330, 0.0912, 0.0480, 0.2009, 0.0834], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0176, 0.0181, 0.0223, 0.0207, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 20:56:22,202 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 20:56:24,362 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:39,404 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8751, 4.2881, 3.0073, 2.3627, 2.5916, 2.5926, 4.6371, 3.5309], device='cuda:3'), covar=tensor([0.3082, 0.0580, 0.2018, 0.3373, 0.3326, 0.2419, 0.0405, 0.1606], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0278, 0.0317, 0.0331, 0.0308, 0.0282, 0.0307, 0.0356], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 20:57:02,095 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 20:57:12,957 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:20,048 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.023e+02 2.384e+02 2.929e+02 5.067e+02, threshold=4.769e+02, percent-clipped=1.0 2023-05-02 20:57:26,447 INFO [train.py:904] (3/8) Epoch 30, batch 1050, loss[loss=0.1554, simple_loss=0.2449, pruned_loss=0.03299, over 16663.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2434, pruned_loss=0.03537, over 3321281.82 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:57:30,781 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-02 20:57:48,287 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295420.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:54,259 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:58:36,214 INFO [train.py:904] (3/8) Epoch 30, batch 1100, loss[loss=0.163, simple_loss=0.2449, pruned_loss=0.04057, over 16539.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2436, pruned_loss=0.0354, over 3329285.29 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:59:38,543 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.061e+02 2.413e+02 2.897e+02 1.352e+03, threshold=4.827e+02, percent-clipped=9.0 2023-05-02 20:59:43,304 INFO [train.py:904] (3/8) Epoch 30, batch 1150, loss[loss=0.1428, simple_loss=0.2396, pruned_loss=0.02298, over 17143.00 frames. ], tot_loss[loss=0.157, simple_loss=0.244, pruned_loss=0.03497, over 3328049.61 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:00:51,733 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9345, 3.0030, 3.2656, 2.0920, 2.8641, 2.2673, 3.4243, 3.3324], device='cuda:3'), covar=tensor([0.0281, 0.0948, 0.0623, 0.2036, 0.0891, 0.1079, 0.0565, 0.1034], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0158, 0.0149, 0.0133, 0.0146, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 21:00:52,295 INFO [train.py:904] (3/8) Epoch 30, batch 1200, loss[loss=0.165, simple_loss=0.2393, pruned_loss=0.04538, over 16716.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2431, pruned_loss=0.03453, over 3324920.25 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:00:54,971 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:00:57,942 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295558.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:01:29,914 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295581.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:01:55,307 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 1250, loss[loss=0.1432, simple_loss=0.2286, pruned_loss=0.02891, over 16319.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.2426, pruned_loss=0.03458, over 3320177.93 frames. ], batch size: 36, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:02:08,667 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0248, 3.1590, 2.8880, 5.2035, 4.3350, 4.4193, 1.9489, 3.4911], device='cuda:3'), covar=tensor([0.1287, 0.0774, 0.1216, 0.0215, 0.0241, 0.0428, 0.1587, 0.0681], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0182, 0.0201, 0.0205, 0.0205, 0.0218, 0.0211, 0.0199], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:02:12,385 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 21:02:18,646 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295617.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:02:21,375 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295619.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:02:26,440 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 21:02:34,348 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295629.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:02:39,600 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6171, 3.6565, 2.0730, 3.8384, 2.9845, 3.7806, 2.1414, 2.9288], device='cuda:3'), covar=tensor([0.0293, 0.0419, 0.1944, 0.0397, 0.0720, 0.0731, 0.1946, 0.0798], device='cuda:3'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0176, 0.0181, 0.0223, 0.0207, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:02:43,093 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:02:58,026 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295645.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:08,457 INFO [train.py:904] (3/8) Epoch 30, batch 1300, loss[loss=0.1541, simple_loss=0.2335, pruned_loss=0.03741, over 16224.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2429, pruned_loss=0.03533, over 3319538.96 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:03:23,105 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 21:03:37,958 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4017, 4.2210, 4.4667, 4.5709, 4.6852, 4.2315, 4.5383, 4.6868], device='cuda:3'), covar=tensor([0.1644, 0.1247, 0.1384, 0.0747, 0.0613, 0.1327, 0.2322, 0.0768], device='cuda:3'), in_proj_covar=tensor([0.0702, 0.0854, 0.0987, 0.0872, 0.0662, 0.0684, 0.0723, 0.0844], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:03:40,228 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-02 21:03:52,643 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:57,718 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:59,007 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295690.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:04:02,885 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295693.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:11,929 INFO [optim.py:368] (3/8) 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 21:04:17,845 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-02 21:04:18,058 INFO [train.py:904] (3/8) Epoch 30, batch 1350, loss[loss=0.1494, simple_loss=0.2467, pruned_loss=0.0261, over 17236.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2433, pruned_loss=0.03486, over 3328794.07 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:04:33,099 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:46,286 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295724.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:17,193 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 1400, loss[loss=0.1702, simple_loss=0.2562, pruned_loss=0.04209, over 16658.00 frames. ], tot_loss[loss=0.157, simple_loss=0.2435, pruned_loss=0.0352, over 3327515.36 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:05:45,942 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9724, 2.1962, 2.3984, 3.4335, 2.2011, 2.4237, 2.3455, 2.3330], device='cuda:3'), covar=tensor([0.1719, 0.3827, 0.3415, 0.0923, 0.4342, 0.2767, 0.3831, 0.3823], device='cuda:3'), in_proj_covar=tensor([0.0428, 0.0481, 0.0393, 0.0343, 0.0450, 0.0551, 0.0454, 0.0565], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:05:52,089 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295772.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:06:30,026 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 1450, loss[loss=0.1332, simple_loss=0.2198, pruned_loss=0.02329, over 16797.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2431, pruned_loss=0.03557, over 3326414.26 frames. ], batch size: 42, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:06:57,425 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:07:02,211 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4690, 3.4402, 3.4893, 3.5688, 3.6127, 3.3003, 3.5372, 3.6760], device='cuda:3'), covar=tensor([0.1395, 0.0958, 0.1136, 0.0670, 0.0651, 0.2373, 0.1462, 0.0827], device='cuda:3'), in_proj_covar=tensor([0.0706, 0.0857, 0.0991, 0.0875, 0.0665, 0.0687, 0.0725, 0.0848], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:07:20,939 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1496, 2.1951, 2.8235, 3.1915, 3.0694, 3.6736, 2.2455, 3.7463], device='cuda:3'), covar=tensor([0.0295, 0.0670, 0.0367, 0.0345, 0.0373, 0.0209, 0.0772, 0.0172], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0202, 0.0190, 0.0196, 0.0214, 0.0170, 0.0207, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 21:07:38,818 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-05-02 21:07:44,322 INFO [train.py:904] (3/8) Epoch 30, batch 1500, loss[loss=0.1568, simple_loss=0.2461, pruned_loss=0.03368, over 16687.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2421, pruned_loss=0.03542, over 3313161.32 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:08:11,943 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:21,931 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,054 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:47,908 INFO [optim.py:368] (3/8) 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 21:08:54,107 INFO [train.py:904] (3/8) Epoch 30, batch 1550, loss[loss=0.161, simple_loss=0.2577, pruned_loss=0.03215, over 17178.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2424, pruned_loss=0.03564, over 3318209.21 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:09:05,604 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295912.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:09:08,344 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295914.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:09:29,567 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295929.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:36,249 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295934.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:09:37,421 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295935.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:02,943 INFO [train.py:904] (3/8) Epoch 30, batch 1600, loss[loss=0.1715, simple_loss=0.25, pruned_loss=0.04651, over 16455.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2445, pruned_loss=0.0364, over 3310997.29 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:10:42,972 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295985.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:10:51,245 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:11:05,441 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 1650, loss[loss=0.1656, simple_loss=0.2567, pruned_loss=0.03726, over 17225.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2457, pruned_loss=0.03695, over 3314178.66 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:11:30,080 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296015.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:00,955 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:06,316 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:16,884 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:24,440 INFO [train.py:904] (3/8) Epoch 30, batch 1700, loss[loss=0.1524, simple_loss=0.2431, pruned_loss=0.03088, over 16750.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2486, pruned_loss=0.03755, over 3309976.25 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:12:36,066 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296063.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:05,160 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.4004, 5.3644, 5.2652, 4.7442, 4.9082, 5.2816, 5.2395, 4.9256], device='cuda:3'), covar=tensor([0.0632, 0.0580, 0.0357, 0.0385, 0.1215, 0.0527, 0.0292, 0.0815], device='cuda:3'), in_proj_covar=tensor([0.0324, 0.0490, 0.0379, 0.0381, 0.0376, 0.0437, 0.0261, 0.0454], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:13:26,990 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 1750, loss[loss=0.1546, simple_loss=0.2408, pruned_loss=0.03418, over 16820.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.249, pruned_loss=0.03731, over 3312679.88 frames. ], batch size: 83, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:13:40,797 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:44,315 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7710, 4.0603, 2.7437, 4.6914, 3.4205, 4.6175, 2.7342, 3.4530], device='cuda:3'), covar=tensor([0.0378, 0.0429, 0.1574, 0.0318, 0.0778, 0.0513, 0.1647, 0.0732], device='cuda:3'), in_proj_covar=tensor([0.0179, 0.0184, 0.0198, 0.0178, 0.0182, 0.0225, 0.0208, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:14:25,219 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6177, 4.9755, 5.2579, 5.2175, 5.2853, 4.9170, 4.6647, 4.7186], device='cuda:3'), covar=tensor([0.0700, 0.0850, 0.0683, 0.0775, 0.0807, 0.0753, 0.1525, 0.0597], device='cuda:3'), in_proj_covar=tensor([0.0448, 0.0510, 0.0488, 0.0452, 0.0537, 0.0519, 0.0593, 0.0416], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 21:14:41,095 INFO [train.py:904] (3/8) Epoch 30, batch 1800, loss[loss=0.1682, simple_loss=0.2581, pruned_loss=0.03909, over 16481.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2494, pruned_loss=0.03684, over 3316533.70 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:15:12,949 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:15:38,575 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 21:15:47,298 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.165e+02 2.480e+02 2.953e+02 5.408e+02, threshold=4.959e+02, percent-clipped=1.0 2023-05-02 21:15:51,605 INFO [train.py:904] (3/8) Epoch 30, batch 1850, loss[loss=0.1785, simple_loss=0.2786, pruned_loss=0.03919, over 16775.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2499, pruned_loss=0.03692, over 3316246.32 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:16:03,685 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296212.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:16:06,023 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296214.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:16:10,949 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8468, 2.0344, 2.4374, 2.7205, 2.7632, 2.7491, 2.0638, 2.9911], device='cuda:3'), covar=tensor([0.0242, 0.0592, 0.0429, 0.0359, 0.0383, 0.0379, 0.0657, 0.0226], device='cuda:3'), in_proj_covar=tensor([0.0202, 0.0202, 0.0190, 0.0196, 0.0214, 0.0171, 0.0207, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 21:16:27,856 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296229.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:17:02,020 INFO [train.py:904] (3/8) Epoch 30, batch 1900, loss[loss=0.1745, simple_loss=0.2761, pruned_loss=0.03642, over 17279.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2496, pruned_loss=0.03622, over 3318013.00 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:17:04,819 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8470, 4.3758, 3.0595, 2.4112, 2.6238, 2.6755, 4.7610, 3.6166], device='cuda:3'), covar=tensor([0.3076, 0.0596, 0.1929, 0.3262, 0.3129, 0.2268, 0.0393, 0.1520], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0279, 0.0317, 0.0331, 0.0309, 0.0283, 0.0307, 0.0358], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:17:09,978 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296260.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:17:12,891 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296262.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:17:45,526 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296285.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:18:06,422 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.062e+02 2.316e+02 2.814e+02 5.874e+02, threshold=4.633e+02, percent-clipped=2.0 2023-05-02 21:18:10,598 INFO [train.py:904] (3/8) Epoch 30, batch 1950, loss[loss=0.1394, simple_loss=0.2289, pruned_loss=0.02493, over 16756.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2497, pruned_loss=0.03591, over 3313828.39 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:18:22,636 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8824, 1.9990, 2.4347, 2.6538, 2.7815, 2.7182, 2.0520, 3.0182], device='cuda:3'), covar=tensor([0.0232, 0.0612, 0.0440, 0.0373, 0.0392, 0.0421, 0.0701, 0.0217], device='cuda:3'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0197, 0.0214, 0.0171, 0.0208, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 21:18:26,937 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:18:52,568 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296333.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:18:56,036 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3388, 2.4059, 2.3671, 4.1656, 2.2714, 2.7338, 2.4201, 2.5289], device='cuda:3'), covar=tensor([0.1486, 0.3791, 0.3391, 0.0614, 0.4441, 0.2798, 0.3932, 0.3803], device='cuda:3'), in_proj_covar=tensor([0.0430, 0.0483, 0.0394, 0.0345, 0.0451, 0.0554, 0.0456, 0.0567], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:19:02,324 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296341.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:19:09,723 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.9478, 4.9120, 4.8034, 4.2566, 4.8547, 2.1238, 4.6225, 4.5698], device='cuda:3'), covar=tensor([0.0226, 0.0146, 0.0232, 0.0414, 0.0151, 0.2790, 0.0192, 0.0269], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0180, 0.0218, 0.0188, 0.0196, 0.0223, 0.0207, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:19:10,933 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9329, 2.0524, 2.4982, 2.7938, 2.8020, 2.8346, 2.1202, 3.1246], device='cuda:3'), covar=tensor([0.0258, 0.0639, 0.0467, 0.0336, 0.0418, 0.0398, 0.0713, 0.0236], device='cuda:3'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0196, 0.0214, 0.0171, 0.0207, 0.0171], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 21:19:21,053 INFO [train.py:904] (3/8) Epoch 30, batch 2000, loss[loss=0.1701, simple_loss=0.2619, pruned_loss=0.03914, over 17070.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2502, pruned_loss=0.03613, over 3306788.05 frames. ], batch size: 55, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:19:37,421 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 21:19:51,092 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:00,817 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7256, 3.9154, 2.6953, 4.6537, 3.2975, 4.5634, 2.7156, 3.3919], device='cuda:3'), covar=tensor([0.0389, 0.0469, 0.1504, 0.0297, 0.0784, 0.0552, 0.1583, 0.0739], device='cuda:3'), in_proj_covar=tensor([0.0180, 0.0186, 0.0198, 0.0179, 0.0183, 0.0226, 0.0209, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:20:09,533 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296389.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:22,980 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.6728, 4.6246, 4.5711, 3.9867, 4.6101, 1.8598, 4.3934, 4.1882], device='cuda:3'), covar=tensor([0.0191, 0.0141, 0.0203, 0.0334, 0.0128, 0.3033, 0.0172, 0.0269], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0179, 0.0217, 0.0188, 0.0196, 0.0223, 0.0207, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:20:25,431 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 2.133e+02 2.447e+02 2.850e+02 4.858e+02, threshold=4.894e+02, percent-clipped=1.0 2023-05-02 21:20:29,080 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8031, 2.6988, 2.1835, 2.3502, 2.9500, 2.7031, 3.4048, 3.3035], device='cuda:3'), covar=tensor([0.0189, 0.0625, 0.0793, 0.0724, 0.0421, 0.0552, 0.0308, 0.0353], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0249, 0.0252], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:20:30,259 INFO [train.py:904] (3/8) Epoch 30, batch 2050, loss[loss=0.156, simple_loss=0.2429, pruned_loss=0.03452, over 17245.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2495, pruned_loss=0.03607, over 3315227.22 frames. ], batch size: 45, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:20:32,301 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:12,415 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:39,722 INFO [train.py:904] (3/8) Epoch 30, batch 2100, loss[loss=0.1627, simple_loss=0.2492, pruned_loss=0.03812, over 16814.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2503, pruned_loss=0.03647, over 3316929.64 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:21:43,701 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3168, 5.2817, 5.0320, 4.4770, 5.1027, 1.9352, 4.8747, 4.8942], device='cuda:3'), covar=tensor([0.0110, 0.0105, 0.0227, 0.0426, 0.0120, 0.2995, 0.0149, 0.0247], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0179, 0.0217, 0.0187, 0.0195, 0.0222, 0.0207, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:22:09,571 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296476.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:33,155 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:36,818 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:39,027 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296497.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:46,673 INFO [optim.py:368] (3/8) 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 21:22:48,895 INFO [train.py:904] (3/8) Epoch 30, batch 2150, loss[loss=0.1868, simple_loss=0.2716, pruned_loss=0.05102, over 16710.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2514, pruned_loss=0.03701, over 3318934.93 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:23:03,719 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:16,224 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296524.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:23,417 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296529.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:23:35,303 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2929, 4.3449, 4.6341, 4.6192, 4.6847, 4.3941, 4.4034, 4.2771], device='cuda:3'), covar=tensor([0.0404, 0.0713, 0.0446, 0.0446, 0.0513, 0.0462, 0.0831, 0.0698], device='cuda:3'), in_proj_covar=tensor([0.0448, 0.0510, 0.0487, 0.0451, 0.0534, 0.0517, 0.0593, 0.0415], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 21:23:43,121 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2537, 5.7657, 5.9184, 5.5582, 5.6961, 6.2911, 5.8012, 5.4438], device='cuda:3'), covar=tensor([0.0912, 0.2044, 0.2706, 0.2267, 0.2691, 0.1030, 0.1594, 0.2460], device='cuda:3'), in_proj_covar=tensor([0.0438, 0.0654, 0.0731, 0.0532, 0.0710, 0.0749, 0.0562, 0.0706], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:23:56,825 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296553.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:23:57,526 INFO [train.py:904] (3/8) Epoch 30, batch 2200, loss[loss=0.18, simple_loss=0.276, pruned_loss=0.04204, over 17160.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2522, pruned_loss=0.03742, over 3325443.21 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:24:04,137 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:13,127 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4211, 2.4101, 2.4036, 4.2561, 2.4224, 2.7305, 2.4746, 2.5736], device='cuda:3'), covar=tensor([0.1505, 0.3851, 0.3448, 0.0625, 0.4242, 0.2874, 0.3731, 0.3868], device='cuda:3'), in_proj_covar=tensor([0.0431, 0.0484, 0.0395, 0.0346, 0.0451, 0.0555, 0.0457, 0.0568], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:24:27,041 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:04,056 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.151e+02 2.512e+02 3.134e+02 5.584e+02, threshold=5.024e+02, percent-clipped=2.0 2023-05-02 21:25:06,331 INFO [train.py:904] (3/8) Epoch 30, batch 2250, loss[loss=0.1853, simple_loss=0.2646, pruned_loss=0.05295, over 16444.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.253, pruned_loss=0.03755, over 3322677.73 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:25:20,037 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:31,126 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-05-02 21:26:17,731 INFO [train.py:904] (3/8) Epoch 30, batch 2300, loss[loss=0.1732, simple_loss=0.2592, pruned_loss=0.04357, over 16460.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2534, pruned_loss=0.03803, over 3318308.83 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:26:41,409 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:45,113 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:05,655 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:24,078 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 2350, loss[loss=0.1557, simple_loss=0.2508, pruned_loss=0.03027, over 17168.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2535, pruned_loss=0.03788, over 3313266.02 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:27:28,738 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:30,521 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.7848, 4.2232, 4.1798, 2.9416, 3.5466, 4.1977, 3.7026, 2.6377], device='cuda:3'), covar=tensor([0.0511, 0.0109, 0.0065, 0.0408, 0.0170, 0.0119, 0.0127, 0.0472], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0105, 0.0117, 0.0100, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 21:28:30,550 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:36,423 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 2400, loss[loss=0.1724, simple_loss=0.2663, pruned_loss=0.03925, over 16726.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2547, pruned_loss=0.03841, over 3320958.47 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:28:54,495 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 21:29:27,114 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 30, batch 2450, loss[loss=0.1557, simple_loss=0.2435, pruned_loss=0.03399, over 16596.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2556, pruned_loss=0.0385, over 3309708.07 frames. ], batch size: 75, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:30:10,974 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1095, 5.0378, 4.9634, 4.4607, 4.6550, 4.9928, 4.9985, 4.6239], device='cuda:3'), covar=tensor([0.0613, 0.0711, 0.0393, 0.0429, 0.1097, 0.0510, 0.0369, 0.0871], device='cuda:3'), in_proj_covar=tensor([0.0329, 0.0499, 0.0385, 0.0388, 0.0382, 0.0445, 0.0265, 0.0461], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:30:48,228 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296848.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:30:51,472 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8220, 4.2347, 2.9915, 2.3258, 2.6917, 2.6332, 4.5912, 3.6369], device='cuda:3'), covar=tensor([0.3080, 0.0641, 0.2139, 0.3375, 0.2971, 0.2177, 0.0473, 0.1448], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0278, 0.0317, 0.0330, 0.0309, 0.0282, 0.0307, 0.0357], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:30:55,421 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296853.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:30:56,382 INFO [train.py:904] (3/8) Epoch 30, batch 2500, loss[loss=0.1824, simple_loss=0.2593, pruned_loss=0.05272, over 16712.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2549, pruned_loss=0.03822, over 3317357.69 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:31:19,657 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:23,956 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:59,216 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8497, 3.7167, 4.0781, 2.3152, 4.1694, 4.2407, 3.2619, 3.4451], device='cuda:3'), covar=tensor([0.0774, 0.0269, 0.0213, 0.1104, 0.0103, 0.0241, 0.0452, 0.0397], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:32:04,914 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 2550, loss[loss=0.1507, simple_loss=0.2447, pruned_loss=0.02832, over 17210.00 frames. ], tot_loss[loss=0.166, simple_loss=0.255, pruned_loss=0.03848, over 3315876.68 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:32:49,096 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296934.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:32:49,212 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3254, 2.3250, 2.2950, 4.1472, 2.2813, 2.6487, 2.4505, 2.5020], device='cuda:3'), covar=tensor([0.1534, 0.4194, 0.3661, 0.0690, 0.4744, 0.3011, 0.3967, 0.4131], device='cuda:3'), in_proj_covar=tensor([0.0432, 0.0485, 0.0396, 0.0347, 0.0453, 0.0557, 0.0458, 0.0569], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:32:54,105 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 2600, loss[loss=0.1589, simple_loss=0.2561, pruned_loss=0.0309, over 17086.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2555, pruned_loss=0.0388, over 3315823.31 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:33:36,744 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:39,651 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:18,418 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:34:24,414 INFO [train.py:904] (3/8) Epoch 30, batch 2650, loss[loss=0.2198, simple_loss=0.2925, pruned_loss=0.07353, over 12400.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2557, pruned_loss=0.03863, over 3309724.65 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:34:46,003 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297019.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:57,886 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 21:35:13,085 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8584, 4.6649, 4.5945, 3.0787, 3.8003, 4.5435, 3.9467, 2.9905], device='cuda:3'), covar=tensor([0.0562, 0.0075, 0.0052, 0.0451, 0.0154, 0.0104, 0.0117, 0.0439], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0137, 0.0105, 0.0117, 0.0100, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 21:35:20,544 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:34,415 INFO [train.py:904] (3/8) Epoch 30, batch 2700, loss[loss=0.1819, simple_loss=0.2606, pruned_loss=0.05155, over 16280.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2563, pruned_loss=0.03825, over 3315003.48 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:36:19,407 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9033, 4.0855, 2.6298, 4.7185, 3.3412, 4.6198, 2.7600, 3.4981], device='cuda:3'), covar=tensor([0.0340, 0.0395, 0.1708, 0.0315, 0.0769, 0.0525, 0.1595, 0.0701], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0186, 0.0200, 0.0180, 0.0185, 0.0227, 0.0210, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:36:23,925 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 30, batch 2750, loss[loss=0.161, simple_loss=0.2638, pruned_loss=0.02912, over 17129.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2563, pruned_loss=0.03757, over 3323491.13 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:37:30,682 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:44,362 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297148.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:50,761 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:51,587 INFO [train.py:904] (3/8) Epoch 30, batch 2800, loss[loss=0.1442, simple_loss=0.231, pruned_loss=0.02867, over 17231.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2562, pruned_loss=0.03744, over 3335113.93 frames. ], batch size: 44, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:38:14,251 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:26,511 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 21:38:49,782 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:57,320 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 30, batch 2850, loss[loss=0.1534, simple_loss=0.2501, pruned_loss=0.02839, over 17129.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2559, pruned_loss=0.03762, over 3329046.44 frames. ], batch size: 48, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:39:15,367 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3586, 3.5059, 3.7505, 2.4456, 3.4175, 3.8605, 3.5332, 2.2271], device='cuda:3'), covar=tensor([0.0544, 0.0205, 0.0068, 0.0466, 0.0140, 0.0105, 0.0118, 0.0520], device='cuda:3'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0135, 0.0104, 0.0116, 0.0099, 0.0131], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 21:39:21,722 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:26,926 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297229.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:40:10,536 INFO [train.py:904] (3/8) Epoch 30, batch 2900, loss[loss=0.1714, simple_loss=0.2465, pruned_loss=0.04815, over 16719.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2537, pruned_loss=0.03744, over 3339098.82 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:40:33,102 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:40:51,681 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=297283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:08,385 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297294.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:09,508 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8426, 5.0335, 5.2390, 4.9299, 5.0416, 5.6501, 5.0785, 4.7693], device='cuda:3'), covar=tensor([0.1381, 0.2293, 0.2498, 0.2343, 0.2587, 0.1019, 0.1746, 0.2702], device='cuda:3'), in_proj_covar=tensor([0.0440, 0.0657, 0.0732, 0.0532, 0.0712, 0.0750, 0.0563, 0.0710], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:41:22,344 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.045e+02 2.480e+02 2.886e+02 5.342e+02, threshold=4.960e+02, percent-clipped=3.0 2023-05-02 21:41:22,359 INFO [train.py:904] (3/8) Epoch 30, batch 2950, loss[loss=0.1718, simple_loss=0.2715, pruned_loss=0.03604, over 16796.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2525, pruned_loss=0.03772, over 3339711.19 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:41:41,509 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:18,964 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 3000, loss[loss=0.1692, simple_loss=0.2663, pruned_loss=0.03606, over 16719.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2532, pruned_loss=0.03832, over 3334098.97 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:42:33,297 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 21:42:42,092 INFO [train.py:938] (3/8) Epoch 30, validation: loss=0.1331, simple_loss=0.238, pruned_loss=0.01415, over 944034.00 frames. 2023-05-02 21:42:42,093 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 21:43:35,187 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297392.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:43:45,091 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6424, 2.4015, 2.3862, 3.9337, 3.1818, 3.8991, 1.4690, 2.7389], device='cuda:3'), covar=tensor([0.1732, 0.0890, 0.1494, 0.0231, 0.0181, 0.0389, 0.2095, 0.1006], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0210, 0.0209, 0.0222, 0.0214, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:43:51,743 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 3050, loss[loss=0.162, simple_loss=0.2592, pruned_loss=0.03242, over 17117.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2532, pruned_loss=0.03828, over 3330823.82 frames. ], batch size: 47, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:01,424 INFO [train.py:904] (3/8) Epoch 30, batch 3100, loss[loss=0.1688, simple_loss=0.2438, pruned_loss=0.04696, over 16416.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2527, pruned_loss=0.03819, over 3331926.37 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:55,841 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.2166, 5.7927, 5.8958, 5.5805, 5.6495, 6.2497, 5.7974, 5.4872], device='cuda:3'), covar=tensor([0.1042, 0.2296, 0.3016, 0.2252, 0.2764, 0.1229, 0.1680, 0.2573], device='cuda:3'), in_proj_covar=tensor([0.0442, 0.0659, 0.0734, 0.0533, 0.0714, 0.0751, 0.0563, 0.0712], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:46:07,372 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 3150, loss[loss=0.173, simple_loss=0.2526, pruned_loss=0.04673, over 16824.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2517, pruned_loss=0.03821, over 3329188.54 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:46:42,075 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297529.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:47:17,203 INFO [train.py:904] (3/8) Epoch 30, batch 3200, loss[loss=0.1556, simple_loss=0.2425, pruned_loss=0.0343, over 16783.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2514, pruned_loss=0.03801, over 3321872.15 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:47:38,107 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 21:47:49,842 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:47:51,694 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:47:51,783 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5047, 4.4443, 4.4376, 4.1573, 4.1988, 4.4697, 4.2542, 4.2540], device='cuda:3'), covar=tensor([0.0657, 0.0747, 0.0335, 0.0338, 0.0816, 0.0472, 0.0586, 0.0696], device='cuda:3'), in_proj_covar=tensor([0.0331, 0.0501, 0.0387, 0.0390, 0.0383, 0.0448, 0.0267, 0.0465], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:48:13,632 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 21:48:27,259 INFO [train.py:904] (3/8) Epoch 30, batch 3250, loss[loss=0.1711, simple_loss=0.2525, pruned_loss=0.04484, over 16847.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2513, pruned_loss=0.03808, over 3317379.52 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:49:20,231 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297642.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:49:36,042 INFO [train.py:904] (3/8) Epoch 30, batch 3300, loss[loss=0.1634, simple_loss=0.2452, pruned_loss=0.04077, over 16899.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2528, pruned_loss=0.03855, over 3312931.89 frames. ], batch size: 109, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:46,297 INFO [optim.py:368] (3/8) 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 21:50:46,313 INFO [train.py:904] (3/8) Epoch 30, batch 3350, loss[loss=0.1794, simple_loss=0.2588, pruned_loss=0.04997, over 16483.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2529, pruned_loss=0.03827, over 3315679.20 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:50:52,756 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 21:50:58,011 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4313, 2.4253, 2.3134, 4.1097, 2.4570, 2.7913, 2.4794, 2.5721], device='cuda:3'), covar=tensor([0.1508, 0.4054, 0.3614, 0.0706, 0.4323, 0.2889, 0.3986, 0.3855], device='cuda:3'), in_proj_covar=tensor([0.0434, 0.0486, 0.0396, 0.0347, 0.0454, 0.0558, 0.0459, 0.0570], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:51:56,062 INFO [train.py:904] (3/8) Epoch 30, batch 3400, loss[loss=0.172, simple_loss=0.2557, pruned_loss=0.04418, over 16456.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2527, pruned_loss=0.0379, over 3302929.47 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:52:57,888 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6860, 3.1737, 3.5496, 2.0602, 3.6494, 3.6814, 3.1266, 2.9262], device='cuda:3'), covar=tensor([0.0721, 0.0297, 0.0235, 0.1117, 0.0139, 0.0218, 0.0420, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0139, 0.0089, 0.0135, 0.0132, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:53:07,131 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 3450, loss[loss=0.1573, simple_loss=0.253, pruned_loss=0.03075, over 17087.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2507, pruned_loss=0.03727, over 3311518.77 frames. ], batch size: 47, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:53:23,341 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9913, 2.2213, 2.3478, 3.5606, 2.2391, 2.4939, 2.3530, 2.3460], device='cuda:3'), covar=tensor([0.1651, 0.3755, 0.3268, 0.0796, 0.4185, 0.2691, 0.3762, 0.3394], device='cuda:3'), in_proj_covar=tensor([0.0434, 0.0487, 0.0396, 0.0348, 0.0454, 0.0559, 0.0459, 0.0571], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 21:54:09,067 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0168, 2.9897, 2.7627, 4.9998, 3.9502, 4.4207, 1.6842, 3.1778], device='cuda:3'), covar=tensor([0.1322, 0.0788, 0.1242, 0.0236, 0.0249, 0.0455, 0.1729, 0.0847], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0211, 0.0210, 0.0222, 0.0214, 0.0202], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:54:17,197 INFO [train.py:904] (3/8) Epoch 30, batch 3500, loss[loss=0.1743, simple_loss=0.2543, pruned_loss=0.04716, over 16715.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2488, pruned_loss=0.03696, over 3323353.43 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:54:27,828 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-02 21:54:36,657 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.5454, 3.5764, 2.2671, 3.8473, 2.9144, 3.7883, 2.3696, 2.9675], device='cuda:3'), covar=tensor([0.0296, 0.0482, 0.1614, 0.0390, 0.0840, 0.0939, 0.1552, 0.0758], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0186, 0.0199, 0.0180, 0.0183, 0.0226, 0.0208, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:54:51,988 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:55:09,860 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8941, 2.3278, 2.6026, 2.9385, 2.8141, 3.4628, 2.5245, 3.4627], device='cuda:3'), covar=tensor([0.0365, 0.0588, 0.0475, 0.0430, 0.0476, 0.0251, 0.0548, 0.0203], device='cuda:3'), in_proj_covar=tensor([0.0206, 0.0204, 0.0194, 0.0199, 0.0218, 0.0174, 0.0210, 0.0174], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 21:55:28,082 INFO [train.py:904] (3/8) Epoch 30, batch 3550, loss[loss=0.1594, simple_loss=0.2597, pruned_loss=0.02961, over 17132.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2475, pruned_loss=0.03675, over 3322869.26 frames. ], batch size: 48, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:55:29,303 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.998e+02 2.319e+02 2.785e+02 6.034e+02, threshold=4.637e+02, percent-clipped=1.0 2023-05-02 21:55:59,067 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:56:38,332 INFO [train.py:904] (3/8) Epoch 30, batch 3600, loss[loss=0.1361, simple_loss=0.2289, pruned_loss=0.02166, over 17235.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2469, pruned_loss=0.03674, over 3321559.65 frames. ], batch size: 45, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:56:42,413 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0167, 5.0655, 5.4288, 5.4364, 5.4547, 5.1381, 5.0583, 4.9082], device='cuda:3'), covar=tensor([0.0401, 0.0523, 0.0447, 0.0408, 0.0477, 0.0404, 0.0965, 0.0474], device='cuda:3'), in_proj_covar=tensor([0.0458, 0.0523, 0.0498, 0.0459, 0.0544, 0.0525, 0.0608, 0.0423], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 21:56:42,454 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:56:54,703 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8903, 2.9539, 2.8001, 2.8801, 3.2850, 2.9783, 3.5933, 3.4006], device='cuda:3'), covar=tensor([0.0196, 0.0463, 0.0502, 0.0436, 0.0280, 0.0419, 0.0244, 0.0307], device='cuda:3'), in_proj_covar=tensor([0.0242, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0251, 0.0253], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 21:57:24,206 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8999, 2.8868, 2.8164, 4.5045, 3.9316, 4.2456, 1.5617, 3.3208], device='cuda:3'), covar=tensor([0.1340, 0.0667, 0.1065, 0.0183, 0.0159, 0.0347, 0.1642, 0.0705], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0210, 0.0209, 0.0221, 0.0213, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:57:50,549 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9984, 2.8904, 2.7274, 4.8253, 3.5445, 4.1926, 1.8373, 3.1329], device='cuda:3'), covar=tensor([0.1408, 0.0941, 0.1350, 0.0268, 0.0272, 0.0488, 0.1758, 0.0900], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0210, 0.0209, 0.0220, 0.0213, 0.0201], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 21:57:53,330 INFO [train.py:904] (3/8) Epoch 30, batch 3650, loss[loss=0.1783, simple_loss=0.2547, pruned_loss=0.05094, over 16438.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.246, pruned_loss=0.03689, over 3314482.88 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:57:55,114 INFO [optim.py:368] (3/8) 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 21:58:14,210 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:58:32,713 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8472, 4.2697, 4.2077, 3.0314, 3.7481, 4.2386, 3.7415, 2.4541], device='cuda:3'), covar=tensor([0.0507, 0.0099, 0.0067, 0.0411, 0.0124, 0.0124, 0.0114, 0.0519], device='cuda:3'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 21:59:06,224 INFO [train.py:904] (3/8) Epoch 30, batch 3700, loss[loss=0.1778, simple_loss=0.25, pruned_loss=0.05274, over 16724.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2457, pruned_loss=0.03848, over 3295118.98 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:59:13,936 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 22:00:17,692 INFO [train.py:904] (3/8) Epoch 30, batch 3750, loss[loss=0.1772, simple_loss=0.2495, pruned_loss=0.05248, over 16468.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2464, pruned_loss=0.03957, over 3270910.50 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:19,717 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.048e+02 2.384e+02 2.845e+02 4.773e+02, threshold=4.768e+02, percent-clipped=0.0 2023-05-02 22:01:30,463 INFO [train.py:904] (3/8) Epoch 30, batch 3800, loss[loss=0.1843, simple_loss=0.2685, pruned_loss=0.05007, over 15562.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2477, pruned_loss=0.04085, over 3276016.76 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:22,049 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5996, 3.3224, 3.7954, 2.0012, 3.8309, 3.8460, 3.2108, 2.9110], device='cuda:3'), covar=tensor([0.0806, 0.0308, 0.0213, 0.1203, 0.0132, 0.0247, 0.0402, 0.0487], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0113, 0.0105, 0.0141, 0.0089, 0.0136, 0.0133, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 22:02:43,875 INFO [train.py:904] (3/8) Epoch 30, batch 3850, loss[loss=0.1747, simple_loss=0.245, pruned_loss=0.05221, over 16896.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.248, pruned_loss=0.04164, over 3283024.56 frames. ], batch size: 116, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:44,948 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.556e+02 2.341e+02 2.538e+02 2.759e+02 9.437e+02, threshold=5.075e+02, percent-clipped=1.0 2023-05-02 22:03:53,265 INFO [train.py:904] (3/8) Epoch 30, batch 3900, loss[loss=0.1592, simple_loss=0.2448, pruned_loss=0.0368, over 16762.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2477, pruned_loss=0.04188, over 3282940.01 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:03:54,944 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8102, 3.9642, 2.9284, 2.4116, 2.5102, 2.5885, 4.1076, 3.4390], device='cuda:3'), covar=tensor([0.2872, 0.0568, 0.1956, 0.3311, 0.3063, 0.2153, 0.0519, 0.1518], device='cuda:3'), in_proj_covar=tensor([0.0340, 0.0279, 0.0317, 0.0333, 0.0312, 0.0283, 0.0309, 0.0359], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 22:05:04,297 INFO [train.py:904] (3/8) Epoch 30, batch 3950, loss[loss=0.1597, simple_loss=0.2416, pruned_loss=0.03886, over 16469.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2477, pruned_loss=0.04261, over 3273410.77 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:05:05,534 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.189e+02 2.590e+02 3.423e+02 7.666e+02, threshold=5.180e+02, percent-clipped=4.0 2023-05-02 22:05:16,591 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 22:05:17,508 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:06:02,127 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1520, 4.1798, 4.4369, 4.4197, 4.4632, 4.1927, 4.2206, 4.1440], device='cuda:3'), covar=tensor([0.0438, 0.0677, 0.0444, 0.0417, 0.0554, 0.0515, 0.0779, 0.0687], device='cuda:3'), in_proj_covar=tensor([0.0456, 0.0520, 0.0495, 0.0458, 0.0541, 0.0523, 0.0603, 0.0422], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 22:06:15,412 INFO [train.py:904] (3/8) Epoch 30, batch 4000, loss[loss=0.1555, simple_loss=0.2393, pruned_loss=0.03583, over 16862.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2479, pruned_loss=0.04315, over 3280804.20 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:06:28,655 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 22:07:25,647 INFO [train.py:904] (3/8) Epoch 30, batch 4050, loss[loss=0.1728, simple_loss=0.2542, pruned_loss=0.04569, over 16329.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2489, pruned_loss=0.04238, over 3279120.92 frames. ], batch size: 35, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:07:27,597 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 1.991e+02 2.231e+02 2.617e+02 4.473e+02, threshold=4.462e+02, percent-clipped=0.0 2023-05-02 22:07:28,586 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4159, 4.7424, 4.9572, 4.8799, 4.9286, 4.6386, 4.3295, 4.4429], device='cuda:3'), covar=tensor([0.0562, 0.0733, 0.0524, 0.0620, 0.0678, 0.0622, 0.1367, 0.0628], device='cuda:3'), in_proj_covar=tensor([0.0455, 0.0519, 0.0493, 0.0457, 0.0540, 0.0522, 0.0601, 0.0421], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-02 22:07:36,118 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 22:07:37,010 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1809, 4.1326, 4.2641, 4.3876, 4.4837, 4.0998, 4.4100, 4.5449], device='cuda:3'), covar=tensor([0.1726, 0.1108, 0.1403, 0.0683, 0.0567, 0.1303, 0.1185, 0.0614], device='cuda:3'), in_proj_covar=tensor([0.0719, 0.0871, 0.1006, 0.0890, 0.0676, 0.0703, 0.0739, 0.0862], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:08:02,575 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7518, 1.8718, 1.6835, 1.4864, 1.9982, 1.6053, 1.5880, 1.9656], device='cuda:3'), covar=tensor([0.0195, 0.0315, 0.0447, 0.0409, 0.0215, 0.0304, 0.0158, 0.0214], device='cuda:3'), in_proj_covar=tensor([0.0241, 0.0250, 0.0240, 0.0241, 0.0251, 0.0249, 0.0250, 0.0252], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:08:22,499 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-05-02 22:08:37,711 INFO [train.py:904] (3/8) Epoch 30, batch 4100, loss[loss=0.186, simple_loss=0.2801, pruned_loss=0.046, over 16483.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2505, pruned_loss=0.04183, over 3269931.77 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:08:58,551 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-05-02 22:09:20,259 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:09:53,959 INFO [train.py:904] (3/8) Epoch 30, batch 4150, loss[loss=0.1746, simple_loss=0.2693, pruned_loss=0.03991, over 16868.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2571, pruned_loss=0.04395, over 3229296.76 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:56,054 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.028e+02 2.326e+02 2.808e+02 4.631e+02, threshold=4.653e+02, percent-clipped=1.0 2023-05-02 22:10:35,420 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:10:55,758 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298543.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:11:12,253 INFO [train.py:904] (3/8) Epoch 30, batch 4200, loss[loss=0.201, simple_loss=0.2927, pruned_loss=0.05462, over 16732.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2636, pruned_loss=0.04513, over 3192335.34 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:08,208 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:22,453 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3649, 4.2804, 4.4618, 4.6051, 4.7188, 4.3044, 4.6635, 4.7901], device='cuda:3'), covar=tensor([0.1901, 0.1375, 0.1441, 0.0702, 0.0585, 0.1184, 0.1102, 0.0602], device='cuda:3'), in_proj_covar=tensor([0.0708, 0.0859, 0.0991, 0.0876, 0.0667, 0.0694, 0.0728, 0.0848], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:12:26,984 INFO [train.py:904] (3/8) Epoch 30, batch 4250, loss[loss=0.1793, simple_loss=0.2738, pruned_loss=0.04233, over 16658.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2678, pruned_loss=0.04567, over 3166859.20 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:28,280 INFO [optim.py:368] (3/8) 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 22:12:40,802 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=298613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:13:04,718 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298629.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:13:31,990 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-05-02 22:13:41,911 INFO [train.py:904] (3/8) Epoch 30, batch 4300, loss[loss=0.1828, simple_loss=0.2754, pruned_loss=0.04509, over 16405.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.269, pruned_loss=0.04466, over 3178700.57 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:13:52,426 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=298661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:35,024 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:36,401 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298690.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:14:56,149 INFO [train.py:904] (3/8) Epoch 30, batch 4350, loss[loss=0.2082, simple_loss=0.2926, pruned_loss=0.06195, over 16992.00 frames. ], tot_loss[loss=0.182, simple_loss=0.272, pruned_loss=0.04594, over 3166109.23 frames. ], batch size: 55, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:14:57,381 INFO [optim.py:368] (3/8) 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 22:16:03,953 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298750.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:16:08,848 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.7743, 4.7573, 4.5342, 3.7523, 4.6811, 1.6240, 4.4337, 3.9988], device='cuda:3'), covar=tensor([0.0068, 0.0060, 0.0172, 0.0329, 0.0066, 0.3407, 0.0103, 0.0348], device='cuda:3'), in_proj_covar=tensor([0.0188, 0.0180, 0.0219, 0.0190, 0.0197, 0.0223, 0.0209, 0.0186], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:16:09,613 INFO [train.py:904] (3/8) Epoch 30, batch 4400, loss[loss=0.1799, simple_loss=0.2766, pruned_loss=0.04159, over 16853.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2745, pruned_loss=0.04747, over 3150762.59 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:21,871 INFO [train.py:904] (3/8) Epoch 30, batch 4450, loss[loss=0.1896, simple_loss=0.2871, pruned_loss=0.04604, over 16707.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2778, pruned_loss=0.04884, over 3168378.73 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:23,559 INFO [optim.py:368] (3/8) 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 22:18:09,765 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298838.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:18:32,516 INFO [train.py:904] (3/8) Epoch 30, batch 4500, loss[loss=0.1772, simple_loss=0.2706, pruned_loss=0.04188, over 16861.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.278, pruned_loss=0.04914, over 3180327.15 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:19,735 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:19:44,880 INFO [train.py:904] (3/8) Epoch 30, batch 4550, loss[loss=0.2169, simple_loss=0.303, pruned_loss=0.06541, over 16589.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2793, pruned_loss=0.05042, over 3189244.12 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:46,102 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.782e+02 2.097e+02 2.367e+02 4.731e+02, threshold=4.194e+02, percent-clipped=0.0 2023-05-02 22:20:57,285 INFO [train.py:904] (3/8) Epoch 30, batch 4600, loss[loss=0.1929, simple_loss=0.2812, pruned_loss=0.0523, over 16572.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2804, pruned_loss=0.05066, over 3201315.31 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:21:11,708 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 22:21:42,882 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298985.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:22:09,580 INFO [train.py:904] (3/8) Epoch 30, batch 4650, loss[loss=0.1944, simple_loss=0.2725, pruned_loss=0.05816, over 16796.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2802, pruned_loss=0.05129, over 3195195.54 frames. ], batch size: 39, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:22:10,878 INFO [optim.py:368] (3/8) 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 22:22:11,253 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2516, 4.3690, 4.1310, 3.7891, 3.8338, 4.2469, 3.8718, 3.9985], device='cuda:3'), covar=tensor([0.0528, 0.0347, 0.0253, 0.0273, 0.0606, 0.0328, 0.0910, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0321, 0.0484, 0.0377, 0.0379, 0.0373, 0.0434, 0.0258, 0.0448], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:23:09,773 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299045.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:23:22,503 INFO [train.py:904] (3/8) Epoch 30, batch 4700, loss[loss=0.1783, simple_loss=0.2624, pruned_loss=0.04709, over 11565.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2772, pruned_loss=0.05025, over 3206156.79 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:23:56,587 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 22:24:36,509 INFO [train.py:904] (3/8) Epoch 30, batch 4750, loss[loss=0.1647, simple_loss=0.2612, pruned_loss=0.03411, over 16809.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2738, pruned_loss=0.04847, over 3194213.93 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:24:37,718 INFO [optim.py:368] (3/8) 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 22:25:07,727 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:25,727 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:49,336 INFO [train.py:904] (3/8) Epoch 30, batch 4800, loss[loss=0.1569, simple_loss=0.2484, pruned_loss=0.03263, over 16499.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2696, pruned_loss=0.04618, over 3201942.41 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:26:37,232 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:37,336 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:38,538 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299187.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:45,005 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299191.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:27:04,368 INFO [train.py:904] (3/8) Epoch 30, batch 4850, loss[loss=0.1891, simple_loss=0.2845, pruned_loss=0.0468, over 16712.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2706, pruned_loss=0.04548, over 3184482.08 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:27:06,459 INFO [optim.py:368] (3/8) 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 22:27:49,335 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:02,974 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299243.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:16,452 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 4900, loss[loss=0.1958, simple_loss=0.2862, pruned_loss=0.05267, over 16673.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2694, pruned_loss=0.04403, over 3191825.24 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:28:48,104 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5594, 3.7631, 3.9309, 2.5026, 3.2053, 2.6382, 3.9082, 4.0995], device='cuda:3'), covar=tensor([0.0257, 0.0774, 0.0618, 0.2002, 0.0878, 0.0891, 0.0583, 0.0895], device='cuda:3'), in_proj_covar=tensor([0.0163, 0.0173, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0187], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 22:29:06,065 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299285.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:29:33,775 INFO [train.py:904] (3/8) Epoch 30, batch 4950, loss[loss=0.1788, simple_loss=0.2719, pruned_loss=0.04283, over 16635.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2683, pruned_loss=0.04306, over 3211238.23 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:29:34,224 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 22:30:02,799 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:17,899 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299333.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:30:36,091 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 5000, loss[loss=0.1576, simple_loss=0.2587, pruned_loss=0.02824, over 16478.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2696, pruned_loss=0.04291, over 3226962.79 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:31:33,580 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:31:46,452 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:31:46,658 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6006, 3.6093, 4.2691, 1.9600, 4.4569, 4.4821, 3.1733, 3.2856], device='cuda:3'), covar=tensor([0.0871, 0.0312, 0.0173, 0.1268, 0.0064, 0.0115, 0.0454, 0.0469], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0140, 0.0088, 0.0134, 0.0133, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 22:32:01,705 INFO [train.py:904] (3/8) Epoch 30, batch 5050, loss[loss=0.1879, simple_loss=0.2779, pruned_loss=0.04894, over 12329.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2701, pruned_loss=0.04287, over 3221958.81 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:32:02,875 INFO [optim.py:368] (3/8) 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 22:33:13,772 INFO [train.py:904] (3/8) Epoch 30, batch 5100, loss[loss=0.1689, simple_loss=0.2562, pruned_loss=0.04083, over 15276.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2685, pruned_loss=0.04208, over 3223675.62 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:33:47,559 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 22:33:57,989 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:34:01,076 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.2422, 2.3047, 2.3604, 3.9579, 2.0961, 2.5996, 2.3496, 2.3900], device='cuda:3'), covar=tensor([0.1803, 0.4375, 0.3504, 0.0747, 0.5494, 0.3191, 0.4514, 0.4149], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0483, 0.0392, 0.0344, 0.0449, 0.0555, 0.0456, 0.0566], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:34:30,232 INFO [train.py:904] (3/8) Epoch 30, batch 5150, loss[loss=0.1707, simple_loss=0.2729, pruned_loss=0.0342, over 16839.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2679, pruned_loss=0.04144, over 3207127.83 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:34:31,340 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 1.896e+02 2.234e+02 2.667e+02 3.487e+02, threshold=4.469e+02, percent-clipped=0.0 2023-05-02 22:35:19,375 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.3003, 5.5490, 5.2909, 5.3499, 5.0719, 5.0466, 4.9245, 5.6760], device='cuda:3'), covar=tensor([0.1248, 0.0845, 0.1005, 0.0862, 0.0851, 0.0715, 0.1181, 0.0771], device='cuda:3'), in_proj_covar=tensor([0.0724, 0.0875, 0.0718, 0.0680, 0.0560, 0.0555, 0.0734, 0.0688], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:35:19,469 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8899, 4.8567, 4.7680, 3.9595, 4.7820, 2.0172, 4.4924, 4.4789], device='cuda:3'), covar=tensor([0.0111, 0.0122, 0.0188, 0.0520, 0.0120, 0.2715, 0.0172, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0189, 0.0195, 0.0221, 0.0206, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:35:33,156 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 5200, loss[loss=0.1706, simple_loss=0.2531, pruned_loss=0.04408, over 16520.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2664, pruned_loss=0.04089, over 3214253.08 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:07,927 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 22:36:48,431 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299599.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:36:55,727 INFO [train.py:904] (3/8) Epoch 30, batch 5250, loss[loss=0.1626, simple_loss=0.2575, pruned_loss=0.03381, over 16913.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2647, pruned_loss=0.04082, over 3214043.42 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:56,956 INFO [optim.py:368] (3/8) 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 22:37:12,569 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 22:37:27,240 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0745, 4.2993, 4.1290, 4.1709, 3.8800, 3.8739, 3.9149, 4.2827], device='cuda:3'), covar=tensor([0.1125, 0.0831, 0.0932, 0.0789, 0.0786, 0.1967, 0.0986, 0.1037], device='cuda:3'), in_proj_covar=tensor([0.0724, 0.0876, 0.0718, 0.0681, 0.0560, 0.0555, 0.0735, 0.0689], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:38:06,894 INFO [train.py:904] (3/8) Epoch 30, batch 5300, loss[loss=0.1519, simple_loss=0.2404, pruned_loss=0.03165, over 16312.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2612, pruned_loss=0.03989, over 3209026.20 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:38:42,816 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:39:18,175 INFO [train.py:904] (3/8) Epoch 30, batch 5350, loss[loss=0.1601, simple_loss=0.2653, pruned_loss=0.02748, over 16850.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2593, pruned_loss=0.03913, over 3215417.21 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:39:19,344 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.953e+02 2.222e+02 2.592e+02 5.137e+02, threshold=4.444e+02, percent-clipped=1.0 2023-05-02 22:40:29,705 INFO [train.py:904] (3/8) Epoch 30, batch 5400, loss[loss=0.1878, simple_loss=0.2747, pruned_loss=0.05044, over 16587.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2621, pruned_loss=0.03982, over 3210605.53 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:40:51,588 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299769.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:10,032 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:34,579 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299798.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:41:46,657 INFO [train.py:904] (3/8) Epoch 30, batch 5450, loss[loss=0.2196, simple_loss=0.3044, pruned_loss=0.06745, over 16399.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2655, pruned_loss=0.04124, over 3208374.04 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:41:47,796 INFO [optim.py:368] (3/8) 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 22:42:27,031 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:27,217 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:51,971 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299847.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:43:02,915 INFO [train.py:904] (3/8) Epoch 30, batch 5500, loss[loss=0.2045, simple_loss=0.2936, pruned_loss=0.05771, over 16825.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2725, pruned_loss=0.04539, over 3174868.91 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:43:11,679 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299859.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:44:09,162 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299895.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:10,687 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1002, 3.3317, 3.5540, 2.0244, 3.0897, 2.3426, 3.4977, 3.7291], device='cuda:3'), covar=tensor([0.0280, 0.0825, 0.0601, 0.2222, 0.0841, 0.1060, 0.0637, 0.0874], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 22:44:14,724 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 5550, loss[loss=0.2596, simple_loss=0.321, pruned_loss=0.09908, over 11303.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2787, pruned_loss=0.04922, over 3160602.75 frames. ], batch size: 250, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:44:23,799 INFO [optim.py:368] (3/8) 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 22:44:48,706 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.2337, 2.0698, 1.8079, 1.8104, 2.3176, 1.9586, 1.9522, 2.3793], device='cuda:3'), covar=tensor([0.0288, 0.0415, 0.0563, 0.0509, 0.0277, 0.0395, 0.0242, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0245, 0.0235, 0.0235, 0.0245, 0.0243, 0.0243, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:44:52,420 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1237, 3.3785, 3.4994, 2.0235, 3.0673, 2.3877, 3.5187, 3.7615], device='cuda:3'), covar=tensor([0.0263, 0.0805, 0.0615, 0.2278, 0.0841, 0.0988, 0.0645, 0.0894], device='cuda:3'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0185], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 22:45:11,558 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 22:45:19,192 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4878, 3.4545, 3.4537, 2.6960, 3.2978, 2.1995, 3.1633, 2.7512], device='cuda:3'), covar=tensor([0.0186, 0.0162, 0.0214, 0.0276, 0.0134, 0.2436, 0.0155, 0.0305], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0189, 0.0194, 0.0221, 0.0206, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:45:32,583 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 5600, loss[loss=0.1909, simple_loss=0.2863, pruned_loss=0.04778, over 16733.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2836, pruned_loss=0.05357, over 3099241.34 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:45:43,606 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1186, 2.4235, 2.5195, 1.9581, 2.6889, 2.7498, 2.4431, 2.3754], device='cuda:3'), covar=tensor([0.0655, 0.0287, 0.0232, 0.0901, 0.0143, 0.0300, 0.0454, 0.0443], device='cuda:3'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0140, 0.0088, 0.0134, 0.0132, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 22:46:27,176 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:47:11,368 INFO [train.py:904] (3/8) Epoch 30, batch 5650, loss[loss=0.2628, simple_loss=0.3263, pruned_loss=0.09963, over 11000.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2878, pruned_loss=0.05678, over 3076831.03 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:47:13,271 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:48:02,792 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.2301, 4.2301, 4.1296, 3.2882, 4.1842, 1.6336, 3.9649, 3.6280], device='cuda:3'), covar=tensor([0.0128, 0.0119, 0.0190, 0.0326, 0.0104, 0.3260, 0.0141, 0.0342], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0177, 0.0216, 0.0188, 0.0193, 0.0220, 0.0205, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:48:24,013 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 22:48:27,593 INFO [train.py:904] (3/8) Epoch 30, batch 5700, loss[loss=0.2126, simple_loss=0.3164, pruned_loss=0.05443, over 16821.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2899, pruned_loss=0.05886, over 3057621.44 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:48:57,337 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:49:13,088 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 22:49:45,341 INFO [train.py:904] (3/8) Epoch 30, batch 5750, loss[loss=0.245, simple_loss=0.3097, pruned_loss=0.09016, over 11287.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2925, pruned_loss=0.06065, over 3032689.14 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:49:49,226 INFO [optim.py:368] (3/8) 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 22:50:18,681 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300125.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:50:35,581 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 5800, loss[loss=0.1772, simple_loss=0.2727, pruned_loss=0.04083, over 16444.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2917, pruned_loss=0.05928, over 3038333.71 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:51:07,774 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300154.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:52:20,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5161, 4.0499, 4.0578, 2.7282, 3.6631, 4.1349, 3.6155, 2.1778], device='cuda:3'), covar=tensor([0.0567, 0.0071, 0.0067, 0.0433, 0.0111, 0.0112, 0.0110, 0.0509], device='cuda:3'), in_proj_covar=tensor([0.0140, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:3') 2023-05-02 22:52:25,620 INFO [train.py:904] (3/8) Epoch 30, batch 5850, loss[loss=0.1988, simple_loss=0.2842, pruned_loss=0.05667, over 16731.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2899, pruned_loss=0.05814, over 3029039.38 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:52:28,949 INFO [optim.py:368] (3/8) 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 22:52:34,780 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 22:52:54,597 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0291, 2.3731, 2.3680, 2.7423, 1.9383, 3.1984, 1.8975, 2.7297], device='cuda:3'), covar=tensor([0.1124, 0.0631, 0.1055, 0.0194, 0.0126, 0.0339, 0.1449, 0.0716], device='cuda:3'), in_proj_covar=tensor([0.0173, 0.0182, 0.0201, 0.0207, 0.0207, 0.0218, 0.0212, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 22:53:48,213 INFO [train.py:904] (3/8) Epoch 30, batch 5900, loss[loss=0.1959, simple_loss=0.278, pruned_loss=0.05686, over 15393.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2894, pruned_loss=0.0579, over 3042362.84 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:54:10,405 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8680, 2.7422, 2.8510, 2.2230, 2.7096, 2.2153, 2.7396, 2.8929], device='cuda:3'), covar=tensor([0.0272, 0.0775, 0.0526, 0.1729, 0.0781, 0.0898, 0.0542, 0.0699], device='cuda:3'), in_proj_covar=tensor([0.0161, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0145, 0.0184], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:3') 2023-05-02 22:54:13,865 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0355, 4.1104, 3.9399, 3.6457, 3.6319, 4.0377, 3.7355, 3.8291], device='cuda:3'), covar=tensor([0.0637, 0.0671, 0.0346, 0.0332, 0.0788, 0.0499, 0.1257, 0.0616], device='cuda:3'), in_proj_covar=tensor([0.0316, 0.0481, 0.0372, 0.0373, 0.0368, 0.0430, 0.0255, 0.0442], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:54:36,278 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 22:55:09,236 INFO [train.py:904] (3/8) Epoch 30, batch 5950, loss[loss=0.1879, simple_loss=0.2817, pruned_loss=0.04702, over 16697.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2898, pruned_loss=0.05657, over 3039321.83 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:55:12,894 INFO [optim.py:368] (3/8) 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 22:55:43,442 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0557, 4.0353, 3.9706, 3.0805, 3.9757, 1.8197, 3.7858, 3.4025], device='cuda:3'), covar=tensor([0.0158, 0.0125, 0.0190, 0.0306, 0.0111, 0.3132, 0.0158, 0.0343], device='cuda:3'), in_proj_covar=tensor([0.0186, 0.0177, 0.0216, 0.0188, 0.0193, 0.0220, 0.0205, 0.0182], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 22:56:29,061 INFO [train.py:904] (3/8) Epoch 30, batch 6000, loss[loss=0.2043, simple_loss=0.2913, pruned_loss=0.05863, over 16673.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2887, pruned_loss=0.05574, over 3062758.50 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:56:29,062 INFO [train.py:929] (3/8) Computing validation loss 2023-05-02 22:56:39,791 INFO [train.py:938] (3/8) Epoch 30, validation: loss=0.1471, simple_loss=0.2591, pruned_loss=0.01755, over 944034.00 frames. 2023-05-02 22:56:39,791 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-02 22:56:51,314 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4160, 2.9812, 2.5837, 2.2694, 2.2944, 2.2819, 3.0319, 2.8828], device='cuda:3'), covar=tensor([0.3034, 0.0889, 0.2080, 0.2818, 0.2711, 0.2451, 0.0733, 0.1543], device='cuda:3'), in_proj_covar=tensor([0.0337, 0.0275, 0.0314, 0.0329, 0.0307, 0.0280, 0.0305, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 22:57:05,046 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-05-02 22:57:57,806 INFO [train.py:904] (3/8) Epoch 30, batch 6050, loss[loss=0.2038, simple_loss=0.2933, pruned_loss=0.05719, over 16299.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2871, pruned_loss=0.05478, over 3075771.55 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:58:01,294 INFO [optim.py:368] (3/8) 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 22:58:29,284 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:58:35,426 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:59:15,660 INFO [train.py:904] (3/8) Epoch 30, batch 6100, loss[loss=0.2107, simple_loss=0.3073, pruned_loss=0.05702, over 16345.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2872, pruned_loss=0.05399, over 3086491.16 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:59:16,246 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300454.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:59:35,197 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-05-02 22:59:47,467 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:00:32,804 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300502.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:00:34,974 INFO [train.py:904] (3/8) Epoch 30, batch 6150, loss[loss=0.201, simple_loss=0.2834, pruned_loss=0.05932, over 11601.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2855, pruned_loss=0.05367, over 3086471.06 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:00:38,227 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 6200, loss[loss=0.1778, simple_loss=0.2695, pruned_loss=0.04304, over 16521.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2839, pruned_loss=0.05327, over 3095749.55 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:13,157 INFO [train.py:904] (3/8) Epoch 30, batch 6250, loss[loss=0.212, simple_loss=0.2866, pruned_loss=0.06877, over 11860.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2838, pruned_loss=0.05321, over 3095859.20 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:16,359 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 2.845e+02 3.157e+02 4.016e+02 6.897e+02, threshold=6.314e+02, percent-clipped=1.0 2023-05-02 23:03:29,106 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1448, 2.1772, 2.6696, 3.1382, 2.9542, 3.5148, 2.2611, 3.5431], device='cuda:3'), covar=tensor([0.0220, 0.0585, 0.0382, 0.0344, 0.0362, 0.0202, 0.0609, 0.0148], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0199, 0.0188, 0.0194, 0.0211, 0.0168, 0.0205, 0.0168], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:03:34,762 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:04:04,753 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-02 23:04:31,361 INFO [train.py:904] (3/8) Epoch 30, batch 6300, loss[loss=0.1938, simple_loss=0.2829, pruned_loss=0.05241, over 16711.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2833, pruned_loss=0.05259, over 3098326.88 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:11,467 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300678.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:05:48,970 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0079, 4.0718, 4.3176, 4.2886, 4.3092, 4.0728, 4.0732, 4.0369], device='cuda:3'), covar=tensor([0.0407, 0.0767, 0.0529, 0.0502, 0.0535, 0.0579, 0.0873, 0.0654], device='cuda:3'), in_proj_covar=tensor([0.0442, 0.0505, 0.0482, 0.0447, 0.0528, 0.0511, 0.0587, 0.0412], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 23:05:50,440 INFO [train.py:904] (3/8) Epoch 30, batch 6350, loss[loss=0.2034, simple_loss=0.2909, pruned_loss=0.058, over 16150.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2842, pruned_loss=0.05405, over 3077105.06 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:53,981 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.820e+02 3.153e+02 4.098e+02 7.645e+02, threshold=6.307e+02, percent-clipped=4.0 2023-05-02 23:06:29,216 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:06:34,904 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 23:07:07,467 INFO [train.py:904] (3/8) Epoch 30, batch 6400, loss[loss=0.1619, simple_loss=0.2489, pruned_loss=0.03743, over 16497.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2843, pruned_loss=0.05515, over 3093873.26 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:07:13,870 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 23:07:42,791 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300777.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:08:00,255 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 23:08:19,373 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 6450, loss[loss=0.184, simple_loss=0.2757, pruned_loss=0.04612, over 17054.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2839, pruned_loss=0.05448, over 3092472.13 frames. ], batch size: 53, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:08:24,279 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.836e+02 3.339e+02 4.246e+02 8.060e+02, threshold=6.678e+02, percent-clipped=2.0 2023-05-02 23:08:57,298 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 23:08:58,763 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5422, 4.6348, 4.9156, 4.8726, 4.8990, 4.6077, 4.5786, 4.4122], device='cuda:3'), covar=tensor([0.0342, 0.0578, 0.0360, 0.0382, 0.0475, 0.0396, 0.0916, 0.0617], device='cuda:3'), in_proj_covar=tensor([0.0445, 0.0508, 0.0485, 0.0450, 0.0531, 0.0514, 0.0590, 0.0414], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 23:09:38,127 INFO [train.py:904] (3/8) Epoch 30, batch 6500, loss[loss=0.194, simple_loss=0.2791, pruned_loss=0.05446, over 16596.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2822, pruned_loss=0.05401, over 3091734.82 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:09:40,791 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.5930, 4.5685, 4.4238, 3.6351, 4.5184, 1.8001, 4.2450, 3.9082], device='cuda:3'), covar=tensor([0.0130, 0.0120, 0.0211, 0.0361, 0.0109, 0.3088, 0.0152, 0.0371], device='cuda:3'), in_proj_covar=tensor([0.0187, 0.0178, 0.0217, 0.0189, 0.0195, 0.0221, 0.0206, 0.0183], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:09:52,598 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300863.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:10:44,195 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.8727, 1.3985, 1.7517, 1.7224, 1.8391, 1.9378, 1.6662, 1.8531], device='cuda:3'), covar=tensor([0.0284, 0.0457, 0.0246, 0.0324, 0.0304, 0.0219, 0.0465, 0.0178], device='cuda:3'), in_proj_covar=tensor([0.0200, 0.0200, 0.0189, 0.0195, 0.0212, 0.0169, 0.0206, 0.0169], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 23:10:58,874 INFO [train.py:904] (3/8) Epoch 30, batch 6550, loss[loss=0.2154, simple_loss=0.3157, pruned_loss=0.05754, over 16286.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2847, pruned_loss=0.05505, over 3079305.66 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:11:01,654 INFO [optim.py:368] (3/8) 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 23:11:52,682 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 23:12:19,421 INFO [train.py:904] (3/8) Epoch 30, batch 6600, loss[loss=0.2311, simple_loss=0.2999, pruned_loss=0.08117, over 11271.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2867, pruned_loss=0.05508, over 3081094.26 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:12:48,783 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:13:38,992 INFO [train.py:904] (3/8) Epoch 30, batch 6650, loss[loss=0.1917, simple_loss=0.2806, pruned_loss=0.05139, over 16925.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2874, pruned_loss=0.05592, over 3072785.99 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:13:43,930 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.674e+02 3.483e+02 3.945e+02 7.489e+02, threshold=6.965e+02, percent-clipped=2.0 2023-05-02 23:14:55,114 INFO [train.py:904] (3/8) Epoch 30, batch 6700, loss[loss=0.1978, simple_loss=0.2806, pruned_loss=0.05752, over 16635.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2864, pruned_loss=0.0563, over 3079656.38 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:15:08,380 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8446, 4.8875, 5.2347, 5.1861, 5.2326, 4.8993, 4.8390, 4.6738], device='cuda:3'), covar=tensor([0.0350, 0.0530, 0.0332, 0.0370, 0.0472, 0.0383, 0.0992, 0.0573], device='cuda:3'), in_proj_covar=tensor([0.0444, 0.0507, 0.0484, 0.0449, 0.0531, 0.0513, 0.0589, 0.0413], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 23:16:11,407 INFO [train.py:904] (3/8) Epoch 30, batch 6750, loss[loss=0.1922, simple_loss=0.274, pruned_loss=0.05523, over 16708.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2846, pruned_loss=0.05593, over 3093607.64 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:15,728 INFO [optim.py:368] (3/8) 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 23:16:39,746 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3636, 4.1741, 4.3620, 4.5225, 4.6705, 4.2949, 4.6552, 4.6709], device='cuda:3'), covar=tensor([0.1851, 0.1422, 0.1697, 0.0824, 0.0681, 0.1230, 0.0827, 0.0833], device='cuda:3'), in_proj_covar=tensor([0.0678, 0.0822, 0.0949, 0.0842, 0.0637, 0.0661, 0.0703, 0.0812], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:16:44,347 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 23:17:07,602 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:17:29,021 INFO [train.py:904] (3/8) Epoch 30, batch 6800, loss[loss=0.1814, simple_loss=0.2821, pruned_loss=0.04037, over 17001.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2846, pruned_loss=0.0559, over 3089339.19 frames. ], batch size: 41, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:17:30,593 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5743, 3.3915, 3.9571, 1.9989, 4.1493, 4.1712, 3.0276, 3.0309], device='cuda:3'), covar=tensor([0.0863, 0.0345, 0.0227, 0.1208, 0.0082, 0.0190, 0.0493, 0.0494], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0112, 0.0104, 0.0141, 0.0089, 0.0135, 0.0133, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 23:17:35,471 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301158.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:18:43,934 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:18:47,466 INFO [train.py:904] (3/8) Epoch 30, batch 6850, loss[loss=0.1865, simple_loss=0.2927, pruned_loss=0.04014, over 17062.00 frames. ], tot_loss[loss=0.199, simple_loss=0.286, pruned_loss=0.05596, over 3106588.00 frames. ], batch size: 55, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:18:51,707 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.931e+02 2.806e+02 3.426e+02 4.107e+02 8.215e+02, threshold=6.851e+02, percent-clipped=6.0 2023-05-02 23:18:54,450 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 23:19:54,278 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-05-02 23:20:04,219 INFO [train.py:904] (3/8) Epoch 30, batch 6900, loss[loss=0.1972, simple_loss=0.2853, pruned_loss=0.05453, over 16713.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2884, pruned_loss=0.05553, over 3111021.60 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:20:35,512 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:21:11,406 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.3795, 4.2478, 4.4391, 4.5775, 4.7190, 4.3117, 4.6785, 4.7411], device='cuda:3'), covar=tensor([0.2040, 0.1300, 0.1553, 0.0785, 0.0659, 0.1149, 0.0828, 0.0779], device='cuda:3'), in_proj_covar=tensor([0.0678, 0.0824, 0.0952, 0.0843, 0.0640, 0.0663, 0.0704, 0.0814], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:21:22,699 INFO [train.py:904] (3/8) Epoch 30, batch 6950, loss[loss=0.1923, simple_loss=0.2854, pruned_loss=0.0496, over 16724.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.29, pruned_loss=0.05707, over 3098815.85 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:21:26,935 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.877e+02 3.278e+02 4.034e+02 6.348e+02, threshold=6.555e+02, percent-clipped=0.0 2023-05-02 23:21:50,651 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:21:50,899 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0453, 2.9596, 2.6564, 2.7743, 3.2866, 2.9143, 3.4749, 3.5421], device='cuda:3'), covar=tensor([0.0112, 0.0480, 0.0567, 0.0502, 0.0310, 0.0454, 0.0333, 0.0283], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0244, 0.0235, 0.0236, 0.0246, 0.0242, 0.0242, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:22:38,930 INFO [train.py:904] (3/8) Epoch 30, batch 7000, loss[loss=0.2086, simple_loss=0.2979, pruned_loss=0.0597, over 16767.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2899, pruned_loss=0.05648, over 3099931.44 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:23:55,571 INFO [train.py:904] (3/8) Epoch 30, batch 7050, loss[loss=0.1904, simple_loss=0.2771, pruned_loss=0.05179, over 17033.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2907, pruned_loss=0.05685, over 3064793.89 frames. ], batch size: 50, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:24:00,750 INFO [optim.py:368] (3/8) 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 23:24:38,166 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:24:41,271 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.6989, 1.8203, 1.6687, 1.5333, 1.9446, 1.6172, 1.5534, 1.8838], device='cuda:3'), covar=tensor([0.0240, 0.0300, 0.0445, 0.0376, 0.0241, 0.0295, 0.0193, 0.0245], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0245, 0.0235, 0.0236, 0.0247, 0.0243, 0.0242, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:25:14,075 INFO [train.py:904] (3/8) Epoch 30, batch 7100, loss[loss=0.1914, simple_loss=0.2809, pruned_loss=0.05091, over 16837.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.289, pruned_loss=0.05686, over 3052907.23 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:25:21,056 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301458.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:25:36,283 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-02 23:26:14,386 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301491.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:21,066 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:24,012 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 23:26:33,762 INFO [train.py:904] (3/8) Epoch 30, batch 7150, loss[loss=0.224, simple_loss=0.3064, pruned_loss=0.07076, over 15454.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.287, pruned_loss=0.05605, over 3065647.21 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:26:36,532 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301506.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:37,391 INFO [optim.py:368] (3/8) 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 23:27:35,591 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4182, 4.4872, 4.2912, 3.9792, 3.9644, 4.4216, 4.1339, 4.1111], device='cuda:3'), covar=tensor([0.0627, 0.0588, 0.0342, 0.0335, 0.0938, 0.0501, 0.0609, 0.0660], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0472, 0.0363, 0.0365, 0.0361, 0.0420, 0.0250, 0.0433], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:27:49,037 INFO [train.py:904] (3/8) Epoch 30, batch 7200, loss[loss=0.1835, simple_loss=0.2768, pruned_loss=0.04508, over 15304.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2849, pruned_loss=0.05428, over 3065526.75 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:28:14,135 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-02 23:28:36,123 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-02 23:28:39,950 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1603, 2.0232, 1.7364, 1.7451, 2.2565, 1.8992, 1.9131, 2.3620], device='cuda:3'), covar=tensor([0.0322, 0.0515, 0.0679, 0.0621, 0.0327, 0.0493, 0.0234, 0.0319], device='cuda:3'), in_proj_covar=tensor([0.0234, 0.0244, 0.0235, 0.0236, 0.0247, 0.0242, 0.0242, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:29:10,980 INFO [train.py:904] (3/8) Epoch 30, batch 7250, loss[loss=0.1796, simple_loss=0.2671, pruned_loss=0.046, over 16693.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2827, pruned_loss=0.05318, over 3068069.85 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:15,148 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.328e+02 2.684e+02 3.448e+02 5.554e+02, threshold=5.368e+02, percent-clipped=0.0 2023-05-02 23:30:27,541 INFO [train.py:904] (3/8) Epoch 30, batch 7300, loss[loss=0.1971, simple_loss=0.2875, pruned_loss=0.05332, over 16908.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.282, pruned_loss=0.05334, over 3066377.93 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:31:22,694 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.6670, 2.7773, 2.5038, 2.6090, 3.0473, 2.6398, 3.0949, 3.2360], device='cuda:3'), covar=tensor([0.0115, 0.0406, 0.0489, 0.0423, 0.0259, 0.0414, 0.0256, 0.0278], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0243, 0.0234, 0.0235, 0.0245, 0.0241, 0.0241, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:31:44,853 INFO [train.py:904] (3/8) Epoch 30, batch 7350, loss[loss=0.187, simple_loss=0.2755, pruned_loss=0.04923, over 16376.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2833, pruned_loss=0.05437, over 3044820.69 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:31:50,961 INFO [optim.py:368] (3/8) 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 23:32:38,309 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7848, 3.8256, 3.9338, 3.6873, 3.8686, 4.2542, 3.9108, 3.6115], device='cuda:3'), covar=tensor([0.2189, 0.2258, 0.2441, 0.2463, 0.2505, 0.1692, 0.1806, 0.2844], device='cuda:3'), in_proj_covar=tensor([0.0433, 0.0642, 0.0715, 0.0524, 0.0693, 0.0733, 0.0553, 0.0698], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-02 23:32:39,775 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.9914, 2.9711, 2.6211, 2.9130, 3.3380, 2.9990, 3.4864, 3.5483], device='cuda:3'), covar=tensor([0.0134, 0.0512, 0.0582, 0.0440, 0.0285, 0.0415, 0.0317, 0.0288], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0244, 0.0235, 0.0235, 0.0246, 0.0242, 0.0241, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:32:41,165 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.3773, 2.9318, 2.7187, 2.3007, 2.3226, 2.3536, 2.9278, 2.8332], device='cuda:3'), covar=tensor([0.2348, 0.0706, 0.1507, 0.2610, 0.2290, 0.2131, 0.0457, 0.1348], device='cuda:3'), in_proj_covar=tensor([0.0338, 0.0276, 0.0316, 0.0330, 0.0307, 0.0281, 0.0305, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 23:32:44,228 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 7400, loss[loss=0.2076, simple_loss=0.2871, pruned_loss=0.06407, over 15314.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2845, pruned_loss=0.05526, over 3030820.81 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:33:05,167 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5487, 3.7838, 2.7731, 2.2790, 2.5422, 2.4579, 4.0473, 3.3952], device='cuda:3'), covar=tensor([0.3187, 0.0637, 0.1994, 0.2872, 0.2556, 0.2299, 0.0484, 0.1347], device='cuda:3'), in_proj_covar=tensor([0.0339, 0.0276, 0.0316, 0.0330, 0.0307, 0.0281, 0.0305, 0.0353], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:3') 2023-05-02 23:33:15,844 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 23:33:42,301 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1816, 5.2100, 5.0289, 4.6436, 4.6682, 5.1172, 4.9914, 4.8240], device='cuda:3'), covar=tensor([0.0586, 0.0494, 0.0331, 0.0326, 0.0992, 0.0461, 0.0323, 0.0654], device='cuda:3'), in_proj_covar=tensor([0.0310, 0.0471, 0.0363, 0.0365, 0.0360, 0.0420, 0.0250, 0.0432], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:33:54,212 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:01,192 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0426, 2.5106, 2.0456, 2.2881, 2.8328, 2.4725, 2.6553, 2.9496], device='cuda:3'), covar=tensor([0.0244, 0.0450, 0.0634, 0.0490, 0.0289, 0.0422, 0.0271, 0.0300], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0244, 0.0235, 0.0235, 0.0246, 0.0242, 0.0241, 0.0244], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:34:02,910 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 23:34:10,887 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:23,455 INFO [train.py:904] (3/8) Epoch 30, batch 7450, loss[loss=0.1895, simple_loss=0.2894, pruned_loss=0.04486, over 15455.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2858, pruned_loss=0.05594, over 3044978.02 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:34:24,186 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3925, 3.3548, 3.4115, 3.4911, 3.5135, 3.3181, 3.5036, 3.5631], device='cuda:3'), covar=tensor([0.1210, 0.0910, 0.1005, 0.0652, 0.0696, 0.2407, 0.1224, 0.0925], device='cuda:3'), in_proj_covar=tensor([0.0672, 0.0814, 0.0943, 0.0837, 0.0633, 0.0655, 0.0697, 0.0805], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:34:24,241 INFO [zipformer.py:625] (3/8) 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] (3/8) 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 23:35:30,253 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:35:45,203 INFO [train.py:904] (3/8) Epoch 30, batch 7500, loss[loss=0.1997, simple_loss=0.2856, pruned_loss=0.05691, over 16580.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2855, pruned_loss=0.05545, over 3044046.42 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:36:57,857 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:05,267 INFO [train.py:904] (3/8) Epoch 30, batch 7550, loss[loss=0.2101, simple_loss=0.2815, pruned_loss=0.06939, over 11362.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2854, pruned_loss=0.05616, over 3022837.64 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:37:07,777 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.823e+02 2.767e+02 3.333e+02 4.100e+02 7.380e+02, threshold=6.666e+02, percent-clipped=1.0 2023-05-02 23:37:46,223 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 23:38:23,954 INFO [train.py:904] (3/8) Epoch 30, batch 7600, loss[loss=0.1948, simple_loss=0.2801, pruned_loss=0.05479, over 17265.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2845, pruned_loss=0.05579, over 3033807.19 frames. ], batch size: 52, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:38:34,197 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301960.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:38:43,721 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 7650, loss[loss=0.1819, simple_loss=0.2735, pruned_loss=0.04517, over 16802.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2852, pruned_loss=0.0563, over 3044362.55 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:39:53,135 INFO [optim.py:368] (3/8) 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 23:39:58,771 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6636, 3.3916, 3.8788, 1.9045, 4.0296, 4.1058, 3.0550, 3.0101], device='cuda:3'), covar=tensor([0.0783, 0.0326, 0.0252, 0.1313, 0.0092, 0.0178, 0.0471, 0.0496], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0112, 0.0104, 0.0140, 0.0088, 0.0135, 0.0133, 0.0133], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 23:41:06,721 INFO [train.py:904] (3/8) Epoch 30, batch 7700, loss[loss=0.2384, simple_loss=0.3087, pruned_loss=0.08408, over 11558.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2854, pruned_loss=0.05718, over 3016234.34 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:41:57,540 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:18,119 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 7750, loss[loss=0.2507, simple_loss=0.3068, pruned_loss=0.0973, over 11537.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.285, pruned_loss=0.05671, over 3023117.21 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:42:33,832 INFO [optim.py:368] (3/8) 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 23:42:39,568 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-05-02 23:43:12,802 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:43:38,208 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 23:43:41,535 INFO [train.py:904] (3/8) Epoch 30, batch 7800, loss[loss=0.1924, simple_loss=0.2828, pruned_loss=0.05106, over 17259.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2853, pruned_loss=0.05648, over 3043302.49 frames. ], batch size: 52, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:44:06,720 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1200, 2.3775, 2.5190, 1.9815, 2.6806, 2.7496, 2.4070, 2.3711], device='cuda:3'), covar=tensor([0.0725, 0.0311, 0.0303, 0.0962, 0.0159, 0.0371, 0.0517, 0.0466], device='cuda:3'), in_proj_covar=tensor([0.0151, 0.0112, 0.0104, 0.0140, 0.0088, 0.0134, 0.0132, 0.0132], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-02 23:44:55,045 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302200.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:45:00,017 INFO [train.py:904] (3/8) Epoch 30, batch 7850, loss[loss=0.2147, simple_loss=0.3131, pruned_loss=0.05812, over 16838.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2865, pruned_loss=0.05636, over 3042514.18 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:45:06,745 INFO [optim.py:368] (3/8) 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 23:45:25,663 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3698, 3.4895, 3.6229, 3.5986, 3.6161, 3.4395, 3.4685, 3.4958], device='cuda:3'), covar=tensor([0.0456, 0.0756, 0.0540, 0.0491, 0.0572, 0.0635, 0.0926, 0.0647], device='cuda:3'), in_proj_covar=tensor([0.0439, 0.0500, 0.0479, 0.0444, 0.0522, 0.0507, 0.0583, 0.0408], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-02 23:46:15,231 INFO [train.py:904] (3/8) Epoch 30, batch 7900, loss[loss=0.1912, simple_loss=0.2823, pruned_loss=0.05007, over 16755.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2854, pruned_loss=0.05545, over 3058159.33 frames. ], batch size: 89, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:46:16,966 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302255.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:46:25,208 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:46:25,358 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:22,383 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 7950, loss[loss=0.1756, simple_loss=0.2612, pruned_loss=0.04501, over 17035.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2856, pruned_loss=0.0558, over 3060777.19 frames. ], batch size: 53, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:47:41,024 INFO [optim.py:368] (3/8) 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 23:48:05,040 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.0500, 5.0346, 4.7951, 4.0990, 4.9598, 1.8790, 4.6840, 4.3904], device='cuda:3'), covar=tensor([0.0119, 0.0099, 0.0220, 0.0428, 0.0096, 0.2965, 0.0150, 0.0324], device='cuda:3'), in_proj_covar=tensor([0.0184, 0.0175, 0.0215, 0.0186, 0.0192, 0.0219, 0.0203, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:48:48,631 INFO [train.py:904] (3/8) Epoch 30, batch 8000, loss[loss=0.2563, simple_loss=0.3173, pruned_loss=0.09772, over 11164.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2864, pruned_loss=0.0568, over 3033443.12 frames. ], batch size: 250, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:48:54,495 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:49:01,919 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.1742, 5.1848, 4.8697, 4.1720, 5.0719, 1.8285, 4.7977, 4.5460], device='cuda:3'), covar=tensor([0.0097, 0.0095, 0.0227, 0.0451, 0.0100, 0.2972, 0.0131, 0.0312], device='cuda:3'), in_proj_covar=tensor([0.0185, 0.0175, 0.0215, 0.0186, 0.0192, 0.0219, 0.0203, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:49:21,391 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.5357, 1.8070, 2.2268, 2.4947, 2.4932, 2.8956, 1.9966, 2.8535], device='cuda:3'), covar=tensor([0.0293, 0.0645, 0.0397, 0.0419, 0.0443, 0.0237, 0.0630, 0.0202], device='cuda:3'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0194, 0.0212, 0.0169, 0.0205, 0.0170], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:3') 2023-05-02 23:49:57,498 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302399.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:50:03,889 INFO [train.py:904] (3/8) Epoch 30, batch 8050, loss[loss=0.188, simple_loss=0.2776, pruned_loss=0.04924, over 16415.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2866, pruned_loss=0.05642, over 3049303.60 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:50:11,660 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.982e+02 3.316e+02 3.990e+02 1.007e+03, threshold=6.633e+02, percent-clipped=4.0 2023-05-02 23:51:10,083 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302447.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:51:21,040 INFO [train.py:904] (3/8) Epoch 30, batch 8100, loss[loss=0.179, simple_loss=0.2771, pruned_loss=0.04042, over 16792.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2859, pruned_loss=0.05552, over 3072237.57 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:51:56,200 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:52:35,993 INFO [train.py:904] (3/8) Epoch 30, batch 8150, loss[loss=0.1711, simple_loss=0.2569, pruned_loss=0.04264, over 16666.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.284, pruned_loss=0.05487, over 3077085.08 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:52:40,390 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3989, 2.7077, 2.3150, 2.4744, 3.0338, 2.6958, 2.9504, 3.1971], device='cuda:3'), covar=tensor([0.0184, 0.0496, 0.0588, 0.0505, 0.0306, 0.0423, 0.0285, 0.0311], device='cuda:3'), in_proj_covar=tensor([0.0233, 0.0244, 0.0234, 0.0235, 0.0246, 0.0242, 0.0242, 0.0245], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-02 23:52:43,485 INFO [optim.py:368] (3/8) 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 23:53:28,271 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302538.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:53:52,238 INFO [train.py:904] (3/8) Epoch 30, batch 8200, loss[loss=0.2126, simple_loss=0.2835, pruned_loss=0.07087, over 11285.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2811, pruned_loss=0.05384, over 3082038.42 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:53:53,909 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:53:55,140 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:54:03,194 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:12,502 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:13,880 INFO [train.py:904] (3/8) Epoch 30, batch 8250, loss[loss=0.1729, simple_loss=0.262, pruned_loss=0.04187, over 12197.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2797, pruned_loss=0.05084, over 3086223.86 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:55:22,071 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.554e+02 2.948e+02 3.684e+02 6.863e+02, threshold=5.896e+02, percent-clipped=1.0 2023-05-02 23:55:23,081 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:02,657 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:35,852 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 8300, loss[loss=0.1706, simple_loss=0.2713, pruned_loss=0.03498, over 16853.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2768, pruned_loss=0.04828, over 3059098.85 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:57:43,508 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302693.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:58:01,686 INFO [train.py:904] (3/8) Epoch 30, batch 8350, loss[loss=0.1988, simple_loss=0.2798, pruned_loss=0.0589, over 11948.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2772, pruned_loss=0.04691, over 3060555.95 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:58:09,434 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.283e+02 2.623e+02 3.174e+02 4.916e+02, threshold=5.246e+02, percent-clipped=0.0 2023-05-02 23:58:23,052 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 23:58:38,861 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-05-02 23:59:22,768 INFO [train.py:904] (3/8) Epoch 30, batch 8400, loss[loss=0.1729, simple_loss=0.2603, pruned_loss=0.04275, over 12230.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2746, pruned_loss=0.04479, over 3055422.52 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:44,411 INFO [train.py:904] (3/8) Epoch 30, batch 8450, loss[loss=0.1491, simple_loss=0.2423, pruned_loss=0.02789, over 12166.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2723, pruned_loss=0.043, over 3055232.60 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:52,080 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.019e+02 2.324e+02 2.808e+02 4.179e+02, threshold=4.647e+02, percent-clipped=0.0 2023-05-03 00:00:54,499 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.7428, 2.7069, 2.4142, 3.9250, 2.3168, 3.8685, 1.5825, 2.9442], device='cuda:3'), covar=tensor([0.1399, 0.0750, 0.1232, 0.0204, 0.0121, 0.0357, 0.1727, 0.0709], device='cuda:3'), in_proj_covar=tensor([0.0175, 0.0182, 0.0202, 0.0206, 0.0207, 0.0219, 0.0211, 0.0200], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-03 00:01:31,125 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302833.0, num_to_drop=1, layers_to_drop={3} 2023-05-03 00:02:04,570 INFO [train.py:904] (3/8) Epoch 30, batch 8500, loss[loss=0.1598, simple_loss=0.2536, pruned_loss=0.03295, over 16709.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2687, pruned_loss=0.04057, over 3062103.03 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:02:08,465 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302856.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:02:48,149 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302881.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:13,951 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-03 00:03:26,596 INFO [train.py:904] (3/8) Epoch 30, batch 8550, loss[loss=0.1661, simple_loss=0.2652, pruned_loss=0.0335, over 15329.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2663, pruned_loss=0.03954, over 3039789.05 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:03:27,650 INFO [zipformer.py:625] (3/8) 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] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.175e+02 2.493e+02 2.958e+02 5.835e+02, threshold=4.986e+02, percent-clipped=2.0 2023-05-03 00:04:45,057 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302942.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:04,930 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 8600, loss[loss=0.1756, simple_loss=0.2743, pruned_loss=0.03847, over 16155.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2665, pruned_loss=0.03881, over 3041352.67 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:05:50,163 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-03 00:05:54,771 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-03 00:06:09,545 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-03 00:06:16,066 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302988.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:06:39,316 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303000.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:06:46,602 INFO [train.py:904] (3/8) Epoch 30, batch 8650, loss[loss=0.1724, simple_loss=0.2627, pruned_loss=0.04102, over 12260.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.265, pruned_loss=0.0374, over 3062887.67 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:58,832 INFO [optim.py:368] (3/8) 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 00:08:10,037 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4712, 4.4827, 4.2891, 3.4244, 4.3693, 1.7885, 4.1278, 3.9867], device='cuda:3'), covar=tensor([0.0130, 0.0131, 0.0213, 0.0298, 0.0132, 0.2917, 0.0160, 0.0285], device='cuda:3'), in_proj_covar=tensor([0.0181, 0.0173, 0.0211, 0.0182, 0.0188, 0.0216, 0.0199, 0.0175], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:08:31,107 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303052.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:08:34,580 INFO [train.py:904] (3/8) Epoch 30, batch 8700, loss[loss=0.1646, simple_loss=0.2548, pruned_loss=0.03719, over 16303.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2617, pruned_loss=0.03589, over 3066124.53 frames. ], batch size: 35, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:08:40,759 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-03 00:10:14,425 INFO [train.py:904] (3/8) Epoch 30, batch 8750, loss[loss=0.1807, simple_loss=0.2847, pruned_loss=0.03838, over 16850.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2618, pruned_loss=0.03579, over 3058495.60 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:24,544 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8348, 3.5012, 4.0080, 2.0555, 4.1386, 4.2043, 3.1890, 3.2105], device='cuda:3'), covar=tensor([0.0700, 0.0274, 0.0184, 0.1129, 0.0067, 0.0152, 0.0396, 0.0428], device='cuda:3'), in_proj_covar=tensor([0.0146, 0.0109, 0.0101, 0.0136, 0.0085, 0.0131, 0.0129, 0.0128], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-03 00:10:25,189 INFO [optim.py:368] (3/8) 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 00:10:38,668 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303113.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:11:13,904 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9247, 2.1687, 2.2682, 3.3524, 2.0881, 2.3749, 2.2968, 2.2433], device='cuda:3'), covar=tensor([0.1575, 0.3913, 0.3450, 0.0817, 0.4916, 0.3021, 0.3784, 0.4032], device='cuda:3'), in_proj_covar=tensor([0.0420, 0.0473, 0.0386, 0.0333, 0.0442, 0.0542, 0.0446, 0.0552], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:11:23,648 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303133.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:12:06,332 INFO [train.py:904] (3/8) Epoch 30, batch 8800, loss[loss=0.1762, simple_loss=0.2653, pruned_loss=0.04351, over 12446.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2606, pruned_loss=0.03515, over 3054103.08 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:12:21,001 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4408, 3.6384, 3.6935, 2.5832, 3.3338, 3.7462, 3.5008, 2.0691], device='cuda:3'), covar=tensor([0.0511, 0.0070, 0.0062, 0.0402, 0.0131, 0.0090, 0.0099, 0.0536], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0134, 0.0103, 0.0115, 0.0098, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-03 00:13:02,844 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303181.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:13:09,907 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1544, 4.0277, 4.2133, 4.3007, 4.4093, 3.9668, 4.4177, 4.4634], device='cuda:3'), covar=tensor([0.1551, 0.1081, 0.1263, 0.0636, 0.0527, 0.1522, 0.0579, 0.0613], device='cuda:3'), in_proj_covar=tensor([0.0659, 0.0799, 0.0922, 0.0824, 0.0623, 0.0645, 0.0682, 0.0789], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:13:50,932 INFO [train.py:904] (3/8) Epoch 30, batch 8850, loss[loss=0.1571, simple_loss=0.265, pruned_loss=0.02466, over 15479.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2636, pruned_loss=0.03453, over 3060030.22 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:14:00,708 INFO [optim.py:368] (3/8) 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 00:15:03,361 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303237.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:15:19,498 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.3755, 3.4986, 2.0566, 3.7967, 2.5681, 3.7742, 2.1610, 2.7697], device='cuda:3'), covar=tensor([0.0337, 0.0380, 0.1635, 0.0293, 0.0847, 0.0535, 0.1664, 0.0818], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0178, 0.0192, 0.0169, 0.0177, 0.0216, 0.0201, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-03 00:15:38,683 INFO [train.py:904] (3/8) Epoch 30, batch 8900, loss[loss=0.1722, simple_loss=0.2811, pruned_loss=0.03163, over 16861.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2652, pruned_loss=0.03447, over 3072304.72 frames. ], batch size: 96, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:15:49,470 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303258.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:17:03,840 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303288.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:17:43,147 INFO [train.py:904] (3/8) Epoch 30, batch 8950, loss[loss=0.1442, simple_loss=0.2402, pruned_loss=0.0241, over 16242.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2642, pruned_loss=0.03454, over 3072607.40 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:17:53,309 INFO [optim.py:368] (3/8) 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 00:18:16,126 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303319.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:18:16,251 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0390, 2.2081, 2.2955, 3.5976, 2.1569, 2.4911, 2.3151, 2.3199], device='cuda:3'), covar=tensor([0.1471, 0.3904, 0.3403, 0.0687, 0.4370, 0.2895, 0.4142, 0.3630], device='cuda:3'), in_proj_covar=tensor([0.0421, 0.0474, 0.0387, 0.0334, 0.0442, 0.0543, 0.0448, 0.0553], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:18:53,964 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303336.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:19:32,194 INFO [train.py:904] (3/8) Epoch 30, batch 9000, loss[loss=0.1364, simple_loss=0.2191, pruned_loss=0.02691, over 17071.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2602, pruned_loss=0.0331, over 3065907.60 frames. ], batch size: 53, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:19:32,194 INFO [train.py:929] (3/8) Computing validation loss 2023-05-03 00:19:42,083 INFO [train.py:938] (3/8) Epoch 30, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01984, over 944034.00 frames. 2023-05-03 00:19:42,084 INFO [train.py:939] (3/8) Maximum memory allocated so far is 17977MB 2023-05-03 00:21:28,479 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303403.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:29,166 INFO [train.py:904] (3/8) Epoch 30, batch 9050, loss[loss=0.1639, simple_loss=0.2558, pruned_loss=0.03604, over 16993.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2604, pruned_loss=0.0334, over 3066364.88 frames. ], batch size: 55, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:21:38,010 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303408.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:39,555 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 2.159e+02 2.466e+02 3.038e+02 8.070e+02, threshold=4.932e+02, percent-clipped=3.0 2023-05-03 00:22:30,393 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8301, 4.8272, 4.5622, 4.0172, 4.6968, 1.9353, 4.4546, 4.3638], device='cuda:3'), covar=tensor([0.0128, 0.0112, 0.0249, 0.0333, 0.0133, 0.2702, 0.0161, 0.0279], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0172, 0.0211, 0.0181, 0.0188, 0.0216, 0.0199, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:23:00,687 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.7395, 2.9796, 3.4128, 2.0731, 2.9611, 2.1413, 3.2621, 3.2470], device='cuda:3'), covar=tensor([0.0274, 0.0948, 0.0507, 0.2154, 0.0788, 0.1054, 0.0647, 0.1074], device='cuda:3'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:3') 2023-05-03 00:23:07,838 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.6469, 2.6277, 1.9534, 2.7807, 2.1585, 2.8214, 2.1881, 2.4184], device='cuda:3'), covar=tensor([0.0345, 0.0382, 0.1248, 0.0279, 0.0704, 0.0491, 0.1280, 0.0619], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0178, 0.0192, 0.0168, 0.0176, 0.0215, 0.0201, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-03 00:23:13,931 INFO [train.py:904] (3/8) Epoch 30, batch 9100, loss[loss=0.1457, simple_loss=0.2519, pruned_loss=0.01976, over 17013.00 frames. ], tot_loss[loss=0.164, simple_loss=0.26, pruned_loss=0.034, over 3063737.94 frames. ], batch size: 97, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:23:33,640 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303464.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:25:12,849 INFO [train.py:904] (3/8) Epoch 30, batch 9150, loss[loss=0.1482, simple_loss=0.2436, pruned_loss=0.02645, over 16625.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.261, pruned_loss=0.03384, over 3077618.03 frames. ], batch size: 57, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:25:25,336 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.268e+02 2.639e+02 3.152e+02 6.290e+02, threshold=5.277e+02, percent-clipped=4.0 2023-05-03 00:26:23,059 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303537.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:26:54,946 INFO [train.py:904] (3/8) Epoch 30, batch 9200, loss[loss=0.1821, simple_loss=0.2808, pruned_loss=0.04176, over 16236.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2574, pruned_loss=0.03318, over 3093168.47 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:27:08,051 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1600, 3.2096, 2.0280, 3.5035, 2.4581, 3.4724, 2.1438, 2.6723], device='cuda:3'), covar=tensor([0.0371, 0.0449, 0.1641, 0.0334, 0.0883, 0.0631, 0.1626, 0.0777], device='cuda:3'), in_proj_covar=tensor([0.0174, 0.0178, 0.0192, 0.0168, 0.0176, 0.0215, 0.0201, 0.0179], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-03 00:27:31,975 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303574.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:27:52,354 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303585.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:28:27,178 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9222, 2.1905, 2.4627, 3.1866, 2.2548, 2.3734, 2.3885, 2.3144], device='cuda:3'), covar=tensor([0.1439, 0.3899, 0.3196, 0.0891, 0.4936, 0.2895, 0.3839, 0.3997], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0471, 0.0386, 0.0332, 0.0441, 0.0540, 0.0445, 0.0551], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:28:29,162 INFO [train.py:904] (3/8) Epoch 30, batch 9250, loss[loss=0.1612, simple_loss=0.2578, pruned_loss=0.03235, over 16164.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2573, pruned_loss=0.03354, over 3066822.28 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:28:41,899 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.148e+02 2.583e+02 3.115e+02 7.022e+02, threshold=5.165e+02, percent-clipped=1.0 2023-05-03 00:28:49,554 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303614.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:29:38,999 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.1831, 4.4272, 4.4747, 3.3907, 3.8246, 4.4307, 3.9718, 2.7188], device='cuda:3'), covar=tensor([0.0420, 0.0054, 0.0037, 0.0307, 0.0119, 0.0084, 0.0070, 0.0445], device='cuda:3'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0134, 0.0103, 0.0114, 0.0097, 0.0130], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-03 00:29:39,023 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 9300, loss[loss=0.1507, simple_loss=0.2498, pruned_loss=0.02578, over 15145.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2555, pruned_loss=0.03301, over 3063380.34 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:30:25,040 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.9178, 4.1242, 4.1974, 3.0603, 3.6748, 4.1693, 3.7491, 2.4629], device='cuda:3'), covar=tensor([0.0437, 0.0058, 0.0045, 0.0341, 0.0128, 0.0106, 0.0091, 0.0462], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0134, 0.0102, 0.0114, 0.0097, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-03 00:31:05,522 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.1496, 4.1642, 4.4593, 4.4379, 4.4405, 4.2266, 4.2089, 4.2180], device='cuda:3'), covar=tensor([0.0379, 0.0790, 0.0512, 0.0542, 0.0627, 0.0525, 0.0892, 0.0475], device='cuda:3'), in_proj_covar=tensor([0.0432, 0.0491, 0.0471, 0.0437, 0.0516, 0.0498, 0.0570, 0.0402], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-03 00:31:53,162 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4420, 3.3918, 3.4798, 3.5485, 3.5928, 3.3712, 3.5668, 3.6499], device='cuda:3'), covar=tensor([0.1330, 0.0908, 0.1097, 0.0717, 0.0608, 0.1917, 0.0929, 0.0781], device='cuda:3'), in_proj_covar=tensor([0.0657, 0.0796, 0.0919, 0.0824, 0.0621, 0.0645, 0.0682, 0.0789], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:32:05,574 INFO [train.py:904] (3/8) Epoch 30, batch 9350, loss[loss=0.1751, simple_loss=0.2659, pruned_loss=0.04212, over 16652.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2554, pruned_loss=0.03312, over 3052858.84 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:32:13,822 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303708.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:32:16,691 INFO [optim.py:368] (3/8) 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] (3/8) Epoch 30, batch 9400, loss[loss=0.1701, simple_loss=0.2749, pruned_loss=0.03272, over 16312.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2555, pruned_loss=0.03274, over 3047636.59 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:33:50,775 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303756.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:33:56,170 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303759.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:35:14,448 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.8446, 3.8239, 4.0137, 3.7530, 3.9555, 4.3391, 3.9882, 3.7266], device='cuda:3'), covar=tensor([0.2131, 0.2367, 0.2092, 0.2447, 0.2599, 0.1519, 0.1612, 0.2553], device='cuda:3'), in_proj_covar=tensor([0.0419, 0.0623, 0.0699, 0.0507, 0.0673, 0.0719, 0.0540, 0.0677], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-03 00:35:25,717 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.4713, 3.1959, 3.4465, 1.8620, 3.5768, 3.6258, 2.9850, 2.9138], device='cuda:3'), covar=tensor([0.0722, 0.0301, 0.0226, 0.1259, 0.0102, 0.0206, 0.0424, 0.0455], device='cuda:3'), in_proj_covar=tensor([0.0144, 0.0107, 0.0098, 0.0134, 0.0084, 0.0128, 0.0127, 0.0127], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:3') 2023-05-03 00:35:26,329 INFO [train.py:904] (3/8) Epoch 30, batch 9450, loss[loss=0.1517, simple_loss=0.2484, pruned_loss=0.02751, over 16321.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2573, pruned_loss=0.03296, over 3057754.02 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:35:29,346 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.8896, 2.1686, 2.4010, 3.1865, 2.2299, 2.3790, 2.3408, 2.2721], device='cuda:3'), covar=tensor([0.1474, 0.3974, 0.2979, 0.0783, 0.4548, 0.2837, 0.3771, 0.4131], device='cuda:3'), in_proj_covar=tensor([0.0418, 0.0471, 0.0386, 0.0332, 0.0441, 0.0540, 0.0445, 0.0550], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:35:36,996 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 2.120e+02 2.504e+02 3.073e+02 7.133e+02, threshold=5.009e+02, percent-clipped=4.0 2023-05-03 00:35:38,587 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303810.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:35:59,035 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-03 00:37:08,345 INFO [train.py:904] (3/8) Epoch 30, batch 9500, loss[loss=0.1535, simple_loss=0.2517, pruned_loss=0.0276, over 16901.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2567, pruned_loss=0.03282, over 3074090.64 frames. ], batch size: 102, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:37:43,361 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303871.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:36,416 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303896.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:54,494 INFO [train.py:904] (3/8) Epoch 30, batch 9550, loss[loss=0.1659, simple_loss=0.2603, pruned_loss=0.03578, over 16792.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2571, pruned_loss=0.03326, over 3087709.36 frames. ], batch size: 76, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:39:08,518 INFO [optim.py:368] (3/8) 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] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303914.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:39:49,602 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303930.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:36,090 INFO [train.py:904] (3/8) Epoch 30, batch 9600, loss[loss=0.1765, simple_loss=0.2716, pruned_loss=0.04072, over 16761.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.258, pruned_loss=0.03381, over 3059200.88 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:40:43,427 INFO [zipformer.py:625] (3/8) 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] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303962.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:41:16,223 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.0633, 4.0633, 3.9678, 3.1646, 3.9913, 1.8447, 3.8183, 3.5383], device='cuda:3'), covar=tensor([0.0125, 0.0107, 0.0202, 0.0264, 0.0126, 0.2956, 0.0137, 0.0339], device='cuda:3'), in_proj_covar=tensor([0.0182, 0.0173, 0.0211, 0.0181, 0.0188, 0.0217, 0.0199, 0.0176], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:41:16,247 INFO [zipformer.py:625] (3/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303974.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:42:24,226 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([5.5292, 5.8875, 5.6173, 5.6845, 5.3188, 5.2927, 5.2704, 5.9834], device='cuda:3'), covar=tensor([0.1335, 0.0947, 0.1075, 0.0786, 0.0793, 0.0687, 0.1320, 0.0889], device='cuda:3'), in_proj_covar=tensor([0.0707, 0.0850, 0.0699, 0.0663, 0.0542, 0.0542, 0.0712, 0.0668], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:42:31,192 INFO [train.py:904] (3/8) Epoch 30, batch 9650, loss[loss=0.1681, simple_loss=0.2636, pruned_loss=0.03625, over 16809.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2598, pruned_loss=0.03427, over 3059239.33 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:42:48,097 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.176e+02 2.625e+02 3.264e+02 9.247e+02, threshold=5.250e+02, percent-clipped=2.0 2023-05-03 00:43:41,208 INFO [zipformer.py:625] (3/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304035.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:44:20,835 INFO [train.py:904] (3/8) Epoch 30, batch 9700, loss[loss=0.1727, simple_loss=0.2667, pruned_loss=0.03931, over 15324.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2588, pruned_loss=0.03414, over 3067720.96 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:44:30,039 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304059.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:46:04,097 INFO [train.py:904] (3/8) Epoch 30, batch 9750, loss[loss=0.1622, simple_loss=0.255, pruned_loss=0.03469, over 12421.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2577, pruned_loss=0.03437, over 3042211.98 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:46:09,339 INFO [zipformer.py:625] (3/8) 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] (3/8) 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] (3/8) Epoch 30, batch 9800, loss[loss=0.1531, simple_loss=0.2639, pruned_loss=0.02118, over 15421.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2578, pruned_loss=0.03368, over 3052036.87 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:48:04,642 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304166.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:48:38,220 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.1526, 3.5417, 3.5678, 2.3424, 3.1560, 3.5644, 3.3814, 2.1608], device='cuda:3'), covar=tensor([0.0590, 0.0064, 0.0066, 0.0454, 0.0154, 0.0106, 0.0100, 0.0487], device='cuda:3'), in_proj_covar=tensor([0.0137, 0.0090, 0.0091, 0.0134, 0.0102, 0.0114, 0.0097, 0.0129], device='cuda:3'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:3') 2023-05-03 00:48:53,964 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.4826, 3.5059, 3.7118, 3.6906, 3.7226, 3.5473, 3.5702, 3.6005], device='cuda:3'), covar=tensor([0.0426, 0.0842, 0.0538, 0.0520, 0.0524, 0.0629, 0.0824, 0.0497], device='cuda:3'), in_proj_covar=tensor([0.0429, 0.0488, 0.0469, 0.0435, 0.0513, 0.0495, 0.0566, 0.0399], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:3') 2023-05-03 00:49:24,991 INFO [train.py:904] (3/8) Epoch 30, batch 9850, loss[loss=0.1615, simple_loss=0.2581, pruned_loss=0.03246, over 16688.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2591, pruned_loss=0.03336, over 3074228.40 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:49:39,349 INFO [optim.py:368] (3/8) 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 00:50:19,594 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304230.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:51:13,301 INFO [zipformer.py:625] (3/8) 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] (3/8) Epoch 30, batch 9900, loss[loss=0.1653, simple_loss=0.2676, pruned_loss=0.03153, over 15322.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2589, pruned_loss=0.033, over 3052936.80 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:52:13,220 INFO [zipformer.py:625] (3/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304278.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:52:42,637 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.4238, 4.2377, 4.4420, 4.5845, 4.7595, 4.3050, 4.7634, 4.7700], device='cuda:3'), covar=tensor([0.1826, 0.1284, 0.1608, 0.0845, 0.0629, 0.1239, 0.0726, 0.0785], device='cuda:3'), in_proj_covar=tensor([0.0650, 0.0786, 0.0904, 0.0815, 0.0613, 0.0636, 0.0675, 0.0781], device='cuda:3'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:52:56,004 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([1.9734, 1.8144, 1.6817, 1.4560, 1.9597, 1.5739, 1.5188, 1.9060], device='cuda:3'), covar=tensor([0.0226, 0.0375, 0.0488, 0.0457, 0.0273, 0.0356, 0.0219, 0.0274], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0239, 0.0229, 0.0230, 0.0241, 0.0238, 0.0233, 0.0236], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:52:59,552 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([2.0000, 1.8441, 1.7068, 1.5304, 1.9984, 1.6544, 1.5654, 1.9818], device='cuda:3'), covar=tensor([0.0231, 0.0379, 0.0451, 0.0424, 0.0272, 0.0336, 0.0206, 0.0262], device='cuda:3'), in_proj_covar=tensor([0.0225, 0.0239, 0.0229, 0.0230, 0.0241, 0.0238, 0.0233, 0.0236], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:3') 2023-05-03 00:53:14,776 INFO [train.py:904] (3/8) Epoch 30, batch 9950, loss[loss=0.1823, simple_loss=0.2755, pruned_loss=0.04453, over 16828.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.261, pruned_loss=0.03341, over 3047337.71 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:53:31,713 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 2.067e+02 2.374e+02 2.801e+02 5.953e+02, threshold=4.748e+02, percent-clipped=1.0 2023-05-03 00:54:21,392 INFO [zipformer.py:625] (3/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304330.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:54:27,930 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([4.8575, 5.0714, 5.2333, 5.0645, 5.1124, 5.6415, 5.1736, 4.9173], device='cuda:3'), covar=tensor([0.0987, 0.1754, 0.2168, 0.1814, 0.2064, 0.0845, 0.1476, 0.2104], device='cuda:3'), in_proj_covar=tensor([0.0413, 0.0617, 0.0693, 0.0502, 0.0668, 0.0712, 0.0535, 0.0671], device='cuda:3'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:3') 2023-05-03 00:55:15,633 INFO [train.py:904] (3/8) Epoch 30, batch 10000, loss[loss=0.1658, simple_loss=0.2561, pruned_loss=0.03772, over 12860.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2598, pruned_loss=0.03301, over 3069286.14 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:56:00,064 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-05-03 00:56:57,637 INFO [train.py:904] (3/8) Epoch 30, batch 10050, loss[loss=0.1888, simple_loss=0.2836, pruned_loss=0.04701, over 16346.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2602, pruned_loss=0.03294, over 3077014.60 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:57:10,561 INFO [optim.py:368] (3/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.064e+02 2.395e+02 2.910e+02 6.626e+02, threshold=4.790e+02, percent-clipped=3.0 2023-05-03 00:57:13,609 INFO [zipformer.py:1454] (3/8) attn_weights_entropy = tensor([3.0565, 3.0490, 1.9560, 3.2904, 2.2960, 3.3355, 2.1177, 2.5710], device='cuda:3'), covar=tensor([0.0404, 0.0451, 0.1641, 0.0303, 0.0944, 0.0552, 0.1617, 0.0853], device='cuda:3'), in_proj_covar=tensor([0.0172, 0.0176, 0.0190, 0.0167, 0.0175, 0.0212, 0.0199, 0.0178], device='cuda:3'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:3') 2023-05-03 00:57:17,588 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-03 00:57:54,740 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-03 00:58:04,534 INFO [scaling.py:679] (3/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-03 00:58:34,282 INFO [train.py:904] (3/8) Epoch 30, batch 10100, loss[loss=0.1495, simple_loss=0.2401, pruned_loss=0.02945, over 16253.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2598, pruned_loss=0.03263, over 3069283.29 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:58:57,144 INFO [zipformer.py:625] (3/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304466.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:59:20,345 INFO [scaling.py:679] (3/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-03 00:59:55,828 INFO [train.py:1169] (3/8) Done!